Pleiotropic functions of Hox genes revealed by mutants for genes in Hox seventh and ninth paralogous groups

The mammalian Hox complex consists of 39 genes that are believed to control the regionalization of the embryo along its major body axes. To investigate the in vivo functions of Hox genes, mice individually mutant for a subset of Hox genes were generated by gene-targeting. Mice with mutations in genes on separate chromosomes were mated to generate multiple mutants to reveal redundant functions not apparent in individual mutants. The analysis of Hoxb9 mutants and Hoxa9/Hoxb9 double mutants demonstrates a role of Hoxb9 in the patterning of the thoracic region and a synergistic and quantitative interaction between Hoxa9 and Hoxb9. Although Hoxa7 mutants do not have any detectable phenotype, the functions of Hoxa7 in the patterning of the upper thoracic region became apparent in Hoxa7/Hoxb7 mutants, in which the effects of the disruption of Hoxb7 on the first two ribs were greatly exaggerated by the addition of the Hoxa7 mutant allele(s). The similarity of the defects found in the Hoxa7/Hoxb7 and Hoxa9/Hoxb9 mutants evokes a modification of the current hypothesis explaining the defects in the axial column in Hox mutants. Limb defects were observed when a Hoxd9 mutation was introduced into the Hoxa9/Hoxb9 mutants. Many triple mutant genotypes have overt forelimb defects caused mainly by a series of skeletal defects in the limb. In particular, the humerus was greatly reduced in length and altered in shape. Hox genes appear to confer positional cues to different portions of the vertebrate limb, as well as to regulate the growth of cells forming different structures of the limbs. Of particular interest, hypoplasia of the mammary gland has been observed in pregnant and lactating females with multiple mutations in Hoxa9, Hoxb9 and Hoxd9. This broadens our concept of Hox gene function from patterning of the developing embryo to include indispensable roles in adults.

PLEIOTROPIC FUNCTIONS OF Hox GENES REVEALED BY MUTANTS FOR GENES IN Hox SEVENTH AND NINTH PARALOGOUS GROUPS By Feng Chen A dissertation submitted to the faculty of The University of Utah in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Human Genetics The University of Utah August 1998 Copyright © Feng Chen 1998 All Rights Reser-led THE CSIVERSITY OF UTAH GRADUATE SCHOOL SUPERVISORY COMMITTEE APPROVAL of a dissertation submitted by Feng Chen This dissertation has been read by each member of the following supervisory committee and by majority vote has been found to be satisfactory. March 18,1998 March 18, 1998 March 18, 1998 March 18, 1998 <.,,/ Ie' .J If t ~ March 18, 1998 Anthea Letsou Shigeru Sakonju Carl S. Thummel THE UNIVERSITY OF UTAH GRADUATE SCHOOL FINAL READING APPROVAL To the Graduate Council of the University of Utah: I have read the dissertation of Feng Chen in its fmal form and have found that (1) its format, citations, and bibliographic style are consistent and acceptable; (2) its illustrative materials including figures, tables, and charts are iIi place; and (3) the fmal manuscript is satisfactory to the supervisory committee and is ready for submission to The Graduate School. March 31, 1998 Date Mario R. Cap , Supervisory Committ Approved for the Major Department Cbair/Dean Approved for the Graduate Council Ann W. Hart Dean of The Graduate School ABSTRACT The mammalian Hox complex consists of 39 genes that are believed to control the regionalization of the embryo along its major body axes. To investigate the in vivo functions of Hox genes, mice individually mutant for a subset of Hox genes were generated by gene-targeting. Mice with mutations in genes on separate chromosomes were mated to generate mUltiple mutants to reveal redundant functions not apparent in individual mutants. The analysis of Hoxb9 mutants and Hoxa91Hoxb9 double mutants demonstrates a role of Hoxb9 in the patterning of the thoracic region and a synergistic and quantitative interaction between Hoxa9 and Hoxb9. Although Hoxa7 mutants do not have any detectable phenotype, the functions of Hoxa7 in the patterning of the upper thoracic region became apparent in Hoxa71Hoxb7 mutants, in which the effects of the disruption of Hoxb7 on the first two ribs were greatly exaggerated by the addition of the Hoxa7 mutant allele(s). The similarity of the defects found in the Hoxa71Hoxb7 and Hoxa91Hoxb9 mutants evokes a modification of the current hypothesis explaining the defects in the axial column in Hox mutants. Limb defects were observed when a Hoxd9 mutation was introduced into the Hoxa91 Hoxb9 mutants. Many triple mutant genotypes have overt forelimb defects caused mainly by a series of skeletal defects in the limb. In particular, the humerus was greatly reduced in length and altered in shape. Hox genes appear to confer positional cues to different portions of the vertebrate limb, as well as to regulate the growth of cells forming different structures of the limbs. Of particular interest, hypoplasia of the mammary gland has been observed in pregnant and lactating females with multiple mutations in Hoxa9, Hoxb9 and Hoxd9. This broadens our concept of Hox gene function from patterning of the developing embryo to include indispensable roles in adults. TABLE OF CONTENTS ABSTRACT ...................................................................................... iv LIST OF FIGURES ................................................................. ......... Vll LIST OF TABLES ........................................................................ IX ACKNOWLEDGMENTS ..................................................... ............ X CHAPTER 1. INTRODUCTION ................................................................................... 1 2. TARGETED MUTATIONS IN Hoxa9 AND Hoxb9 REVEAL SYNERGISTIC INTERACTIONS ......................................... 9 Abstract ................................................................... 10 Introduction ............................................................... 10 Materials and Methods .................................................. 12 Results ..................................................................... 14 Discussion ................................................................. 16 Acknowledgments .............................. ....... ...... . . . . . . .. ....... .. 19 References ............................................................... .... 19 3. Hoxa9, Hoxb9, AND Hoxd9 FUNCTION IN THE MAMMARY GLAND AND THE PATTERNING OF THE AXIAL AND APPENDICULAR SKELETON .. ...... ... ........ .... ...... ................. ..... .............. ...... ........ .......... .... 21 Abstract ................................................................... 22 Introduction ............................................................... 22 Materials and Methods ................................................... 24 Results ..................................................................... 26 Discussion ............ '" .................................................. 47 Acknowledgments ............................... ...... ....... . . . . . . . ......... 50 4. Hoxa7 AND Hoxb7HAVE REDUNDANT FUNCTIONS IN THE PATTERNING OF THE UPPER THORACIC REGION .................................................................................................... 51 Abstract .................................................................... 52 Introduction ............................................................... 52 Materials and Methods ................................................... 54 Results ..................................................................... 56 Discussion ................................................................. 70 Acknowledgments ............................................................ 75 5. THE GENERATION OF Hoxa13 DEFICIENT MICE AND MICE WITH LONG-RANGE CHROMOSOMAL DELETIONS AND/OR DUPLICATIONS IN THE Hoxb COMPLEX .............................................................................. 76 Abstract .................................................................... 77 Introduction ............................................................... 77 Materials and Methods .................................... ......... . . . ... 79 Results ....................................................................... 87 Discussion ................................................................. 89 Acknowledgments ... . . . . .. . . . . . . . . . . . . . . . . . . .. ... ........... . . . .. . . ........ .. 92 6. DISCUSSION ..................................................................... 93 Hox genes and the development of axial structures ............. ... 94 Hox genes and the development of the vertebrate limbs ........... 95 Functions of Hox genes other than patterning ......... ............... 98 Relations of different Hox genes ....... .............. .. ........... .... ... . .. 10 I Hox genes and evolution ........................ ............... .... .......... 103 REFERENCES ...... ......... ... ... ............ .......... .................. ................. 104 vi LIST OF FIGURES Figure Page 1.1 The Drosophila Hom-C genes and the mouse Hox complex................. 8 2.1 Disruption of the Hoxb9 gene and analysis of the mutant genotype........ 11 2.2 Defects in the thoracic skeleton in Hoxb9-1- mice............................ 12 2.3 Sagittal sections of Hoxb9+l- and Hoxb9-1- newborn mice ......... ....... l3 2.4 The expression of Hoxb9 in wild-type embryos............................. 16 2.5 The expression pattern of Hoxb8 does not change in Hoxb9 mutants.... 17 2.6 Skeletal defects in Hoxa91Hoxb9 mutant mice............ ...................... 17 3.1 Targeted mutations in Hoxa9, Hoxb9, and Hoxd9 affect the patterning of the thoracic skeleton ................... ................................. 28 3.2 Triple homozygous mutants have overt limb defects............... ........ 32 3.2 The survival rate of newborn pups decreases dramatically in some triple mutant genotypes.............. ........................................ 36-38 3.4 Mammary gland hypoplasia in Aabbdd, aaBbdd, and aabbdd females during pregnancy and around parturition............................... 43 3.5 Expression of Hox genes in the mammary gland................................... 46 4.1 Disruption of the Hoxb7 gene... ....................... ........ ...... ........... 58 4.2 Analysis of the Hoxa7 and Hoxb7 mutant genotypes...................... 60 4.3 Skeletal preparations of Hoxa7 mutants and Hoxb7 mutants.. ............ 62 4.4 Skeletal preparations of Hoxa71 Hoxb 7 double mutants........... ..... .... 67 4.5 Histological analysis of Hoxa71Hoxb7 double mutants... ...... ..... .... 69 5.1 5.2 5.3 5.4 6.1 6.2 Disruption of the Hoxal3 gene ............................................. . Insertion of loxP sites into the Hoxb7 gene ............................... . Insertion of loxP sites into the Hoxb9 gene ................................. . Analysis of genotypes ............................................................. . Hox genes in limb morphogenesis ......................................... . Hox genes and the mammary gland ............................................ .. Vlll 82 84 86 91 97 100 LIST OF TABLES 2.1 Skeletal phenotypes in mice deficient for Hoxb9 ... ......................... . 12 2.2 Thymus weight comparison between Hoxb9 homozygous mutants with and without first and second rib fusions .................... . 15 2.3 Summary of skeletal phenotypes in Hoxa91Hoxb9 mice ................ . 18 3.1 Summary of skeletal phenotypes in Hoxa9, Hoxb9, and Hoxd9 triple mutants ....................................................................... . 27 4.1 Summary of skeletal phenotypes in Hoxa7, Hoxb7, and Hoxa71Hoxb7 mutant mice ............................................................. .. 63 6.1 Functional hierarchies of Hoxa9, Hoxb9, and Hoxd9 ................. ... . 101 ACKNOWLEDGMENTS I am grateful to my advisor, Dr. Mario R. Capecchi, for providing me a great opportunity to do my thesis work in his lab, which will undoubtedly continue to have a profound influence in my life. I deeply appreciate all the help provided by Drs. Anthea Letsou, Shigeru Sakonju, Carl S. Thummel, Gary C. Schoenwolf on my thesis committee and Dr. Raymond F. Gesteland on my prelim exam committee. The work presented in this dissertation would not be possible without help from postdoctoral fellows, graduate students, technicians, and administrative assistants in Mario's laboratory . I am thankful to China where I was born and educated through college. I thank the US and her people for providing me a great opportunity to further my education. I thank my parents, relatives and friends for continual support. I can never thank my wife, Li, enough for making my life meaningful and enjoyable with her love. I can always find happiness being with her regardless of the ups and downs of my research. CHAPTER 1 INTRODUCTION Genetic information is being generated at an increasing pace, partially as a result of the Human Genome Program. It is important and challenging to assign functional roles to nucleotide sequences in the human genome, transcribed and nontranscribed alike. Although gene functions can be studied at various levels by a variety of methods, precise in vivo functions of a gene have to be revealed in the context of the normal physiological environment inside the organism itself. Humans can not be used for many experimental purposes for obvious ethical considerations. The mouse Mus musculus has become a valuable experimental organism for at least five important reasons. First, there is a great similarity between the human genome and the mouse genome. Second, the embryogenesis and physiology of the two organisms share a great deal of similarity. Third, this animal has a high reproductive potential and requires relatively inexpensive care. Fourth, the mouse has been scientifically studied for a long time, so that a comprehensive system of knowledge on its genetics, physiology, and embryology has been established. Fifth, the invention of the transgenic technique and later the gene-targeting technique in mouse has opened up seemingly unlimited possibilities for manipulating the genome of this animal, firmly establishing its role as an excellent animal model. Our laboratory has employed the gene-targeting technique to generate an extensive collection of murine Hox gene mutants to determine the in vivo functions of these genes and to study the mechanism of development. Gene-targeting is a process in which embryonic stem (ES) cells are genetically modified by homologous recombination and used to produce mutant mice carrying such precisely engineered modifications of the genome (Capecchi, 1989). Briefly, this technique consists of the following critical steps: the first step is to alter a cloned DNA sequence of a chosen locus by standard molecular cloning technology. The modified DNA sequence, called a targeting vector, usually contains a selectable marker such as the neomycin resistance gene (ned). Most targeting vectors also have the HSVI and HSV2 thymidine kinase (tk) genes flanking the genomic sequences. The second step involves the introduction of the modified DNA into ES cells and the subsequent homologous recombination between the exogenous DNA and the endogenous chromosomal sequence. The result of the second step is ES cells carrying 2 exogenously modified sequences in the endogenous locus. These ES cells are usually enriched by a positive-negative selection scheme, taking advantage of the presence of the positive selectable marker and the tk genes in the targeting vector (Mansour et ai., 1988). In the third step, ES cells containing the desired genomic modification are microinjected into mouse blastocysts to generate germline chimeras. Finally, heterozygous mutants obtained from the germline chimeras can be interbred to generate homozygotes for the desired modification. The discovery that the bacteriophage PI Cre-ioxP site-specific recombination system works efficiently in mammalian cells further expands the already extensive applications of the gene­targeting technique. The 38 kDa Cre recombinase catalyzes site-specific DNA recombination between 34 bp repeats called ioxP. Appropriate uses of this system have already generated mice with "clean" deletions, with tissue-specific alleles, and with long-range deletions difficult to achieve by classical gene-targeting methods (Ramirez-Solis et ai., 1995). Theoretically, nearly every form of chromosomal rearrangement, naturally occurring or artificially designed, can be generated by these techniques. The improvement of these techniques has been and will continue to be important in the study of in vivo functions of Hox genes as well as other mammalian genes. Hox genes encode transcription factors of the Antennapedia homeodomain class. There are four linkage groups in the mammalian Hox complex containing at least 39 genes. The four linkage groups, designated as Hoxa, b, c and d, are located on four separate chromosomes. Individual members of the four linkage groups have been classified into 13 paralogous families with respect to DNA sequence and chromosomal location (Fig. 1.1). Interestingly, in a given cluster, a 3' Hox gene is activated prior to and in a more anterior region of the embryo than its 5' neighbor. This correlation between the expression pattern of the Hox genes and their relative chromosomal positions is called temporal and spatial colinearity (Duboule and Dolle, 1989; Capecchi, 1997). The Drosophila Hom-C genes, homo logs of the mammalian Hox genes, are transcription factors acting as master switches that can initiate genetic cascades determining the identities for each parasegment (Akam, 1987; Gehring, 1987). 3 Figure 1.1 The Drosophila Hom-C genes and the mouse Hox compex. The mouse Hox complex consists of at least 39 genes distributed on four linkage groups. These four linkage groups, designated as Hoxa, b, c, and d, locate on four separate chromosomes. Individual Hox genes have been classified into 13 paralogous groups on the basis of sequence similarity and chromosomal location. The transcription direction of all the Hox genes is from the 13th paralogous group toward the first paralogous group. 4 5 Drosophila Chromosome lab ;. Dfd Scr Antp Ubx Abd-A Abd-B ANT-C - -- _II- -• BX-C 3 Mouse ~ l. ~ I I I I Al A2 A3 A4 A5 A6 A7 A9 AIO All A13 Evxl Hoxa - - -. 6 BI B2 B3 B4 B5 B6 B7 B8 B9 B13 Hoxb - CJ 11 C4 C5 C6 C8 C9 CIO Cll CI2 C13 Hoxc - CJ l1li CJ 15 Dl D3 D4 D8 D9 DIO DII DI2 D13 Evx2 Hoxd CJ 1111 CJ -. 2 anterior 3' ~ 5' posterior Mutations in these genes can result in transfonnations of one parasegment to another (Lewis, 1978). Mutational analysis in the mouse has demonstrated that Hox genes are also master switches controlling the regionalization of the embryo along its major body axes. Mutations in the Hox genes can result in transfonnation of axial structures, malformation, deletion as well as duplication of structures in the vertebral body (Chisaka and Capecchi, 1991; Lufkin et al., 1991; Chisaka et ai., 1992; Le Mouellic et ai., 1992; Condie and Capecchi, 1993; Dolle et al., 1993; Gendron-Maguire et ai., 1993; Jeannotte et al., 1993; Ramirez-Solis et aI., 1993; Rijli et ai., 1993; Small and Potter, 1993; Davis and Capecchi, 1994; Kostic and Capecchi, 1994; Satokata et al., 1995; Suemori et al., 1995; Boulet and Capecchi, 1996; Goddard et aI., 1996; Barrow and Capecchi, 1996; Chen and Capecchi, 1997). A number of hypotheses have been formulated or adapted to explain the phenotypes observed in the mammalian Hox mutants. Among them, the Hox code model asserts that regional (segmental) identity is determined by the specific combination of Hox expression. In a given segment, the most posterior expressing Hox gene is believed to have dominant roles in the specification of segmental identity, a hypothesis called "posterior prevalence" (Duboule, 1991; Krumlauf, 1993). These hypotheses predict that loss-of-function mutations in Hox genes would lead to anterior transformation near the anterior boundary of the disrupted gene. Such anterior transformations in the axial column have been observed in Hox loss-of-function mutants, lending support to these hypotheses. Accumulating evidence, however, suggests that this is a tendency rather than a rule. The observation of posterior transfonnation, deletions, duplications and other malformations of structures in many Hox loss-of-function mutants calls for modifications of or alternatives to these hypotheses. Furthermore, defects in the limbs of Hox loss-of-function mutants generally do not resemble any form of transfonnation and seem to be more difficult to explaine by any simple models. In experiments described in this dissertation, mice individually mutant for a subset of Hox genes were generated by gene-targeting. Mice with mutations in genes on separate chromosomes were used to generate multiple mutants to reveal redundant functions not apparent in individual mutants. Mice homozygous for a disruption in Hoxb9 show defects in the development of the first and second ribs and the sternum. 6 Over half of the homozygous mutants, as well as some heterozygotes, also have an eighth rib attached to the sternum, suggesting that Hoxb9 plays a significant role in the specification of thoracic skeletal elements. Mice homozygous for a Hoxa9 disruption have an extra pair of ribs on the first lumbar vertebra. They do not have overlapping phenotypes with the Hoxb9 mutants (Fromental-Ramain et al., 1996; Peterson, Chisaka, and Capecchi, unpublished data; Chen and Capecchi, 1997). To reveal potential interactions between the paralogous Hox genes Hoxa9 and Hoxb9, mice heterozygous for both mutations were intercrossed. Mice homozygous for both mutations show more severe phenotypes than predicted by the addition of the individual mutant phenotypes. Both the penetrance and the expressivity of the rib and sternal defects are increased, suggesting synergistic interactions between these genes. Although Hoxa7 mutants do not have detectable defects, the functions of Hoxa7 in the patterning of the upper thoracic region were revealed in Hoxa71Hoxb7 mutants, in which the effects of the disruption of Hoxb 7 on the first two ribs were greatly exacerbated by the addition of the Hoxa7 mutant allele(s). The similarity of the defects found in the Hoxa71Hoxb7 and Hoxa9lHoxb9 mutants evokes a modification of the current hypothesis explaining the defects in the axial column in Hox mutants. Mice carrying all possible combinations of disrupted alleles for Hoxa9, Hoxb9, and Hoxd9 have been generated by appropriate matings starting from the three individual mutants. Triple homozygous mutants and mutants with certain combinations of mutant alleles have extensive rib fusions and overt defects in their limbs. These overt forelimb defects were caused mainly by a series of skeletal defects in the limb. In particular, the humerus was greatly reduced in length and altered in shape. Hox genes appear to confer positional cues to different portions of the vertebrate limb, as well as to regulate the growth of cells forming different structures of the limbs. Of particular interest, females with multiple mutant alleles in these three genes can not raise their own pups. Pups born to these females died of apparent malnutrition, which can be rescued by fostering. The mammary glands of these females were found to be underdeveloped during pregnancy and around parturition. The demonstration of the expression of these Hox genes in the fetal mammary gland primordia and in adult mammary glands further indicates direct functional roles of 7 these Hox genes in controlling the development and functions of the mammary gland. This broadens our concept of Hox gene function from patterning of the embryo to including indispensable roles in adults. Finally, the functional importance of a group of paralogous genes, as manifested by the relative contribution to the mutant phenotypes, varies in different body regions, indicating that these genes form distinctive hierarchies in different regIOn. 8 CHAPTER 2 TARGETING MUTATIONS IN Hoxa9 AND Hoxb9 REVEAL SYNERGISTIC INTERACTIONS The following chapter is a reprint of an article coauthored by Mario R. Capecchi and me. This article was originally published in Developmental Biology, Volume 120, pages186-196, January, 1997 (Copyright 1997 by Academic Press). It is reprinted here with the permission of the coauthor and the Academic Press. LJEVELOPMENTAL BIOLOGY 181, 1~6~1~611~97J ARTICLE NO. D8968440 Targeted Mutations in Hoxa-9 and Hoxb-9 Reveal Synergistic Interactions Feng Chen and Mario R. CapecchP Department of Human Genetics, Howard Hughes Medical Institute, University of Utah School of Medicine, Salt Lake City, Utah 84112 10 Mice were generated with a targeted disruption of the homeobox-containing gene horb-9. Mice homozygous (or this mutation show defects in the development of the flrst and second ribs. In most cases the flrst and second ribs are fused near the point at which the flrst and second pairs of ribs normally attach to the sternum. Abnormalities of the sternum accompany the rib fusions. These include abnormal attachment of the ribs to the sternum, a reduction in the number o( intercostal segments of the sternum, and abnormal growth of the intercostal segments. Over half of the homozygous mutants, as wen as some heterozygotes, also have an eighth rib attached to the sternum. These results show that horb.9 plays a significant role in the specification of thoradc skeletal elements. To reveal potential interactions between the paralogous Bor genes hortl·9 and horb.9, mice heterozygous for both mutations were intererossed. Mice homozygous for both mutations show more severe phenotypes than predicted by the addition of the individual mutant phenotypes. Both the penetrance and the expressivity of the rib and sternal defects are increased, suggestiog synergistic interactions between these genes. In particu1ar, the sternum defects are greatly exacerbated. Interestingly, the defects in horb.9 and boXtl·91 hoxb·9 mutant mice are concentrated along the axial column at points of transition between vertebral types. 0199' Ac:ad.mic Press INTRODUCTION Hox genes encode transcription factors belonging to the Antennapedia homeodomain class. The mammalian HOJ( complex contains 39 genes distributed on 4 linkage groups designated as HoxA, B, C, and D. This organization is be­lieved to have arisen early in vertebrate phylogeny by qua­druplication of an ancestral complex common to verte­brates and invertebrates (Pendleton et al., 19931 Holland and Garda-Fernandez, 19961. Based on DNA sequence and the position of the genes on their respective chromosomes, individual members of the 4 linkage groups have been clas­sified into 13 paralogous families. Members of a paralogous family often share similar gene expression patterns. In Drosophila the homologous genes (Home genes) are used to pattern the developing embryo along its rostrocau­dal axis (Akam, 1981; Gehring, 1981). Mutations in some of these genes change the identity of one parasegment into that of a neighboring parasegment (Lewis, 1918). Mutational analysis in the mouse has demonstrated that the HOJ( genes, alone or in concert with other HOJ( genes, are also used to regionalize the embyo along its major body axes (Chisaka , To whom correspondence should be addressed. andCapecchi, 1991; Lufkinet al., 1991, Chisakaet a1., 1992; LeMouellic ee al, 1992; Condie and Capecchi, 1993} Dolle ee al., 1993; Gendron-Maguire et al" 1993; Jeannotte et al., 1993, Ramirez·Solis et al., 1993J RijU et a1., 1993, Small and Potter, 1993, Davis and Capecchi, 1994, Kostic and Ca· pecchi, 1994, Satokata et 01., 1995; Suernori et al" 1995; Boulet and Capecchi, 19961 Goddard et a1, 1996). Thus mu­tations in the 3' Hox genes affect the fortnation of anterior structures whereas disruption of 5' genes gives rise to poste­rior abnortnaUties.Regionalization of the embryo by Hox genes appears to be accomplished by the controlled tempo­ral and spatial activation of these genes such that a 3' gene is activated prior to and in a more anterior region of the embryo than its 5' neighbor (Duboule and Dolle, 1989; Gra· ham et a1.. 1989; Duboule, 1994; Capecchi, 19911. However, Hox genes function as components of highly integrated cir­cuits such that paralogous genes, adjacent genes on the same linkage group, and even nonparalogous genes in sepa­rate linkage groups interact poSitively, negatively, and in parallel with each other to orchestrate the morphological regionalization of the embryo (Condie and Capecchi, 1994; Rancourt et a1.. 1995; Davis et a1.. 1995; Horan et a1., 199503, bl Barrow and Capecchi, 1996; Davis and Capecchi, 1996; Favier et a1., 1996; Fromental·Ramain et al., 19961. OOl2·1606/97 $25,00 Copynght () 1997 by Academic Prus Al! rights of reproorn:tion in olny form reserved. Targeted Mutations in Hoxa-9 and Hoxb-9 A Hoxb-9 genomic DNA RI I 12kb RI I RI RV E ES E E XBRI I II I i II Lu .. IWd:I n ... 1 I ~ 5' flanking probe 3' flanking probe B Targeting vector homologous recombination E ES E BRI B X I II I t I I I TK2 ----... TK1 MC1 neopoly A C Targeted allele ~ RI AV E ES E BRI B XBRI -...... I I I II I tl I II I .. .... _- ~ ----... IWd:I MC1 neopoly A AI RI I I 8.8kb RI AI I I 4.3kb D E F CD (wf) ..C.. l +/+ -/- +/- +/- +/- -/- +/- +/- +/+ +/- -/- +/- +/+ -333bp -188bp FIG. 1. Disruption of the hoxb-9 gene and analysis of the mutant genotype. IA-C} Diagrams of the wild-type hoxb-91ocus, the targeting vector, and the targeted allele, respectively. The black box indicates the homeobox of hoxb-9. Hoxb-9 is transcribed from left to right. The bars indicate restriction fragments produced by BeoRI digest. RI, EcoRI; RV, EcoRVi Il, Be047m; S, SolI; X, Xhol; B, BomHI. (0, III Southern transfer analyses of the targeted cell line and of the intercross genotypes, respectively. Genomic DNA was digested with BeaRI and probed with the 3' flanking probe. The 12- and 4.3-kb bands were from the wild-type and mutant alleles, respectively. IF) PCR analysis of the intercross genotypes. The wild-type band is 188 bp and the mutant band is 333 bp. Although all 39 Hox genes are likely to be involved in the formation of the axial skeleton, the specific malformations caused by mutations in individual Hox genes are difficult to predict. Some of this difficulty can be attributed to over­lap of function between genes of the same paralogous group or between genes of different paralogous groups. The rostro-caudal direction of apparent homeotic transformations of vertebrae seen in Hox mutants is also unpredictable. In fact, the two polarities can appear in the same mutant IJeannotte et al., 1993; Small and Potter, 1993). Often the defects are observed at the anterior limit of the Hox gene expression pattern, a phenomenon that has led to the proposal of the Copynght ~ 1997 by M<ldcmlc Press. All rights of rerroduc:tion in any form res.erved. 11 Chen and Capecchi FIG.2. Defects in the thoracic skeleton in noxb-9-/- mice. lA-E) Ventral views of the thoracic skeletons with the vertebral columns removed. (F-J) Lateral views of the upper thoracic regions of the same embryos showing different rypes of first and second rib fusions. The black arrow points to a ventral rib element which connects only to the sternum. posterior prevalence modellDuboule, 1991}. However, even this correlation has many exceptions. A model that, in its broad sense, is likely to be correct is that the combination of Hox genes expressed in a given region determines the identity of structures in that region. However, such combi­natorial models should not be viewed as a code involving TABLE 1 Skeletal Phenotypes in Mice Deficient for hoxb-9 +/+ Phenotypes In ~ 11) First and second rib fusion 0 0 Bilateral 0 0 Unilateral 0 0 Articulating eighth nb 0 2 (7.4%1 Both sides 0 2 One side 0 0 20 (83%) 18 2 13 (54%1 12 1 simple addition of elements, because these gene products, as already pointed out, are likely to interact positively, nega­tively, and in parallel with each other. Simple combinatorial models based on overlapping Hox gene expression patterns have in fact been woefully inaccurate in predicting vertebral identities. Instead, the interactions among Hox genes are sufficiently complex that their role in specifying the verte­bral column must be functionally determined on a case by case basis through analyses of both individual and appro­priate combinations of Hox gene mutations. Herein we describe the phenotypic consequences of dis­rupting hoxb-9 in mice, as well as the effects of combining hoxb-9 and hoxa-9 mutations. These mice show defects in the formation of the thoracic skeleton. MATERIALS AND METHODS Targeting Vector A 9.9-kb genomic clone containing the hoxb-9 gene was isolated from a h DNA Iihrary prepared from mouse CC 1.2 embryo-derived stem IES) cells and used to construct a replacement-type targeting Copyright.~ 191)7 hy Academic PrCMI. All Tii'!tW. of rl'l,fmhtctiotl in ;:Iny 'nun reserved 12 Targeted Mutations in Hoxa·9 and Hoxb·9 vector. In order to disrupt tbe hoxb·9 coding sequence, tbe MClneo polYIAJ cassette IThomas and Capeccbi, 1987) was inserted into tbe Ec047m site witbin tbe bomeodomain. Completion of tbe targeting vector involved flanking the hoxb·9 genomic sequences with the HSV1 and HSV2 thymidine kinase genes (Fig. 11. Electroporation and the Generation of hoxb-9 Mutant Mice Tbe targeting vector was linearized by digestion witb Xhol and electroporated into RI ES cells IDeng and Capecchi, 1992, Nagy et ol., 1993). ES cells containing a disruption of the hoxb·9 gene were enricbed by positive-negative selection IMansour et oJ., 19881. DNA samples isolated from colonies of ES cells were digested with EeoR! and probed Witb 5' and 3' flanking probes (Fig. I). Two per­cent of tbe ES cell lines contained tbe desired hoxb-9 mutation. The targeted £S cell line (1g6) was used to generate chimeric mice that transmitted the hoxb-9 mutant allele to tbeir progeny (Capec­cbi, 1989, 1994). The Generation of Mice Deficient for Both hoxa-9 and hoxb·9 The hoxa-9 mutant mice used in this study were generated by Peterson, Chisab, and Capecchi (unpublished datal. Since both hoxa-9-/- miee and hoxb-9-/- mice are fertile IFromental-Ra­main et ai., 19961, compound heterozygotes for hoxa-9 and hoxb-9 mutations (hoxa-9+/- andhoxb-9+/-1 were obtained from crosses between these homozygous mutant mice. Genotype Analysis DNA was prepared from tail biopsies of adult and newborn mice and from yolk sacs of embryos_ Hoxb-9 genotypes were determined either by Southern transfer analysis with the 3' flanking prnbe or by amplification of DNA fragments using the polymerase chain reaction IPCRI. Hoxa-9 genotypes were determined by PCR. The sequences of the PCR primers used for this analysis were: hoxb-9 forward primer, 5'CTCCAATGCCAGGGGAGTAG3'; hoxb-9 re­verse primer, 5'CTTCTCTAGCTCCAGCGTCTGG3'1 MClneo reverse primer for hoxb-9. 5'GTGTTCGAATTCGCCAATGAC­AAGa'; hoxa-9 forward primer, 5'CGCfGGAACTGGAGAAGG­AGTITCTG3'/ hoxa-9 reverse primer, ATCCTGCGGTTCTGG­AACCAGA Tea' I MClneo reverse primer for hoxa-9. 5'TC!' ATC­GCCTTCTTGACGAGTTC3'. An example of the hoxb-9 genotyping results is given in Fig. 1. Histology Newborn mice were euthanized by asphyxiation with CO" fixed in 4% formaldehyde in phosphate-buffered saline [PBSI overnight FIG. 3. Sagittal sections of hoxb-9+/- and -/- newborn mice. Sections iAI and IBI, IC) and (Dl. [EI and (F), IG) and [HI were from similar poSitions of the mice. Sections from top to bottom were taken progressively further away from the midlines of tbe mice. Tbe space in the upper thoracic region and the size of the thymus are reduced in the h01(b-9-/- mutants with rib fusions. 13 at room temperature (Manley and Capecchi. 19951. and embedded in paraffin according to standard protocols. Ten·micrometer serial sagittal sections were collected and regressively stained with hema· toxylin and eosin IChisaka and Capecchi. 19911. Newborn and adult whole mount skeletons were prepared as described by Mansour et a!. 119931. Whole Mount in Situ HrbridiZtltion Whole mount in situ hybridization on £9.0-£12.5 embryos was performed as described (Carpenter et al .• 1993; Manley and Capec· chi, 19951. The concentration of digoxigenin-UTP·labeled RNA probes in the hybridization mixture was approximately 0.4 pg/ml. Alkaline phosphatase-conjugated anti-digoxigenin Fab fragment was used at a 1:4000 dilution. The templatea for hoxb·9, hoxb-B. and hoxb·7 RNA in situ probes were as follows: a 517-bp fragment in the 3' untranslated region (UTR/ of hoxb-9, a 356·bp SacI-Kpnl fragment from the 3' UTR of hoxb-B. and a 524-bp fragment from the 3' UTR of hoxb·7. respectively. RESULTS Generation of hozb-9 MutlUJt Mice The structure of the targeting vector used to disrupt the boxb-9 gene in RI £S cells is shown in Fig. 1. The insertion of the nec cassette into the boxb·P homecbox terminates the protein prior to all three h~es of the homeodomain and therefore should render the gene product nonfunctional with respect to DNA binding. Southern transfer analyses, using 5' and 3' probes fImking the targeting vector as well as an internal probe, were used to ensure that no rearrange· ments of the boxb·Plocus had occurred other than the de· lIired nec insertion into the homeodomain. A representative targeted ce1lline was used to produce chimeric males that passed the boxb-9 mutation through the gennline. Hoxb-9 MutlUJts Are Viable IUJd Fertile Mice heterozygous for the boxb·9 mutation were inter­crossed to produce homozygotes. Adult boxb·P-/- mice were obtained at a frequency predicted from a Mendelian distribution of mutant and wild-type alleles, indicating no lOllS of mutant alleles as a consequence of ernbryonic or postnatal lethality . Hoxb·9-/ - mice appear outwardly nor· mal and animals of both sexes are fenile. Hoxb-9 MutlUJt Mice Display Skeletal Defects in the Thoracic Region The skeletons of hoxb·9 mutant mice showed malforma· tions in the thoracic region (Table I, Fig. 21. Of 24 hoxb· 9- homozygotes examined, 20 had fusions of the first and second ribs. Ninety percent of the £naions were bilateral. The sternum of a normal mouse consists of six ossified segments, the manubrium, four sternebrae, and the xiphoid process. In animals with first and second rib fusions on both Chen and Capecchi sides of the body, the normal cartilage centers associated with the attachment of the second pair of ribs are absent IFig. 21. As a result, the upper sternum has one less ossified segment. The absence of that ossified segment causes the upper rib cage to be shonened by approximately 35 % in the distance from the rrtanubrium to the point where the third rib articulates with the sternum. This shonening is in part due to the loss of two growth plates normally present at the ends of each ossified segment. Segmentation of the sternum results from the attachment of the ribs to the sternum which locally inhibits the hypertrophy of the canilaginous cells of the sternum near the attachment site IChen, 1953). Thus, the observed abnormal segmentation patterns of the sternum can be understood in terms of the abnormal pat­terns of first and second rib attachment to the sternum. The rrtaiority of the first and second rib fusions occurred at the point of attachment with the sternum IFig. 2GJ. About 25 % of the fusions occurred by alternative pathways. For example, in some animals the first and second ribs fused prior to attachment to the sternum, formed a common ven· tral rib, and then attached to the sternum (Fig. 2HI. In other mutants, the first rib branched after fusion. In some of these cases both branches were articulated with the sternum; in others only one branch was attached. Finally, iIi two homo­zygotes the first rib attached to the sternum but lost connec· tion with either the second rib or the dorsal ponion of the first rib, implicating remodeling after first rib attachment (Fig. 211. Thineen of the 24 boxb·9-/- mice examined had an eighth rib attached to the sternum on one or both sides (Figs. 1B, 10, and 2E). A similar phenotype has been described in boxc-B and boxc-9 mutant mice (LeMouellic et aI., 1993, Suemori et aI .. 1995). Such eighth rib attachment appears to occur independendy of the first and second rib £naions, since some animals had eighth rib articulations with the sternum but did not show first and second rib fusions (Fig. 2EJ. Approximately 7% of the boxb·P- heterozygotes also showed eighth rib attachment to the sternum. This pheno­type was never observed in wild.type control animals. The thoracic vertebral bodies of boxb-9-/- mice appear normal. Thus, the defects appear to be restricted to the patterning of the ribs and sternum ILe., the ventral aspects of the tho· racic axial columnl. Fonnation of the appendicular axis ap­pears normal in boxb-9 mutant homozygotes. Histological &o.mination of hoxb-9-1- Mice Sagittal and parasagittal sections of newborn boxb-9 ho­mozygous mutant mice reveal a significant reduction in the size and an alteration in the shape of the thymus relative to wild.type and hoxb·9 heterozygous littermates IFig. 31· The conex and medulla of the thymus in hoxb·9-/- mice, however, appear normal. We presume that the altered shape of the thymus is an indirect effect of the reduced space in the upper thoracic region in hoxb·9 mutant homozygotes. Consistent with this hypothesis, hoxb·9 expression is not observed in tissues contributing to the formation of thymus. Copyri&ht e 1997 by Academic Press, An rightll of reproduction in any form reserved. 14 Targeted Mutations in Hoxa-9 and Hoxb-9 TABLE 2 Thymus Weight Comparison between hoxb-9 Homozygous Mutants with and without the First and Second Rib Fusions· Genotype With rib fusion Yes No No Yesc Yes Yes Yes Thymus weight "T'Ig) 0.022 0_024 0.027 0.026 0.022 0.026 0.021 Body weight "B" Igl 3.9 3.6 4.2 4.4 4.5 4.5 4.2 TIB" 0.0056 0.0067 0.0064 0.0060 0.0049 0.0058 0.0050 Note. Thymus weight was independently normalized to heart weight and the results were indistinguishable from those shown above . • These are 8.day-old littermates from a cross between two hoxb- 9 homozygous mutants. "If we use TIBI,. for the average value of TIB for mutants with the rib fusions (Nos. 1,4,5,6, and n we get TIBI ..... 0.0055, if we use TIBu for the average value of TIB for mutants without the rib fusions INos. 2 and 31, we get TI BOA .. 0.0066. The difference be· tween these two values is 20%. , No. 4 had rib fusions that were not as severe as the other rib fusions observed in this litter. Also, mutant animals do not appear to be compromised with respect to their immune function. For enmple, fluo­rescence- cytometric analysis of blood from mutant and control mice showed a normal distribution of T and B lym. phocytes (data not shown). Finally, as previously men· tioned, the rib fusions which are responsible for the reduc­tion in the size of the upper thoracic cavity are not present in all hoxb-9 mutant homozygotes. When the weight of excised thymic tissues was compared between hoxb-9 mu­tant homozygous mice with and without the first and sec­ond rib fusions, the average ratio of thymus to body mass in mutants with the rib fusions was 20% lower than in mutants with normal ribs (Table 21. E13.S embryos were immunostained with the 2H3 anti· body directed against a subunit of the neurofilament protein (Dodd et al., 1988) to reveal possible changes in the pattern of neurons in the peripheral nervous system. No neuronal defects were apparent inhoxb·9 mutant homozygotes either in the body wall or in the limbs (data not shown). Hoxb-9 Is Expressed at the Axial Level from Wbich Affected Skeletal Elements Arise The defects in the upper thoracic region of hoxb-9- homo­zygotes are found at Significantly more rostral levels than the reported anterior limits of expression for hoxb-9 or its paralogues at E12.5 (Bogarad et al .. 1989; Burke et a1., 1995). To resolve this discrepancy, we reexamined the expression of hoxb-9 at E9.5, E1O.5, ElLS, and E12.5 using whole mount RNA in situ hybridization IFig. 41. In E9.5 embryos, the ex· pression of hoxb-9 was strong in the neural tube with an anterior limit at the level of somites 7-8 (pV3). At this time, a high level of expression in the somites was detected in the posterior portion of the embryo but lower levels were seen in the rostral portion of the embryo. Expression in the somites could, however, be readily detected up to the level of the seventh to eighth somite. At E 10.5, the anterior limit of expression in the neural tube shifted rostrally by one to two somites. Expression in the spinal ganglia was also apparent. Paraxial mesoderm expression appeared to be enhanced in ventral portions of the somites. At El1.5 and E12.5, changes in neural tube expression were not apparent. However, the intensity of the RNA hybridization signal decreased progreso sively in the anterior region of the embryos in paraxial meso­derm- derived structures which included the vertebral bodies and rib primordia_ Hoxb-9 expression was evident in the kid­neys from E12.5 onward. From these studies, it is evident that at E9.5, the anterior limit of hoxb-9 expression is sufficiently anterior to cover the region of thoracic defects observed in hoxb-9 mutant mice. The anterior limit of expression in paraxial meso· derm-derived structures then shifts caudally. However, hoxb-9 expression is apparent in the regions of the rib and sternum primordia (i.e., the tissues affected by the hoxb-9 mutation). The boxb-9 Mutation Does Not Affect the .Expression 0/ Neighboring Hox Genes It bas been reported that in some cases a mutation in one Hox gene can affect the expression of a neighboring Hox gene (Suemori et al., 1995; Barrow and Capecchi, 19961 Bou­lefand Capecchi, 1996). To determine if the hoxb-9 muta­tion affected hoxb-8 or hoxb-7 expression, we conducted whole mount RNA in situ hybridization experiments on EIO.5 and E 12.5 embryos of all three hoxb-9 genotypes (i.e., +1+; +1-; -1-). Neither the anterior limits of hoxb-8 ex­pression in the neural tube or in paraxial mesoderm-derived structures nor the overall level of hoxb-8 expression was distinguishable in embryos of these three hoxb-9 genotypes {Fig. 51. Similarly, the expression pattern of hoxb-7 was not affected in hoxb-9 heterozygous and mutant homozygous embryos (data not shown). Thus, neither the presence of the neo insertion in hoxb-9 nor the absence of functional hoxb-9 protein appears to affect the expression of the neigh­boring Hox genes. As a further test for relationships between hoxb-9 and neighboring Hox genes, hoxb-9/hoxb-8 and hoxb-9/hoxb- 7 transheterozygotes were constructed by crossing boxb-9 mutants with hoxb-8 (Greer and Capecchi, data not shown) or hoxb-7 (Chen and Capecchi, unpublished datal mutants. At low frequencies [20%), hoxb-9/hoxb-7 transheterozy­gotes (Le., hoxb·9+1-; hoxb-7-/+ mice) and (14%) of hoxb- 9/hoxb-B transheterozygotes showed first and second rib fusions very similar to those observed in hoxb-9- homozy­gotes. Such fused ribs have never been observed in mice heterozygous for hoxb-9. hoxb-8. or hoxb·7 alone. Such non- Copyright 1997 by Academic Press. All rights of reproouction in any form reserved. 15 allelic noncomplementation suggests that hoxb·7, hoxb·8, and hoxb·9 gene products directly or indirectly interact to specify the upper thoracie region of the mouse IRancourt et al., 19951. Hoxa-9/hoxb-9 Double Mutants Hoxa·9 mutant mice have been described by Fromental· Ramain et al. 11996). Since both hoxa·9-j- and hoxb·9-/ - mice are fertile, compound heterozygotes for both hoxa· 9 and hoxb·9 could be obtained from crosses between hoxa· 9 and hoxb-9 mutant homozygotes. These compound het­erozygotes appeared outwardly normal and were fertile. Mice of all nine possible genotypes were obtained from crosses between such compound heterozygotes, and the nine genotypes were obtained at the expected Mendelian ratios, indicating that all genotypes, including double mu­tant homozygotes, were viable. Skeletal Defects in hOXQ-9!hoxb-9 Double Mutants Hoxa·9 mutant homozygotes have an extra pair of fully grown ribs on the 21st vertebra (I.e., show an anterior homeosis of the Ist lumbar vertebra to a thoracic vertebral Fromental-Ramain et aI., 1996). Hoxa-9 heterozygotes occa­sionally have small rib anlage on the Ist lUmbar vertebra (Peterson, Crusan and Capeccru, unpublished results). Hoxa-9 heterozygotes carrying one or two mutant alleles of hoxb-9 had a pair of full-size 14th ribs (Table 3, Fig. 6). Thirty-eight percent of hoxa-9+/-, hoxb-9+/- mice had pairs of fully grown ribs. The penetrance increased to 53% in hoxa-9+/-; hoxb-9-/- mice, indicating a quantitative effect of the hoxb·9 mutant alleles on mediation of this homeosis in mice heterozygous for the hoxa-9 mutation. An increase in the penetrance of the 1st and 2nd rib fusion phenotype was also observed when hoxb-9 homozygous mutant mice received one or two copies of the hoxa·9 mu­tant allele (Table 3 and Fig. 6). Approximately 83 % of hoxb- 9 homozygous mice show Ist and 2nd rib fusions (Table 11, while 93% of the hoxa-9+/-1 hoxb-9-/- mice and 100% of thehoxa-9-/-; hoxb-9-J- mice show rib fusions. More­over, the rib fusions in hoxa-9/hoxb-9 double mutants are more severe. Some double mutants even had the 3rd rib fused to the 1st and 2nd rib fusion (Fig. 6). Exacerbation of sternum defects is also evident in the hoxa-9Ihoxb-9 double mutants (Fig. 6H/. In such animals the length of all of the intercostal segments is greatly reduced. As is evident from Figs.6F, 6G, and 6H, the expressivity of the sternum defects varies in the hoxa-9/hoxb-9 double mutants, suggesting that other members of this paralogous family may play im­portant roles in patterning the thoracic vertebrae. The more severe abnormalities evident in hoxa-9-/-; hoxb-9-/­mice compared to defects observed in hoxa·9 and hoxb-9 single mutant homozygotes support the hypothesis that these two genes function synergistically to pattern the tho­racic vertebrae. Chen and Capecchi FIG. 4. The expression of hoxb·9 in wild-type embryos. The ante­rior limit of neural tube expression Iwhite arrow I and the anterior limit of paraxial mesoderm expression Iblack anowl are at approxi­mately pya at E9.5IAJ. The anterior limit of neural tube expression shifts rostrally by one to two somites at ElO.5 (BI and remains unchanged at EII.SIc) and EI2.SIDI. The anterior limit of paraxial mesoderm expression shifts caudally in the upper thoracic region between E9.5 and ElO.5. The intensity of the signal from mesoderm expression decreases progressively and becomes restricted to struc· tures that appear to be primordia of the venebral bodies Iblack triangle) and ribs (white triangle I. DISCUSSION Eighty-three percent of the hoxb-9 homozygous mutants examined had first and second rib fusions while 54% of homozygotes and a small number of heterozygotes showed abnormal attachment of the eighth rib to the sternum. Both of these defects can be interpreted in terms of anterior ho­meotic transformations. In many hoxb-9 mutant homozy­gotes, the articulation angle and position of sternal attach­ment of the second rib resembles those of the first rib. Nor­mally, the first rib has two attachment points with the Copynght t,~ 19':>7 by AC;ldl.'mic PrcSfl All ri~hts of reproduction in .my form I).;scrvcd. 16 17 Target.ed Mutations in Hoxa-Y and Hoxb-Y 6 E (AA;BB) FIG. 5. The expression pattern of hoxb-8 does not change in hoxb-9 mutants. (A) hoxb-9+/+, IB) hoxb-9+/-, ICI hoxb-9-/-. White arrows indicate the anterior limits of neural tube expression. Black arrows indicate the anterior limits of mesoderm expression. Copyrigiu ,_, 191;17 hy Ac;\dcmit:: Pn.'"s. All rl,i\hfS of rcpmJuclion III .my form rl'::l'rveu. TABLE 3 Summary of Skeletal Phenotypes in Ihoxa-9; hoxb-9) Mice Phenotypes First and second rib fusion Articulating eighth rib 14 pairs 01 ribs Aa;Bb (n = 1610 o 116%) 6 (38%) aa,Bb (n = III o 1 (9%) 111100%1 Aa;bb In = IS) 14 (93%) 8 {53%I 8 (53%1 Chen and Capecchi aa;bb In 9) 9 (100%)" 5 (56%) 9 (IOO%) a A, hoxa·9 wild.type allele; a, hoxa·9 mutant allele; B, hoxb·9 wild·type allele, b, hoxb-9 mutant allele; b Three (aa;bb) mice had the third rib fused to the second rib. When this happened, the TI ribs either fused to the 1'2. ribs or lost contact with the sternum and any other ribs. sternum: one at the tip of the manubrium, the other at the body of the manubrium. The other ribs have only one point of articulation. In mice with the first and second rib fusions, the second rib tends to have two attachment points that resemble those of the first rib. The abnormal attachment of the eighth rib to the sternum can also be regarded as the eighth rib acquiring seventh rib character, since in the mutant it morphologically resembles the seventh rib. The formation of the sternum is also abnormal in boxb- 9 mutant homozygotes. However, it is not clear whether these defects are a direct consequence of the boxb-9 muta­tion or the consequence of aberrant articulations with defec· tive ribs. Segmentation of the sternum clearly results from articulation with the ribs IChen, 19521. Further, defective segmentation would also be expected to alter th,(growth of the sternebrae. Thus, the fact that we observe sternum de· fects in the upper thoracic region of boxb-9 mutant homozy­gotes can be understood either in terms of defective rib patterning or in terms of a direct role of boxb-9 in stemum formation. However, the more extensive sternum defects observed in boxa-9Iboxb-9 double mutants support the sec­ond hypothesis. In most of the boxa-9Ihoxb-9 double mu­tants, the entire sternum is hypoplastic and malformed. Such extensive sternum malformations are not readily in· terpreted as resulting solely from the defects in the articula· tions of the first, second, or eighth ribs, but instead suggest an additional role for boxa-9 and boxb-9 in sternum forma­tion. This is an important distinction since ribs and ster­num originate from separate mesodermal lineages. It is curious that in boxb-9 mutant homozygotes, defects are observed in the formation of the first, second, and eighth ribs with intervenirtg ribs appearing normal. However, closer examination indicates that the angle of articulation of the ribs with the sternum is abnormal along the entire ribcage (Fig. 2). This indicates that the hoxb-9 mutation contributes to mispatterning of the ribs, and possibly the sternum, from Tl through T8, with the defects only beirtg more apparent at the ends of this block of vertebrae. A consequence of the rib fusions seen in some boxb·9 mutant mice is a reduction in the size of the upper thoracic cavity. Interestingly, in those hoxh-9 mutant homozygotes that have first and second rib fusions, the size of the thymus is also reduced. The reduction in the size of the thymus appears to result as a secondary response to the reduction in the size of the upper thoracic cavity since in boxb-9 mutant mice lacking rib fusions the thymus is normal in size. These observations suggest that the size of the thymus is regulated with respect to the size of the thoracic cavity. Hoxa-9/hoxb-9 Double Mutants Mice homozygous for mutations in either boxa-9 and boxb-9 do not show overlappirtg phenotypes. Yet double mutants show exacerbation of all of the defects observed in either boxa-9- or hoxb-9- hornozygotes. In addition, both the penetrance and expressivity of the defects correlate with the number of mutant alleles present in the mouse. For example, mice heterozygous for either the hoxa-9 or the hoxb-9 mutation never show full-size 14th ribs. However, such ribs are apparent in some hoxa-9. boxb-9 compound heterozygous mice and the frequency is increased in hoxa· 9+/-, hoxb-9-/- mice. Conversely, hoxa-9-/- mice have never been observed to have 1st and 2nd rib fusions. How­ever, the frequency of such fusions in boxb-9-/- mice pro· gressively increases with the addition of one and then two hoxa-9 mutant alleles. Similar observations have been made in mice containirtg combinations of hoxa-3 and hoxd-3 mu­tant alleles (Condie and Capecchi, 19941. These observa· tions again emphasize the extensive quantitative interac­tions amortg Hox genes required to specify the vertebrate body plan. Each Hox gene appears to have individual unique functions as well as more extensive roles in combination FIG. 6. Skeletal defects in the hoxa-9/hoxb·9 mutant mice. lA-D) Skeletal defects in mice with lour combinations 01 hoxll-9 and hoxb· 9 mutant alleles Isee text for details). IE-H) Alterations in the thoracic skeleton in hOXIl.9-/-, hoxb-9-/- mice. Various rib fusions, abnonnal rib attachments to the sternum, and general malfonnations of the sternum are apparent. Cenotype symbols: A, hoxa-9 wild· type allele, a, hoxa·9 mutant allele; B, hoxb-9 wild-type allele; h, hoxb-9 mutant allele. "LJ" indicates the first lumbar vertebra which bears ribs, Copyri~ht ('I 1997 hy Ac;;demic Press. AU rights uf reproduction in .1r;y form reserved 18 Targeted Mutations in Hoxa·9 and Hoxb·9 with other Hox genes. The concentration of Hox gene prod· ucts in different cells must be very tightly regulated since the reduction of Hox protein concentration in heterozygotes can result in marked phenotypic consequences. Genetic interactions between hoxa·9 and hoxd·9 have been reported IFromental-Ramain et a1., 1996). In this case, the interactions are observed in more caudal aspects of the vertebral column compared to those seen in hoxa-9/hoxb· 9 double mutants. Specifically, Fromental-Ramain et a1. 119961 observed in mice mutant for both hoxa-9 and hoxd-9 an exacerbation of defects in the lumbosacral axial skeleton. Thus, the domain of influence of hoxa-9 extends from the first thoracic vertebra, when interacting with hoxb-9, to sacral vertebrae, when interacting with hoxd-9. As the functions of increasing numbers of Hox genes are unraveled, patterns are emerging. With respect to the forma­tion of the axial column, there appear to be hot spots for the accumulation of defects. Such defects are particularly evident at the boundaries between changes in vertebral type, such as the base of the skull and the first cervical vertebra, C7 and TI, T7 and 1'8, T13 and Ll, L6 and 51, and S4 and Cal. For example, mutations in horb-2, horb- 4, bord-l, hord-3, and hoxd-4 show defects in the formation of the first cervical vertebra, the atlas (Barrow and Capecchi, 1996/ Ramirez-Solis et al., 1993; Condie and Capecchi, 1993; Horan at aI., 1995a, bJ. Mutations in hoxa-4, bora-5, hoxa-6, boxb-5, boxb-6, borb-7. borb-8, and boxb-9 have defects inC7, TI, or both and so on. Why is this the case! Part of the answer is operational. It is very apparent when C7 acquires ectopic ribs or Tlloses ribs. On the other hand, transformations of C5 to C4 are more difficult to score. Second, we can anticipate that the generation of the major morphological differences associated with the different ver­tebral classes will require the concerted activity of more Hox genes than the generation of the smaller differences within a vertebral class. Finally, the vertebral classes may be formed as units using prescribed developmental pro· grams. Consistent with this hypothesis, the expression pat­terns of Hox genes in the prevertebrae of the chick and mouse maintain register with the type of vertebrae rather than the position of the vertebra along the axial column IBurke et aJ.. 19951. In mice in which the continuity of the normal program has been disrupted by a Hox mutation, the discontinuities in the specification of cells are likely to be most apparent at the transitions between vertebral types, making such boundaries particularly vulnerable to dysmor· phology. ACKNOWLEDGMENTS We especially acknowledge Osamu Chisaka for cloning the hoxb- 9 mouse genomiC sequences and for preparing the hoxb·9 targeting vector. We thank M. Allen, C. Lenz, G. Peterson, S. Barnett, E. Nakashima, and M. Wagstaff for excellent technical assistance and L. Oswald for help with preparation of the manuscript. REFERENCES Akam, M. E. (19871- The molecular basis for metameric pattern in the Drosophila embryo. DeyelopmetJt WI, 1-22. Barrow, J. R., and Capecchi, M. R. (19961. Targeted disruption of the hoxb-2 locus in mice interferes with expression of hoxb-l and hoxb·4. Development, 122,3817-3828. Bogarad, L. D., Utset, M. P., Awgulewitsch, A., Miki, T., Hart, C. P., and Ruddle, F. H. (19891. The developmental expression pattern of a new murine homeobox gene: Hox·2.5. Dey. BioI. 133,537- 549. Boulet, A. M., and Capecchi, M. R. (1996). Targeted disruption of hoxc-4 causes esophageal defects and vertebral transformations. Dey. BioI. 177,232-249. Burke, A. c., Nelson, C. E., Morgan, B. A., and Tabin, C. 119951. Hox genes and the evolution of vertebrate axial morphology. De· velopment 121,333-346. Capecc.bi, M. R. (19891. Altering the genome by homologous recom­bination. Science 244, 1288-1292. Capecchi, M. R. 119941. Targeted gene replacement. Sci. Am. 270, 52-59. Capecchi, M. R. (19971. The role of Hox genes in hindbrain develop­ment_ In "Molecular and Cellular Aspects of Neural Develop­ment" /W. M, Cowan, T. M. Jessell, and S. L. Zipursky, Eds.l, Oxford Univ. Press, New York, in press. Carpenter, E. M., Goddard, J. M., Chisalca, 0., Manley, N. Ro, and Capecc.bi, M. R. (19931. Loss of Hcu:-AIIHo.x-l.t;! function results in the reorganization of the murine hindbrain. Development 118, 1063-1075. Chen, J. M. 119521. Studies on the mozphogenesl.s of the mouse atemum L Normal embryonic development. n. Experiments on the origin of the sternum and its capacity for self.·differentiation in vitro. T. Anat. 86, 373-40l. Chen, y. M. (1953). Studies on the mozphogenesis of the mouse sternum I. Normal embryonic development. n. Experiments on the closure and segmentation of the sternal bands. T. Anal. 87, 130-149. Chisalca, 0., and Capecc.bi, M. R. 119911. Regionally restricted de­velopmenral defects resulting &om targeted disruption of the mouse homeobox gene hcu:-l.S. Nature 350, 473-479. Chisaka, 0., Muscl, T. S., and Capecchi, M. R. (1992). Develop­mental defects of the ear, cranial nerves and hindbrain resulting from targeted disruption of the mouse homeobox gene Hox-I.t;. Nilture 355, 516-520. Condie, B. G., and Capecchi, M. R. (1993). Mice homozygous for a targeted disruption of Hoxd-3IHox·4.1 J exhibit anterior transfor­mations of the first and second cervical vertebrae, the atlas and the axis. Development 119, 579-:-595. Condie, B. G., and Capecchi, M. R.II994). Mice with targeted dis· ruptions in the paralogous genes hoxa·3 and hoxd·3 reveal syner· gistic interactions. Nature 370, 304-307. Davis, A. P., and Capecchi, M. R.119941. Axial homeosis and appen­dicular skeleton defects in mice with a targeted disruption of hoxd·l1. Development 120, 2187-2198. Davis, A. P., Witte, D. P., Hsieh, L. H., Potter, S. S., and Capecchi, M. R. 119951. Absence of radius and ulna in mice lacking hoxa· 11 and hoxd-11. Nature 375,791-795. Davis, A. P., and Capecchi, M. R. 119961. A mutational analysis of the 5' Hox D genes: Dissection of genetic interactions during limb development in the mouse. Development 122, 1175-1185. Deng. C., and Capecchi, M. R. (19921. Reexamination of gene laI' geting frequency as a function of the extent of homology between Copyright C [997 by Academic Press, All rights of reproduction in an)' form regerved. 19 the targeting vector and the target locus, Mol. CelL Biol. 12, 3365-3371, Dodd, I" Morton, S, B" Karagogeos, D" Yamamoto, M" and Jessell, T, M, 119881, Spatial regulation of axonal glycoprotein expression on subsets of embryonic spinal neurons, Neuron 1, 105-116, Dolle, P., Dierich, A., Le Meur, M., Schimmang, T" Schuhbaur, B" Chambon, P., and Duboule, D, (19931, Disruption of the Hoxd· 13 gene induces localized heterochrony leading to mice with neotenic limbs, Cell 75, 431-441. Duboule, D" and Dolle, p, 119891. The structural and functional organization of the murine Hox gene family resembles that of Drosophila homeotic genes, EMBO T, 8, 1497-1505, Duboule, D. fi 991 I, Patterning in the vertebral limb, Curro Opin, Genet, Dev, 1, 2.11-2.16. Duboule, D, 119941. Temporal colinearity and the phylotypic pro· gression: a basis for the stability of a vertebrate Bauplan and the evolution of morphologies through heterochrony. Development (Sup pl.}, 135-142.. Favier, B., Rijli, F. M., Frornentai·Ramain, C., Fraulob, Y., Cham· bon, P., and Dolle, P./19961. Functional cooperation between the non.paralogous genes Hoxa·la and Hoxd·ll in the developing forelimb and axial skeleton. Development 122, 449-460. Fromental·Ramain, C., Warot, x., Lakkaraju, S., Favier, B., Haack, H., Birling, C., Dierich, A., Dolle, P., and Chambon, P. 11996). SpeciBc and redundant functions of the paralogous Hoxa·9 and Hoxd·9 genes in forelimb and axial skeleton patterning. Develop. ment 12.2., 461-472.. Gehring. W. J. 11987). Homeoboxes in the study of development. Science 2.36,1145-1252. Gendron, M. M., Mallo, M., Zhang. M., and Gridley, T. 11993). Hoxa-2 mutant mice exhibit homeotic transformation of skeletal elements derived from cranial neural crest. Ce1175, 1317-1331. Goddard, I. M., Rossel, M., M.an.Iey, N. R., and Capecchi, M. R. (19961. Mice with targeted disruption of ba1cb·l faU to form the motor nucleus of the vnth nerve. Development, 12.2.,3217-32.2.8. Graham, A., Papalopulu, N., and Krumlauf, R. (19891. The murine and Drosopbila homeobox gene complexes have common fea· tures of organization and expression. cell 57, 367-378. Holland, P. W. H., and Garcia·Fernandez, I. (1996). Hox genes and chordate evolution. Dw. BioL 173,382-395. Horan, G. S. B., RamJrez.Solis, R., Featherstone, M. S., Wolgemuth, D. I., Bradley, A., and Behringer, R. R. 119958). Compound mu­tants for the paralogous hoxa4, hoxb4. and hoxd4 genes show more complete homeotic transformations and a dose·dependent increase in the number of vertebrae transformed. Genes Dev. 9, 1667-1677. Horan, G. S. B., Kovacs, E. N., Behringer, R. R., and Featherstone, M. S.II995b). Mutations in paralogous Hox genes result in over· lapping homeotic transformations of the axial skeleton: Evidence for unique and redundant function, Dev. BioI. 169, 359-372. leannotte, L., Lemieux, M., Charron, I., Poirier, F., and Robertson, E. 1.11993). Specification of axial identity in the mouse: Role of the Hoxa·5 (Hoxl31 gene. Genes Dev, 7, 2085-2.096. Chen and Capecchi Kostic, D., and Capecchi, M, R, [19941. Targeted disruptions of the murine hoxa·4 and hoxa·6 genes result in homeotic transforma­tions of components of the vertebral column. Mech, Dev. 46, 231-247. Le Mouellic, H., Lallemand, Y., and Brulet, P. [19921, Homeosis in the mouse induced by a null mutation in the Hox-3.1 gene. Cell 69,251-2.64. Lewis, E. B. (1978). A gene complex controlling segmentation in Drosopbila. Nature 278, 565 -5 70. Lufkin, T., Dierich, A., LeMeur, M., Mark, M., and Chambon, p, (1991). Disruption of the Hox·l.6 bomeobox gene results in de­fects in a region corresponding to its rostral domain of expression. Ce1166, 1105-1119. Manley, N. R., and Capecchi, M, R. (1995). The role of hoxa·3 in mouse thymus and thyroid development. Development 121, 1989-2003. Mansour, S. L., Thomas, K. R., and Capecchi, M. R. (19881. Disrup· tion of the proto-oncogene int·2 in mouse embryo.derived stem cells: A general strategy for targeting mutations to non·selectable genes. Nature 336, 348-352.. Nagy, A., Rossant, I., Nagy, R., Abramow·Newerly, W., and Roder, I. C. /1993). Deriva tion of completely cell culture-derived mice from early.passage stem cells. Proc. Natl. Acad. Sci. USA 'Xl, 8414-842.8. Pendleton, J. W., Nagai, B. L, Murtha, M. T., and Ruddle, F. H. 119931. Expansion of the Hox gene family and the evolution of chordates. Proc. Natl. Acad. Sci. USA 90, 6300-6304. Ramf:rez-Solis, R., Zheng, H., Whiting, J., Krumlauf, R., and Bradley, A. f!9931.'Hoxb4IHox-2.61 mutant mice show bomentic trans· formation of a cervical vertebra and defects in the closure of the sternal rudiments. Ce1173, 279-294. Raneourt, D. B., Tsuzuki, T., and Capecchi, M. R. 119951. Genetic interaction between bozb·5 and hox'b-6 is revealed by nonallelic noncomplementstion. Genes Dw. 9, 108-12.2.. Rijli, F. M., Mark, M., Lakkaraju, S., Dierich, A., Doll~, P., and Chambon, P.1(993). A bomeotic transformation is generated in the rostral branchial region of the head by disruption of Hoxa-2. which acts as a selector gene. Cell 75, 1333-1349. Satokata, I., Benson, G., and Maas, R. (19951. Sexually dimorphic sterility phenotypes in HoxalQ. deficient mice. Nature 374, 460- 463. Small, K. M., and Potter, S. S. (19931. Homentic transformations and limb defects in Hoxll·l1 mutant mice. Genes Dev. 7,2.318- 2.32.8. Suemori, H .• Takahashi, N., and Noguchi, S. (19951. Hoxc·9mutant mice sbow anterior transformation of the vertebrae and malfor. mation of the sternum and ribs. Meeh. Dev. 51,265-273. Thomas, L R., and Capecchi, M. R. (1987). Site·directed mutagene· sis by gene targeting in mouse embryonic· derived stem cells. Cell 51,503-512. Received for publication August 20, 1996 Accepted September 21, 1996 Copyright" 1997 by Academic Press. Ail rights of reprodu!.':tioll in any form reserved 20 CHAPTER 3 Hoxa9, Hoxb9, AND Hoxd9 FUNCTION IN THE MAMMARY GLAND AND THE PATTERNING OF THE AXIAL AND APPENDICULAR SKELETON Abstract Mice carrying all possible combinations of disrupted alleles for Hoxa9, Hoxb9, and Hoxd9 have been generated by appropriate matings starting from the three individual mutants. Triple homozygous mutants (aabbdd*) and mutants with certain combinations of mutant alleles have extensive rib fusions and overt defects in their limbs. However, even triple homozygous mutants show only a moderate reduction in viability. Male triple mutants * * , except for aabbdd, are normal in fertility. Mothers with the genotypes of Aabbdd, aaBbdd, and aabbdd, cannot raise their own pups. Pups born to these females die of apparent malnutrition, which can be rescued by fostering. The mammary glands of these females are underdeveloped during pregnancy and around parturition. These Hox genes are expressed in the fetal mammary gland primordia and in adult mammary glands, further indicating direct functional roles for these Hox genes in controlling the development and function of the mammary gland. Relative importance of these genes, as manifested by their contribution to the mutant phenotypes, varies in different body regions, indicating that these genes form distinctive functional hierarchies in different regions. Introduction The mammalian Hox complex contains at least 39 genes that are divided into four linkage groups on four separate chromosomes. These four linkage groups (or clusters) are referred to as HOXA, HOXB, HOXC, and HOXD in human, and as Hoxa, Hoxb, Hoxc, and Hoxd in mouse. The four linkage groups are believed to have arisen by quadruplication and divergence of a single ancestral complex common to vertebrates and invertebrates (Pendleton et al., 1993; Holland and Garcia-Fernandez, 1996). Individual genes of the four linkage groups have been classified into 13 * In this chapter, we use A, B, and D to designate the wild-type alleles of Hoxa9, Hoxb9 and Hoxd9, respectively; we use a, b, and d to designate the mutant alleles. ** Mice with at least one mutant allele in each of the three loci, including the eight genotypes from AaBbDd to aabbdd. 22 paralogous families on the basis of sequence similarity and position within each linkage group. Studies on mouse gain-of-function and loss-of-function mutations have demonstrated that Hox genes collectively control the identity of the various regions along major body axes (Chisaka and Capecchi, 1991; Lufkin et ai., 1991; Chisaka et ai., 1992; Le Mouellic et al., 1992; Condie and Capecchi, 1993; Dolle et al., 1993; Gendron-Maguire et ai., 1993; Jeannotte et al., 1993; Ramirez-Solis et al., 1993; RijIi et al., 1993; Small and Potter, 1993; Davis and Capecchi, 1994; Kostic and Capecchi, 1994; Satokata et al., 1995; Suemori et al., 1995; Boulet and Capecchi, 1996; Goddard et ai., 1996; Barrow and Capecchi, 1996; Chen and Capecchi, 1997; Godwin and Capecchi, 1998). Paralogous genes and nonparalogous genes in the same cluster or in different clusters interact with each other in the regionalization of the embryo (Condie and Capecchi, 1993; Horan et al., 1995; Rancourt et al., 1995; Davis and Capecchi, 1995 and 1996; Fromental-Ramain et al., 1996; van der Hoeven et ai., 1996; Chen and Capecchi, 1997). The relative importance of genes functioning in the same embryonic region, however, can be very different. Studies of mouse mutants involving the 5' Hoxd genes and the analysis of human HOXD13 mutants have indicated that Hoxd13 may have a dominant role over the other 5' Hoxd genes in limb patterning, and is thus on the top of the functional hierarchy in this region (Muragaki et al., 1996; Zakany et al., 1996; van der Hoeven et al., 1996). Studies of the single and compound mutants of Hoxa9 and Hoxd9 revealed their synergistic interactions in the patterning of the lumbosacral axial skeleton and the stylopodallimb skeleton (Fromental-Ramain et at., 1996). Synergistic interactions of Hoxa9 and Hoxb9 have also been illustrated by the abnormal patterning in the thoracic region in single and compound mutants of Hoxa9 and Hoxb9 (Chen and Capecchi, 1997). In this report, we will describe the abnormalities found in the axial and appendicular structures of the triple mutants for Hoxa9, Hoxb9, and Hoxd9. We will further discuss the interactions and apparent functional hierarchies among these genes. Although many studies have established that Hox genes play key roles in embryonic development of the mammals, their functions in adult animals are less certain. The observation that some Hox genes are expressed in kidney, hematopoietic 23 cell lineages and the detection of Hoxc6 expression in adult murine mammary gland have led to the speculation that some Hox genes may have functional roles in these adult tissues (Friedmann et al., 1994). Misexpression of Hox genes was frequently found to be involved in human or murine leukaemias. For example, the misexpression of Hoxb8 due to viral insertion and the activation of Hoxa9 (or human HOXA9) by proviral integration or chromosomal translocation are believed to be the causes of some myeloid leukaemias (Blatt and Sachs, 1988; Nakamura et al., 1996; Borrow et al., 1996). Our studies on the triple mutant females show that disruptions of Hox genes can cause hypoplasia of mammary glands in pregnancy and postpartum. Thus, our results indicate that a subset of Hox genes is involved in the control of growth and differentiation of the mammary gland. Materials and Methods The generation of mice deficient for Hoxa9, Hoxb9, and Hoxd9 Individual mutants carrying disruptions in their respective homeoboxes for Hoxa9, Hoxb9, and Hoxd9 were generated (Chen and Capecchi, 1997, Chisaka, Goddard, Spyropoulos, and Capecchi, unpublished data) by gene~targeting and ES cell technology as described (Capecchi, 1989). Matings between the individual mutants were carried out to generate double mutants and, subsequently, triple mutants. Genotypes of Hoxa9 and Hoxb9 were determined by PCR as described (Chen & Capecchi, 1997; Chapter. 2 of this dissertation). The sequences of the PCR primers used for the PCR analysis of the Hoxd9 genotypes were: Hoxd9 forward primer: S'AGCGAACTGGATCCACGCTCGCTCCA3', Hoxd9 reverse primer: S'GACTTGTCTCTCTGT AAGGTTCAGAATCC3', neo reverse primer for Hoxd9: S'CGTGTTCGAA TTCGCCAA TGACAAGAC3'. Histology Newborn and adult whole-mount skeletons were prepared as described by Mansour et at., 1993. Mammary gland biopsies were fixed in Carnoy's fixative or in 4% fonnaldehyde in phosphate-buffered saline (PBS) overnight at room temperature. Fat was removed from the tissues using three separate I-hour washes of acetone. For 24 whole-mount preparation, the mammary glands were rehydrated through an ethanol series, then stained in hematoxylin for 10 minutes and destained in running deionized water for 10 minutes. After the staining, the whole-mount mammary glands were photographed and stored in 80% ethanol at 4°C. For histological sections, the mammary glands were embedded in paraffin by a MVP (modular vacuum processor, Instrumental Laboratory) following the acetone washes. Ten-micrometer sections were collected and regressively stained with hematoxylin and eosin (Chisaka and Capecchi, 1991). RNA whole-mount in situ hybridization RNA whole-mount in situ hybridization on E9.0-E12.5 embryos was performed as described (Carpenter et al., 1993; Manley and Capecchi, 1995). The concentration of digoxigenin-UTP labeled RNA probes in the hybridization mixture was approximately 0.4 ug/ml. Alkaline phosphatase-conjugated anti-digoxigenin Fab fragment was used at a 1 :3000 dilution. The templates for Hoxa9, Hoxb9, and Hoxd9 RNA in situ probes were as follows: a 0.6 kb BgnI-EcoRI fragment from Hoxa9; a 0.5 kb fragment in the 3' untranslated region (UTR) of Hoxb9; a 0.8 kb PstI-EcoRI fragment from the 3' UTR of Hoxd9. RT-PCR (reverse transcription-polymerase chain reaction) Total RNA was isolated from E12.5 embryos (without heads), adult liver biopsies, and adult mammary glands of different stages using the TRIzol reagents (BRL) according to the manufacturer's guidelines. The RNA samples were treated with RNase free DNase (BMB) at 37°C for 40 minutes, followed by phenol­chloroform extraction and ethanol precipitation. One and a half ug of total RNA was used in the reverse transcription reaction with poly(dT) 12-1 8 primers. A second set of mixtures was prepared with the same amount of RNA as the first set, and with all the reagents except for the reverse transcriptase. The second set of samples was used to demonstrate that the amplification in the RT-PCR was solely from the cDNA but not from any DNA contamination. The amount of reverse transcription mixture used in the PCR reaction was equalized by the amplification of the p-actin cDNA. PCR was 25 performed as follows: 5 minutes at 94 DC, then 30 cycles of: 95DC, 20 seconds, 65DC, 20 seconds, 72DC, 20 seconds, then a 5 minutes extension at 72DC. Primers used and product sizes are: j3-actin forward: 5'CTCCATCGTGGGCCGCTCTAG3', reverse: 5'GTAACAATGCCATGTTCAAT3', 137 bp band; Hoxa9 forward: 5'CGCTGGAACTGGAGAAGGAGTTTCTG3', reverse: ATCCTGCGGTTCTGGAACCAGATC3', 123 bp band; Hoxb9 forward: 5'CAGGGAGGCTGTCCTGTCTAATC3', reverse: 5'CTTCTCTAGCTCCAGCGTCTGG3', 177 bp band; Hoxc9 forward: 5'GCAACCCCGTGGCCAACTGGATCC3', reverse: 5' AAGACGGTGGGCTTTTCTCT A TCTTGT3' , 376 bp band; Hoxd9 forward: 5' AGCGAACTGGATCCACGCTCGCTCCA3', reverse: 5'GACTTGTCTCTCTGTAAGGTTCAGAATCC, 196 bp band. Results Mutations in Hoxa9. Hoxb9, and Hoxd9 synergistically affect the development of the axial skeleton Fusions between the first and second pairs of ribs were found in Hoxb9 homozygous mutants (Chen and Capecchi, 1997), but not in AABbDD, aaBBDD, AABBdd, AaBbDd, or aaBbDd mice. However, in 50% of the AABbdd, 67% of the AaBbdd and 100% of the aaBbdd mice studied, first and second rib fusions were observed. In addition, a more severe rib fusion phenotype involving the first three pairs of ribs was seen in 100% of the aabbdd mice, 13% of the aabbDd mice, and 25% of the Aabbdd mice (Table 3.1 and Fig. 3.1). All the aabbdd mice studied have eighth ribs articulating with the sternum (Table 3.1 and Fig. 3.1). Taken together, these data suggest that all of these genes are involved, but of different relative importance, in the patterning of the thoracic region. Hoxb9 is the most indispensable among the three in the upper thoracic region, since among the three individual homozygous mutants, only Hoxb9-/- has rib fusions in the upper rib cage. Since some AABbdd and some AaBbdd mice have rib fusions, but none of the aaBbDD or aaBbDd mice have rib fusions, the disruption of Hoxd9 appears to cause more severe phenotypic 26 Table 3.1. Summary of skeletal phenotypes in Hoxa9. Hoxb9. and Hoxd9 triple mutants: Thoracic region (number of animals) Forelimbs (number of arms) Genotype Rib fusions articulating eighth humerus defects zeugopod defects rib no £1-2* rl-3** no yes no minor mod- severe no yes erate AaBbDd 2/2 212 4/4 4/4 aaBbDd 515 4/5 115 6/10 4110 10/10 AabbDd 112 112 2/2 2/4 214 4/4 AaBbdd 113 2/3 213 113 4/6 2/6 6/6 Aabbdd 6/8 2/8 3/8 5/8 12116 4116 15116 1116 aaBbdd 4/4 2/4 214 8/8 5/8 3/8 aabbDd 7/8 1/8 3/8 5/8 9116 7/16 16116 aabbdd 8/8 8/8 16116 5/16 11116 * rl-2, rib fusions of the first two pairs of ribs. * * rl-3, rib fusions ofthe first three pairs of ribs. Figure 3.1 Targeted mutations in Hoxa9, Hoxb9, and Hoxd9 affect the patterning of the thoracic skeleton. A, a rib cage from a wild-type control; B, a rib cage from a triple homozygous (aabbdd) mouse. In all the aabbdd mice studied, the rib fusions in the upper thoracic region involved at least the first three pairs of ribs (as indicated by r1, r2, and r3). All of these aabbdd mice had the eighth pair of ribs articulating to the sternum. 28 29 +1+ consequence in the upper thoracic region than Hoxa9 (Table 3.1). Thus, the functional importance of these genes in the patterning of the thoracic region forms a hierarchy with the order of Hoxb9>Hoxd9>Hoxa9. Functional and structural defects are apparent in the limbs of the triple mutants Neither single mutants for Hoxa9, Hoxb9, or Hoxd9, nor double mutants for Hoxa91Hoxb9 have any apparent limb defects. Malformations in the humerus, such as decreased length and reduction of the deltoid tuberosity, have been observed in double mutants of Hoxa91 Hoxd9 (Fromental-Ramain et ai., 1996). All aabbdd mice have overtly visible deformation in their shortened forelimbs. In many cases their arms were misaligned causing them to walk on the side or even the back of their palms (Fig. 3.2A, B). The grip strength of the forelimbs of these aabbdd mice, as measured by a grip strength meter (SD Instruments, San Diego, CA), is only about 50% to 60% of the average value of the wild-type controls. The wild-type controls we used in this experiment and other experiments in this paper are age-matched F 1 offspring of 129/SvJ X C57BLl6J (we will simply call them wild-type controls in the rest of this chapter). There are two reasons for using these FI controls. First, the mutant colony is in a 129/SvJ and C57BLl6J hybrid background. Second, wild-type littermate controls for the aabbdd mice are nearly impossible to get, considering the low probability (64 X 64) of having an aabbdd mouse and a wild-type mouse in the same litter from a triple heterozygous intercross. Abnormalities were found throughout the forelimb skeletons of the aabbdd mice (Fig. 3.2C, D). The medial comer of the scapula blade is enlarged. The humerus is affected the most severely. It is much shorter than normal and appears to be twisted in some cases. The deltoid tuberosity is completely missing in all humeri from aabbdd mice. In about 70% of the forelimbs from aabbdd mice examined, various defects were found in the elbow joint, the radius, and the proximal portions of the ulna. The radius was frequently shorter, more curved, and attached to a more proximal position of the humerus, creating a bigger gap between the radius and the ulna (Fig. 3.2C, D). 30 Figure 3.2 Triple homozygous mutants have overt forelimb defects. A is a wild-type mouse. B is a triple homozygous mutant. All of the triple homozygous mutants have deformed forelimbs. Some of them walk on the side or even the back of their palms. The overt defects appear to be largely due to skeletal abnormalities in their forelimbs. C, a wild­type control; B, a triple homozygous mutant (See text for details; s: scapula, h: humerus, r: radius, u, ulna). The arrow points to the ectopic ossification. 31 32 Interestingly, ectopic ossification was found in some adult forelimb skeletons of aabbdd mice (Fig. 3.2D). It remains to be determined whether the ectopic ossification came from the remodeling of the affected skeletal elements, like the humerus, or from the ossification of other tissues, like the tendons. Comparison revealed a general trend of increase in the penetrance and severity of the limb phenotypes in the triple mutants from AaBbDd to aabbdd (Table 3.1). The contributions of the three individual mutant alleles to the limb phenotypes, however, are not equivalent. Genotype aabbDd and genotypes Aabbdd and aaBbdd have the same number of mutant alleles, but differ in the fact that aabbDd mice are heterozygous for Hoxd9, whereas Aabbdd and aaBbdd mice are homozygous for Hoxd9. Nine out of 16 humeri from aabbDd mice have no obvious defects, while all the limbs from the other two genotypes showed various degrees of defects in the humerus. Even 100% of the AaBbdd mice have humerus phenotypes. Furthermore, the humerus defects in AaBbdd, Aabbdd, and aaBbdd mice are generally more severe than the defects in the affected limbs of aabbDd mice (Table 3.1 and data not shown). It is obvious that the disruption of Hoxd9 leads to much more severe consequences in the humerus than does the disruption in Hoxa9 or Hoxb9. Comparison also revealed that the disruption of Hoxa9 is slightly more damaging to the normal patterning of the humerus than is the disruption of Hoxb9. It appears that these three genes form a functional hierarchy in the patterning of the limb skeleton in the order of Hoxd9 > Hoxa9 > Hoxb9. The humerus is affected the most among all the structures in the forelimb. This is consistent with the speculation that the position of the primary functions of a 5' Hox gene along the proximo-distal axis of the limb is colinear with the position of the gene in the complex (Davis et al., 1995; Rijli and Chambon, 1997). The shortened and twisted humerus in aabbdd mice affects the geometry of the entire forelimb, while the shorter and frequently more curved radius exacerbates the problem. We believe that the skeletal defects cause, though not solely, the walking defects and other functional defects in the forelimb of these mutants. Abnormalities in the soft tissues of the forelimb of these mutants have not yet been found. About 10% of the aabbdd mice also have obvious walking defects in their hindlimbs. The number of weaned aabbdd 33 mice observed is about 56% of the number predicted from the parental genotypes, indicating a moderate decrease in the viability of the aabbdd mice. Decrease in the viability of mice of other triple mutant genotypes is not apparent (data not shown). Fertility of the male aabbdd mice is slightly affected The fertility of the male triple mutants, except for aabbdd, is normal. Three out of 15 aabbdd males are sterile, as demonstrated by the absence of vaginal plug and/or pregnancy after a period of at least 80 days with at least 2 fertile females. The fertile aabbdd mice are generally not as active sexually as wild-type males or males with fewer mutant alleles, as demonstrated by fewer vaginal plugs or pregnancies resulting from matings between the aabbdd mice and fertile females. To date, no correlation between the severity of the limb phenotypes and the reproductive performance has been found. The fertility of the female triple mutants is affected to various degrees in accordance with their genotypes In our facility, an Fl female of 129/SvJ X C57BU6J, on average, has 8.2 newborns per litter among which 7.5 pups survive to weaning. The average number of weaned pups resulting from a pregnancy decreases dramatically when more and more mutant alleles are added. Most significantly, only three pups have been weaned (from one Aabbdd female) from 58 pregnancies without human intervension in Aabbdd, aaBbdd, and aabbdd females (Fig. 3.3A). Unlike some other mutants (Ormandy et ai., 1997), this problem is not restricted to first pregnancies (Fig. 3.3B). Pups born to these females usually died shortly after birth. Only a small number of these pups died more than 2 days after birth. Many dying pups were cannibalized by their mothers. The decrease in the average number of newborns per pregnancy in the triple mutants may result, at least partly, from the fact that many dying pups were cannibalized before being documented. Examination of the recovered dead pups showed signs of dehydration and malnutrition with no or very little milk in their stomachs. 34 Figure 3.3 The average nwnber of pups weaned from a pregnancy (first and subsequent pregnancies) decreases dramatically when the females carry a greater number of mutant alleles (A). This decrease is not restricted to first pregnancies (B). In (B), for each maternal genotype, the number before the comma is the number of first pregnancies studied; the number after the comma is the number of later pregnancies studied. Under the care of their own mothers, pups born to Aabbdd, aaBbdd, and aabbdd females essentially fail to survive due to apparent malnutrition. However, a fairly high percentage of pups survive if switched to the care of lactating wild-type females (C). 35 IoTj CD S .~.- CD D OQ CD :::J Z 0 ('J) q ~ 'i:j 00 - CD 8 ".-..... (I) :::J - $:::J: S ~ 0"' CD ('J) .., ~ ::I 0 ('J) ~ 0.. 'i:j "0 @ ~ "0 OQ en :::J § .(..J.. . CD til "-' Fl+I+(n=9) AaBbDd(n=lO) aaBbDd(n=3 AabbDd(n=49) AaBbdd(n=14) aabbDd(n=52) Aabbdd(n=33) aaBbdd(n=14) aabbdd(n=11) Average # of offspring per pregnancy o - N W ~ ~ ~ ~ 00 ~ > w ~ -- Z Z ('t) ('t) 0~0 " 00~ " a'("'t)I j ::I ::I e- ('t) '"0 '"0 O'q ('t) ('t) ('t) "-"-I ""I ::l ~ 0 s:.:> V'l ..... ..... ".<..!.. ('t) ""I "'t:j '"('0t) a 0 -p--.- ::l CJ l!lIl s::: g. ~ ~ ('t) ('t) ('t) § ""I § 0 ('t) -. (p't.) . p.. '"0 '"0 '"0 ~ c: c: O'q '"0 '"0 ::l '"0 '"0 s:.:> ::l ('t) ('t) (') -""I ""I ..... ~ ('t) .s:..:.>. .V..'.l. "V-'l" ('t) ""I "'t:j "'t:j 0 0 AaBbDd(n=6,4 ) aaBbDd(n=13,18) AabbDd(n=23,26) AaBbdd(n=5,9) aabbDd(n=30,22) Aabbdd(n=18,15) aaBbdd(n=6,8) aabbdd(n=7,4) Average # of offspring per pregnancy ...... tv !..;.) ..j'::>. Vl 0\ ! ! ! ! ! ! -l 00 I I tJj VJ -..J 38 c 9 8 7 >. u § bs::I ) 6 e 0.. I-< (\) 0.. 5 bI) s:: "R ~ tj..j 4 0 ...... 0 '#: (\) 3 ~ I-< (\) ~ 2 1 0 Flf+l+ Aabbdd aaBbdd Aabbdd (n=9) (n=33,f=8) (n=14,f=I) (n= 11 ,(;=4) Female genotype (n: # of pregnancies; f=# of fostering) ~ Newborn per PG I Weaned pup per PG o Newborn per fostering II Weaned pup per fostering Dead pups recovered from mothers of Aabbdd, aaBbdd, and aabbdd can be of any genotype predicted from the parental genotypes. No drastic overrepresentation of any genotype over the rest has been observed (data not shown). For comparison, we call AabbDd, aaBbDd females crossed with aabbdd males as type I matings, while we call Aabbdd, aaBbdd females crossed with males with five or fewer mutant alleles in these three loci and aabbdd females crossed with males with four or less mutant alleles as type II matings. Pups born from type I matings, as a group, have equal or more mutant alleles than pups born from type II matings. However, the majority of the former survived to weaning whereas only three pups from the latter did without human interference. These data suggest that the maternal genotype is much more important than the offspring genotypes in determining the chance of survival for the offspring. This is further supported by the results of fostering experiments. In the fostering experiments, pups born to Aabbdd, aaBbdd, and aabbdd mutant females were switched to and cared for by lactating wild-type females. About 65% of the pups born to these mutant females thrived under the care of the foster females (Fig. 3.3C). This number is an underestimate, since it was not always possible to reach the starving pups of mutant mothers immediately after birth and quickly find a suitable fostering female. Rapid increase in the volume of milk in the stomach was observed in pups that eventually survived. Apparently, pups of these mutant mothers are viable, if provided with appropriate nutritional sources, namely milk from a foster female. Therefore, the fatality of the mutant pups with their own mothers is not due to the pups' genotypes or any irreversible damage inflicted on them during pregnancy or parturition. The ability of these newborns to obtain sufficient milk from the foster females further leads to the speculation that qualitative or quantitative changes in milk production in the mutant mothers are responsible for the fatality of the newborns. In reverse fostering experiments, we removed the pups born to Aabbdd, aaBbdd, and aabbdd females and substituted newborn (from 1-4 day old) Swiss Webster pups. All the wild-type pups switched to the mutant females were healthy and had a full stomach of milk at the time of the switch. However, all of them showed signs of dehydration and many of them died in the first couple of days after the switch. After a few days, the surviving pups were grossly runted. Their bodyweights were 39 only 30%-60% of those of their littermates that remained with their wild-type mothers. In one reciprocal switch experiment, four 2-day-old wild-type pups were switched to the care of an Aabbdd female whose newborn pups were put under the care of the wild-type pups' mother. Eight days later, one surviving wild-type pup had gained less than 0.5 gram and weighed only 23.8% of the average weight of its littermates who remained with their own mother, and only 59% of the weight of an eight-day-old pup of the Aabbdd female switched to the care of the wild-type mother. This pup, along with about half of the total number of wild-type pups switched to the care of post­partum Aabbdd, aaBbdd, and aabbdd females survived to weaning. The survival of some wild-type pups under the care of these mutant females, together with our observation of the normal maternal behavior of the females show that abnormal maternal behavior is not the cause for the fatality of the pups born to these mutant females. Since a portion of wild-type pups with some degree of post-natal development survived under the care of the Aabbdd, aaBbdd, and aabbdd females, milk from these mutant females is not likely to have drastic qualitative changes, such as lack of essential nutrients or the presence of toxic components (as seen in the mutant "toxic milk" for example). It appears that a quantitative reduction of milk production in these mutant mothers caused the fatality of their own pups and the grossly retarded development of the surviving wild-type pups under their care. The survival of these wild-type pups also suggests that sufficient milk supply is much more critical for the survival of the neonatal pups than for pups with some post­natal development. In one experiment, pups born to a Aabbdd female were fostered for a few days by a wild-type female then switched back to the care of the mutant mother. They survived to weaning, indicating that pups with some postnatal development, mutant or wild-type alike, have a better chance of survival than neonatal pups under the care of the mutant mothers. These data suggest that insufficient milk production in these mutant females is the primary cause for the fatality of their pups. 40 Mammary gland hypoplasia appears to cause the reduced milk production Since reduction in the quantity of milk production appears to be the cause for the inability of these mutant females to raise pups, there could be morphological changes in their mammary glands. We isolated mammary glands from wild-type controls, Aabbdd, aaBbdd, and aabbdd females. The mammary glands from the mutant females, when compared to the wild-type controls, showed hypoplasia of varying severity. Around parturition, the epithelial ductal system in the wild-type gland occupies nearly the entire fat pad, whereas the mutant ductal system has considerably less branching and does not fill the fat pad. The number of lobulo­alveolar structures in the mutant glands is greatly reduced (Fig. 3.4A, B). In many cases, the morphological characteristics of these mutant mammary glands resemble those of the mammary glands from wild-type mice in mid- to late pregnancy. It is possible that the mutant mammary glands with less branches and alveoli cannot produce as much milk as the wild-type mammary glands. Therefore, the reduction of milk production in mutant females is likely caused by mammary gland hypoplasia around parturition, but not by neurological or other physiological conditions. It is demonstrated clearly on the sections that the lobulo-alveolar structures in the mutants are not just reduced in numbers, but also different in cellular appearance. There is a high amount of intracellular fat droplets in the alveolar cells of the mutant glands (Fig. 3.4A, B, C, D). In fact, this resembles the morphology of alveolar cells in wild-type females in mid- to late pregnancy. Therefore, the hypoplasia may be the result of a delay of growth and differentiation in the mutant mammary glands. The morphological evaluation revealed that the mammary glands from the 12.5 day-pregnant aabbdd female are significantly underdeveloped as compared to the stage-matched controls (Fig. 3.4E, F). The mutant mammary glands are thinner, with less branching in the epithelial ductal system. Under higher magnification, it is clear that lobulo-alveolar structures are present, but with less branching and in reduced numbers. In the wild-type mammary glands, the lobulo-alveoli already cover the entire length of the branches, whereas in the mutant mammary glands, the lobular­alveoli cover primarily only the regions near the end buds (Fig. 3.4G, H). 41 Figure 3.4 Hypoplasia is evident in the mammary glands of the Aabbdd, aaBbdd, and aabbdd mutant females during pregnancy and around parturition. E and F are whole mount preparations; A, B, G, and H are sections under low magnification; C, D, I, and J are sections under high magnification. E, G, and I are from the fourth mammary glands of l2.5-day pregnant wild-type females. F, H, and J are from the fourth mammary glands of l2.5-day pregnant aabbdd females. A and C are from the fourth mammary gland of a wild-type female shortly postpartum. B and D are from the fourth mammary gland of an aabbdd female shortly postpartum. 42 43 Under high magnification, we can see that although lobulo-alveolar in mutant mammary glands are reduced in number, they do not seem to have abnormal cellular morphological characteristics (Fig. 3.4I, J). Whole-mount morphological comparison of mammary glands from mutant virgins and wild-type virgins did not reveal any noticeable difference (data not shown), though this does not exclude the possible existence of cellular lesions in these mutant mammary glands. The mutant mammary glands may have intrinsic defects affecting their ability to proliferate at a high rate during pregnancy and early lactation. Genes in the ninth Hox paralogous group may directly control the development and adult functions of the mammary gland Mammary gland hypoplasia observed in these mutant females is likely to be a direct result of the loss of functions of the genes in the ninth paralogous group. Hoxb9 and Hoxd9 expression was observed in the mammary gland primordia in E12.5 embryos using RNA in situ hybridization (Fig. 3.5A, B). Cells expressing Hoxb9 or Hoxd9 form circles surrounding the primitive nipples. Interestingly, this expression pattern in the mammary primordia is reciprocal to that of Bmp-2. At this stage, Bmp-2 is expressed in the mammary epithelia located in the center of the circle of Hoxb9 or Hoxd9 expressing cells (data not shown). The expression of these two Hox genes is in the mesenchyme that is believed to have inductive functions. Very weak signals surrounding the primitive nipples were found when a Hoxa9 in situ probe was used in E12.5 embryos. This indicates that the expression of Hoxa9 in the mammary tissues may follow a different time course. Alternatively, the Hoxa9 probe may not be as sensitive as the other two probes used. RT -PCR was carried out to investigate the expression of these Hox genes in adult mammary glands. The amplification of fJ-actin was used as a control for quality and quantity of templates in each sample. The expression of Hoxa9, Hoxb9, and Hoxd9 was found in mammary gland samples from 8-week-old virgin, 6.5 day pregnant, 12.5 day pregnant, 2 day lactating, and 7 day lactating females, as well as E12.5 embryos, but not in liver that serves as a good negative control (Fig. 3.5C). 44 Figure 3.5 RNA whole-mount in situ hybridization on EI2.S day embryos (A: Hoxb9; B: Hoxd9; Arrows point to the dense mesenchyme surrounding the primitive nipples). C shows the results of the RT-PCR analyses with templates from various sources (V: mammary gland of a virgin mouse, P6.S: mammary gland from a 6.5-day-pregnant mouse, P12.5 mammary gland from a 12.5-day-pregnant mouse, L2: mammary gland from a 2-day­lactating mouse, L7: mammary gland from a 7-day-Iactating mouse, E: E12.5 embryos without heads, Lvr: adult liver, N: no template control). 45 46 The expression of Hoxa9, Hoxb9, and Hoxd9 in embryonic mammary gland primordia and in adult mammary gland tissues suggests that they have direct functional roles in the mammary gland, which, if disrupted, can cause hypoplasia of the gland. Discussion Functional redundancy and hierarchies of Hox genes The phenotypic consequences of the individual disruptions of Hoxa9, Hoxb9, and Hoxd9 are limited. Disruption of Hoxa9 causes the transformation of Ll to T14; disruption of Hoxb9 causes fusions between the first and second pairs of ribs; disruption of Hoxd9 causes transformation in the sacral region (Fromental-Ramain et al., 1996; Chen and Capecchi, 1997; Reynolds and Capecchi, unpublished data). Studies of the compound mutants clearly show that the functions of these genes are not limited to the regions that are affected in the individual mutants. Defects in the limbs and in nursing ability have not been found in any of the single mutants, but were observed in the double and more clearly in the triple mutants (Fromental-Ramain et aI., 1996; Chen and Capecchi, 1997; this Chapter). Nearly all the phenotypes observed in the single mutants became more severe and/or had higher penetrance, when additional mutant alleles in the paralogous genes are present. Obviously, these genes have functions in regions showing mutant phenotypes. However, the relative importance of these genes, as manifested by their contributions to the mutant phenotypes, varies in different regions, indicating that these genes form distinctive functional hierarchies in different regions. As shown in the results section, in the upper thoracic regIOn, the functional importance follows the order of Hoxb9>Hoxd9>Hoxa9. In the patterning of the forelimbs, the order changes to Hoxd9>Hoxa9>Hoxb9. The same order applies to the determination of the nursing ability of the triple mutant females. Functional hierarchy of Hox genes on the Hoxd cluster was described in limb development where Hoxd13 appears to have a prevalent role over other 5' Hoxd genes through binding site occupancy or cofactor titration (Zakany and Duboule, 1996; van der Hoeven et al., 1996). Similar mechanisms may be used in the establishment of the functional hierarchies for the ninth paralogous group. Other mechanisms may rely on the difference among these genes in the timing 47 of expression, the local concentration, or the efficiency of transactivating downstream genes. The high degree of genetic redundancy found in the Hox paralogous genes conceivably arose from the quadruplication of the ancestral Hox cluster. Truly redundant genes are evolutionarily unstable. If divergence in functions does not occur in the quadruplicated genes, some of the redundant members would be lost in evolution by the accumulation of mutations. The absence of some Hox genes (such as Hoxa8 and Hoxa12) is likely the result of such a process. Defects observed in individual mutants for Hoxa9, Hoxb9, and Hoxd9, which would in some way affect the fitness of the carrier, prove that these genes are not truly redundant. The compound mutant studies have shown that all three genes have functions in limb patterning. However, the lack of phenotypes in the limbs of the individual mutants indicates that each of them is dispensable when its paralogous genes are normal. This is achieved either by complementation from the redundant functions of its paralogs existing in the wild-type situation, or by the activation of alternative pathways not used under normal circumstances. If the functions of these three genes are truly redundant regarding limb patterning, why didn't one or more of these genes lose their functions in limb patterning during evolution? One possibility is that lesions disrupting the limb patterning function are very likely to destroy other nonredundant functions of the gene. Besides patterning the developing embryos, Hox genes have functions in the mammary tissue in adult mammals The temporal and spatial colinearities of the expression of the clustered Hox genes result in a series of highly ordered combinations of Hox expression along the antero-posterior (AP) axis. The hypothesis that the Hox code, the specific combination of Hox expression, determines the regional (segmental) identity along the AP axis, has support from studies of the disruption or ectopic expression of some Hox genes. Although similar Hox codes may exist along the proximo-distal axis of the appendicular structures, the functions of Hox genes in these structures have been better explained in terms of controlling the localized growth and/or recruiting cells that are involved in the formation of the new structures (Davis and Capecchi, 1995). 48 Similar growth controlling functions of Hox genes in developing embryos may have been utilized in the adult mammary glands, especially during pregnancy and early lactation when the mammary tissue undergoes rapid proliferation. Mammary gland hypoplasia observed in compound mutant females during pregnancy and postpartum suggests that these Hox genes promote cellular proliferation of the mammary tissue under normal conditions. Consistent with this speculation, abnormal activation of murine Hoxa9 and human HOXA9 have been linked to some myeloid leukaemias, in which the deregulated Hox gene may have caused the uncontrolled proliferation of the leukocytes (Blatt and Sachs, 1988; Nakamura et ai., 1996; Borrow et ai., 1996). Of particular interest, Hoxal overexpression has been detected in some murine mammary gland cancers (Friedmann et ai., 1994). This is consistent with our observation that loss-of-function mutations in a subset of Hox genes caused hypoplasia of the mammary gland epithelia. If the proliferation-promoting function of the Hox genes proves to be responsible for these cancers, understanding the molecular pathways involved may have direct therapeutical indications for the treatment of these cancers. The absence of obvious hypoplasia in the mammary glands of the virgin compound mutant females indicates that the functions of these genes may be more important during pregnancy and early lactation when cellular proliferation is at a level comparable to that found in rapidly growing embryos. Expression of Hoxc9 was also detected in the adult mammary glands (data not shown). It may also playa role in the control of the mammary gland growth. Mutants homozygous for all four genes in the ninth paralogous group may have even more severe problems in their mammary glands. However, such a mutant may be very difficult or even impossible to obtain due to the likely reduction in fertility and viability of the intermediate genotypes on the basis of our preliminary results (Boulet, Chen and Capecchi, unpublished data). Expression of Hoxc6 in the adult mammary gland has been reported (Friedmann et ai., 1994). It is possible that Hox genes outside the ninth paralogous group may also have functional roles in the control of mammary gland development and adult functions. However, mammary gland defects have not been directly demonstrated in previous studies of Hox mutants. This could be due to two reasons. 49 First, the putative functions of some Hox genes in the mammary gland may be redundant, so that the disruption of one gene may not be sufficient to cause noticeable defects in the mammary gland. Second, the potential mammary gland defects caused by some other Hox genes may be masked by other defects, such as early lethality. Therefore the discovery of the functions in the mammary gland for Hoxa9, Hoxb9, and Hoxd9 could mean either that they are more important than other Hox genes in the mammary gland, or that the effects of the combination of the disruptions in these three genes cause obvious mammary gland defects that are not masked by other defects. The normal expression of these Hox genes in the stroma of the mammary gland may control the expression of some growth factor or paracrine genes. These growth factors or paracrines, in turn, regulate the growth and differentiation of the epithelia through the effects of their receptors in the epithelia. Many studies have shown that hepatocyte growth factor (HGF), a mesenchymal- or stromal- derived multipotent factor, mediates epithelial-mesenchymal interaction. The mammary gland is one of the systems whose development requires intimate interaction between epithelium and mesenchyme. Both HGF and its receptor, c-met are expressed in the adult mammary glands. In cell or organ cultures of mammary gland, HGF promoted branching of the epithelial ductal system (Niranjan et a/., 1995). We are currently investigating whether the growth controlling function of the Hox genes in the mammary gland works through the HGF, c-met pathway. Acknowledgments We thank O. Chisaka, J. Goddard, L. Reynolds, C. Peterson, and D. Spyropoulos for their contributions in generating the triple mutant colony for Hoxa9, Hoxb9, and Hoxd9; M. Allen, S. Barnett, C. Lenz, G. Peterson, M. Wagstaff, and 1. Hayes for technical assistance; C. Daniel for suggestions; A. Godwin, D. Wellik, K. Thomas, S. Stadler, B. Ruch, and G. Gaufo for helpful discussion and comments on the manuscript. F. Chen is supported, in part, by a research fellowship from the University of Utah. 50 CHAPTER 4 Hoxa7 AND Hoxb7 HAVE REDUNDANT FUNCTIONS IN THE PATTERNING OF THE UPPER THORACIC REGION Abstract Mice were generated with targeted disruptions in Hoxa7 and Hoxb7 respectively. Mice carrying the Hoxa7 mutation are healthy. No defects in the skeleton or other tissues have been found in these mutants. Twelve percent of Hoxb7- /- mutants show first and second rib defects similar to that observed in mice homozygous for Hoxb9 (Chen and Capecchi, 1997). Hoxa7-/- mice and Hoxb7-/­mice are fertile and were used to generate double mutants to reveal potential interactions between these two paralogous genes. Mice homozygous for both mutations have first and second rib defects with higher penetrance and increased expressivity, indicating a functional role of Hoxa7 in the patterning of the upper thoracic region and a synergistic interaction between these two genes. Although Hoxb6, Hoxa7, Hoxb7, and Hoxb9 have distinctive anterior expression limits in axial mesoderm, the disruptions of these genes all cause (or contribute to) the first and second rib fusions. A hypothesis is presented to explain these data and the observation that axial defects in these and other Hox mutants are concentrated along the axial column at points of transition between vertebral types. Introduction The mammalian Hox complex contains 39 genes distributed on four linkage groups designated as Hoxa, b, c, and d. Based on DNA sequence similarity and the position of the genes on their respective chromosomes, individual members of the four linkage groups have been classified into 13 paralogous families. Interestingly, in a given cluster, a 3' Hox gene is activated prior to and in a more anterior region of the embryo than its 5' neighbor. This correlation between the expression pattern of the Hox genes and their relative chromosomal positions is called temporal and spatial colinearity (Duboule and Dolle, 1989; Graham et al., 1989; Duboule, 1994; Capecchi, 1997). Hox genes encode transcription factors belonging to the Antennapedia homeodomain class. In Drosophila, the homologous genes (Hom-C genes) are used to pattern the developing embryo along its rostro-caudal axis (Akam, 1987; Gehring, 1987). Mutations in some of these genes change the identity of one parasegment into 52 that of a neighboring parasegment (Lewis, 1978). Mutational analysis in the mouse has demonstrated that the Hox genes, alone or in concert with other Hox genes, are also used to regionalize the embryo along its major body axes (Chisaka and Capecchi, 1991; Lufkin et ai., 1991; Chisaka et ai., 1992; Le Mouellic et ai., 1992; Condie and Capecchi, 1993; Dolle et ai., 1993; Gendron-Maguire et ai., 1993; Jeannotte et ai., 1993; Ramirez-Solis et ai., 1993; Rijli et ai., 1993; Small and Potter, 1993; Davis and Capecchi, 1994; Kostic and Capecchi, 1994; Satokata et ai., 1995; Suemori et ai., 1995; Boulet and Capecchi, 1996; Goddard et al., 1996 Barrow and Capecchi, 1996; Chen and Capecchi, 1997; Godwin and Capecchi, 1998). The "posterior prevalence" hypothesis postulates that in a given region, the most posterior expressing Hox gene (the most 5' and most recently activated gene due to the spatial and temporal colinearity), tends to take a dominant role in determining the regional identity (Duboule, 1991; Krumlauf, 1993). This hypothesis, combined with the hypothesis that the identities of individual segments are determined by specific combination of Hox expression (Hox code), predicts that loss-of-function mutations in Hox genes will lead to anterior transformations near the anterior expression boundary of the disrupted gene. Anterior transformations in the axial column near the anterior expression limit of the gene have been observed in many Hox gene loss-of-function mutants (Le Mouellic et ai., 1992; Dolle et al., 1993; Condie and Capecchi, 1993). Accumulating evidence, however, suggests that this is a tendency rather than a rule. Posterior transformations, deletions, additions, and malformations of structures as well as defects in regions other than the anterior expression domains have been often observed in Hox loss-of-function mutants. This can not be satisfactorily explained by the "posterior prevalence" hypothesis (Lufkin et al., 1991; Chisaka and Capecchi, 1991 and 1992; Jeannotte et al., 1993; Ramirez-Solis et al., 1993; Small and Potter, Condie and Capecchi, 1993 and 1994; Davis and Capecchi, 1994; Davis et al., 1995). Therefore, pattern formation along the AP axis in vertebrates appears to be more complicated than that in Drosophila. Herein we describe the phenotypic consequences of disrupting Hoxb 7 in mice, as well as the effects of combining Hoxa7 and Hoxb7 mutations. These mice show defects in the formation of the thoracic skeleton. Interestingly, although Hoxb6, 53 Hoxa7, Hoxb7, Hoxa9, and Hoxb9 have distinctive anterior expression limits along the antero-posterior axis, the disruptions of these genes all contribute to the defects in the extreme anterior end of the rib cage. Furthermore, the defects in Hoxb6, Hoxb 7, Hoxb9, Hoxa71Hoxb7 and Hoxa91Hoxb9 mutant mice are concentrated along the axial column at points of transition between vertebral types. In our hypothesis, disruptions of these Hox genes could initiate regeneration-like adjustments in tissues confronted with discontinuities in axial identity. Morphological abnormalities could be avoided in the position of the initial disturbance of positional cues but become apparent near major body transition zones. The combined effects of Hox functions and the cell cycle length of the cells in the primitive streak could cause the periodic reiteration of developmental abnormalities. Materials and Methods Targeting vector The Hoxa7 targeting vector was constructed by Dusan Kostic (Kostic, 1993). In brief, a neor expression cassette (KT3NP4neopa) was inserted into the homeodomain-coding region (homeobox) of Hoxa7. Three genomic clones containing the Hoxb7 gene were isolated from a lambda bacteriophage DNA library prepared from mouse CC1.2 embryo-derived stem (ES) cells (Rancourt and Capecchi, unpublished results). A 7.5kb HindIII-EcoRI and a 2.6kb EcoRI-SacI subclones were used to construct a replacement-type targeting vector. A ned expression cassette (KT3NP4neopa) was inserted into the homeobox of Hoxb7. Completion of the targeting vector involved flanking the Hoxb7 genomic sequences with the HSVI and HSV2 thymidine kinase (tk) genes. Electroporation and the generation of mutant mice The generation of targeted ES cells with a disruption in Hoxa7 was described by Dusan Kostic (Kostic, 1993). For Hoxb7, the targeting vector was linearized by digestion with NotI and electroporated into Rl ES cells (Deng and Capecchi, 1992; Nagy et al., 1993). ES cells containing a disruption of the Hoxb7 gene were enriched by positive-negative selection (Mansour et af., 1988). DNA samples isolated from 54 colonies of ES cells were digested with EcoRV and probed with a 3' flanking probe (Fig. 4.1). About 2% of the ES cell lines contained the desired Hoxb 7 mutation. The targeted ES cell lines were used to generate chimeric mice that transmitted the Hoxa7 and Hoxb7 mutant alleles to their progeny respectively (Capecchi, 1989 and 1994). Since both Hoxa7-1- mice and Hoxb7-1- mice are fertile (See results), compound heterozygotes for Hoxa7 and Hoxb7 mutations (Hoxa7+1- and Hoxb7+1-) were obtained from crosses between mice homozygous for either the Hoxa7 mutation or the Hoxb 7 mutation. Genotype analysis DNA was prepared from tail biopsies of adult and newborn mice, and from yolk sacs of embryos. Genotypes were determined either by Southern transfer analysis or by amplification of DNA fragments using the polymerase chain reaction (PCR). Hoxa7 genotypes can be determined by a BamHI digestion and a XhoI-BamHI probe. Hoxb7 genotypes can be determined by an EcoRV digestion and a 3' flanking probe (EcoRI-BgnI). The sequences of the PCR primers used for genotyping Hoxa7 and Hoxb7were: Hoxa7 forward primer: 5'GACCCGACAGGAAGCGGG3'; Hoxa7 reverse primer: 5'GTAACTAAAATCAATGAGTCTC3'; neo reverse primer for Hoxa7 and Hoxb7: 5'GTCCAATCAATTGGAAGTAGCC3'; Hoxb7 forward primer: 5'GCCTGACCGAAAGCGAG3'; Hoxb7 reverse primer: 5'CCACTTCATGCGCCGGTTCTG3'. Histology Newborn mice were euthanized by asphyxiation with C02, fixed in 4% formaldehyde in phosphate-buffered saline (PBS) overnight at room temperature (Manley and Capecchi, 1995), and embedded in paraffin according to standard protocols. Ten micrometers serial sagittal sections were collected and regressively stained with hematoxylin and eosin (Chisaka and Capecchi, 1991). Newborn and adult whole-mount skeletons were prepared as described by Mansour et al., 1993. Neurofilament staining in embryos with a monoclonal antibody 2H3 (University of 55 Iowa Hybridoma Center, IW) was performed with minor modifications according to the methods described before (Chisaka and Capecchi, 1991). Results The generation of the Hoxa7 and Hoxb7 mutant mice The structures of the targeting vectors used to disrupt Hoxa7 and Hoxb7 in ES cells are described in Materials and Methods. The insertions of the neD' expression cassettes into the homeoboxes of Hoxa7 and Hoxb 7 terminate the corresponding proteins prematurely and disrupt the homeodomains. Therefore, the insertions render the genes nonfunctional with respect to DNA binding (Fig. 4.1). Southern transfer analyses, using 5' and/or 3' flanking probes as well as internal probes were performed to ensure that no rearrangements of these loci had occurred other than the desired neor insertions into the homeoboxes. Representative targeted ES cell lines were used to produce chimeric males that passed the Hoxa7 mutation and the Hoxb7 mutation, respectively, through the germline (Fig. 4.2). Hoxa7 mutants have no apparent abnormalities Mice heterozygous for the Hoxa7 mutation were intercrossed to produce homozygotes. Adult Hoxa7-1- mice were obtained at a frequency predicted from a Mendelian distribution of mutant and wild-type alleles, indicating no loss of mutant alleles as a consequence of embryonic or post-natal lethality. Mice heterozygous and homozygous for the Hoxa7 mutation appear outwardly normal and animals of both sexes are fertile. Neither gross anatomy nor general histology showed any abnormality in the Hoxa7-1- mice (data not shown). A total of 53 mice (8 +1+, 29 +1-, 16 -1-) were used for skeletal preparation. No apparent defects in the skeletons of +1- or -1- mice were found (Fig. 4.3, Table 4.1). Hoxb 7 mutants have rib fusions in the upper thoracic region Mice heterozygous for the Hoxb 7 mutation were intercrossed to produce homozygotes. Adult Hoxb7-1- mice were obtained at a frequency predicted from a 56 Figure 4.1 Disruption of the Hoxb7 gene. A-C, Diagrams of the wild-type Hoxb7 locus, the targeting vector, and the targeted allele. Hoxb 7 is transcribed from left to right. HB: homeobox, B: BgnI, H: HindIII, Rl: EeoRl, S: Sad, Sp: SphI, X: Xhol, V: EeoRV, TK: Thymidine kinase. 57 A. Hoxb7 genomic DNA HVB S B. Targeting vector _III TK2 v I.. . 6.8 kb HB .. Homologous recombination 3' flanking probe SpB V 1 I I f RI V S II--Pr--I ------i_ ~ J(T31>P4neopa TKI HV B S Sp B V SI RI V C. Targeted allele ....... .1-1 J...I -I.I_J...I __ --L-I ..1-1 ~I __ I-. _--.JI'-IIIIlIIII!III!!!III!l! ~•~. I S RI VB I I II ..... m .. V 3.4 kb I ... V J Vi 00 Figure 4.2 Analysis of the Hoxb7 mutant genotypes. DNA samples isolated from colonies of ES cells were digested with EcoRV and probed with a 3' flanking probe. The 6.8 kb band was from the wild-type allele. The 3.4 kb band was from the targeted allele. 59 60 -6.8 kb -3.4 kb Figure 4.3 Skeletal preparations of the Hoxa7 mutants and the Hoxb7 mutants. A, wild-type control. B, Hoxa7-/-; C, Hoxb7-/-. Hoxa7-/- mice do not have any detectable defects. Twelve percent of Hoxb7-/- mice, including the one shown here, have defective first and second ribs. 61 62 Table 4.1: Summary of the skeletal defects in Hoxa7, Hoxb7, and Hoxa71Hoxb7 mutant mice: Mice with upper Total number of GENOTYPE rib cage defects* mice studied Penetrance Hoxa7+1+ 0 8 0 Hoxa7+/- 0 29 0 Hoxa7-/- 0 16 0 Hoxb7+1+ 0 9 0 Hoxb7+/- 0 34 0 Hoxb7-1- 4 27 15% Hoxa7+1-; Hoxb7+1- 0 13 0 Hoxa7-/-; Hoxb7+/- 1 t 1 9% Hoxa7+1-; Hoxb7-/- 5 11 45% Hoxa7-1-; Hoxb7-/- 12 17 71% I *Includes rib fusion, abnormal rib attachment, disappearance of ribs, and the accompanying sternum defects. 63 Mendelian distribution of mutant and wild-type alleles, indicating no loss of mutant alleles as a consequence of embryonic or postnatal lethality. Mice heterozygous and homozygous for the Hoxb7 mutation appear outwardly normal and animals of both sexes are fertile. A total of 63 mice (9 +/+, 34 +/-, 26 -/-) were used for skeletal preparation (Table 4.1). Three of the 26 Hoxb7-/- mice (12%) tested have abnormal rib patterning in the upper thoracic region, a phenotype very similar to that observed in the Hoxb9 mutants (Fig. 4.3, and Chen and Capecchi, 1997; Chapter 2). Two of these three Hoxb7-/- mice have fusions between the first and the second pairs of ribs. The fused ribs attach to the sternum at the place where the first pair of ribs normally attaches. Rib fusions lead to abnormalities in the segmentation of the sternum, such as a reduction of the number of sternal segments from seven to six. The first pair of ribs in the last one of the three was reduced to small knobs leaving only six pairs of ribs attaching to the sternum. No other skeletal abnormality was found in the Hoxb7+/­and -/- mice. Gross anatomy did not show any apparent defects in the internal organs of these mice (data not shown). The generation of Hoxa71Hoxb7 double mutants Since Hoxa7-/- mice and Hoxb7-1- mice are fertile, compound heterozygotes for both Hoxa7 and Hoxb7 could be obtained from crosses between Hoxa7-/- and Hoxb7-/- mice. These compound heterozygotes appeared outwardly normal and were fertile. Mice of all nine possible genotypes were obtained at the expected Mendelian ratios, indicating that all genotypes, including double mutant homozygotes, were viable. Defects in the Hoxa7/Hoxb7 double mutants Compound heterozygotes (Hoxa7+/-; Hoxb7+/-) do not show any detectable phenotype. After one Hoxa7 allele is added to the compound heterozygotes, one out of 11 (9%) tested had first and second rib fusions. This observation indicates that Hoxa7 has functional roles in the patterning of the upper thoracic region, which were not revealed in the Hoxa7-/- mice possibly due to overlapping functions provided by Hoxb7. Twelve percent of Hoxb7-1- mice have defects in the upper thoracic region. 64 An increase in the penetrance of the first and second rib defects was observed when Hoxb7-1- mice received one or two copies of the Hoxa7 mutant allele (Table 4.1, Fig. 4.4). Approximately 12% of Hoxb7-1- mice show first and second rib defects, while 45% of the Hoxa7+1-; Hoxb7-1- mice and 71 % of the Hoxa7-1-; Hoxb7-1- mice show rib defects (Table 4.1). These observations support the hypothesis that these two genes function synergistically and quantitatively in the patterning of the upper thoracic region. Sagittal and parasagittal sections of newborn Hoxa7-1-; Hoxb7-1- mutant mice reveal a significant reduction in the size and an alteration in the shape of the thymus relative to that of wild-type controls. The histological features of the thymus of these mice, however, appear normal (Fig. 4.5A, B). A very similar phenotype has been described in the study of the Hoxb9 mutants, in which the thymus phe{lotype has been attributed to the reduced space in the upper thoracic region due to rib fusions (Chen and Capecchi, 1997). The smaller thymus observed in the Hoxa71Hoxb7 double mutant mice also appears to be a result of space constraints caused by the first and second rib defects. Thus, it appears to be an indirect result of the disruption of Hox gene functions in this region. E12.5-E14.5 embryos were immunostained with the 2H3 antibody directed against a subunit of the neurofilament protein (Dodd et al., 1988) to reveal possible changes in the pattern of neurons in the nervous system. In one set of experiments, aldan blue was used to stain the cartilage of the embryos after they were immunostained by the 2H3 antibody. The stained cartilage serves as a landmark when comparing the spatial patterning of the nerves. No neuronal defects were apparent in Hoxa7-1-; Hoxb7-1- mutants either in the body wall or the limbs (data not shown). Transverse sections of newborn Hoxa7-1-; Hoxb7-1- mutant mice were stained with hematoxylin and eosin (H&E) to examine the spinal cord, dorsal root ganglia, and other neural tissues at the cellular level. No neural defects were found in the mutants (data not shown). 65 Figure 4.4 Skeletal preparations of the Hoxa71Hoxb7 double mutants. A-D, Ventral view of the rib cages. E-H, Side view of the rib cages. Hoxa7+1-;Hoxb7+1- mice do not have any detectable defects (A and E). Only one out of 11 Hoxa7-1-;Hoxb7+1- mice has defective first and second ribs, the sample shown here is one without first and second rib defects (B and F). Forty-five percent Hoxa7+1-;Hoxb7-1- have defective first