An Evolutionary Switch in Tissue-specific Gene Expression ABUNDANT EXPRESSION OF (YI-ANTITRYPSIN IN THE KIDNEY OF A WILD MOUSE SPECIES*

al-Antitrypsin (AT), one of the major proteinase in- hibitors in mammalian serum, is generally considered to be synthesized exclusively in the liver. We have found that a wild-derived Mus species, Mus caroli, expresses AT mRNA in kidney at levels approaching that in liver; no other mouse, inbred or wild-derived, exhibits this striking property. Liver and kidney mRNAs from M. caroli encode very similar AT poly- peptides that are distinct from that encoded by Mus musculus liver mRNA. In vivo, liver AT is secreted into the bloodstream, while kidney AT, which is processed differently from the liver protein, is excreted into the urine. Analysis of RNA from a hybrid between M. musculus and M. caroli indicates that a cis-acting genetic element may be responsible for the difference in AT expression. Restriction enzyme digestion pat- terns of AT genomic sequences in M. caroli DNA are considerably different from those in M. musculus; in addition, these sequences are undermethylated in liver DNA from M. musculus and in liver and kidney DNA from M. caroli, reflecting the respective patterns of expression. Further studies of the altered tissue specificity of AT expression that is apparent in these two related species should lead to new insights into the nature and evolution of genetic determinants of tissue-specific phenotypes.

addition, these sequences are undermethylated in liver DNA from M. musculus and in liver and kidney DNA from M. caroli, reflecting the respective patterns of expression. Further studies of the altered tissue specificity of AT expression that is apparent in these two related species should lead to new insights into the nature and evolution of genetic determinants of tissuespecific phenotypes.
One feature of gene expression in higher eucaryotes is its tissue specificity; the levels of expression and/or the patterns of regulation of most genes vary in striking fashion from tissue to tissue. While a great deal of information correlating tissue-specific patterns of expression to molecular properties of the respective genes has been accumulating (1, 2), fundamental knowledge about the genetic elements that determine tissue specificity is still lacking. A promising approach t o the problem involves introducing isolated genes into appropriate cells in culture (3) or into animal germ lines (4, 5). Analysis of the expression of the newly acquired genes, and modified derivatives, is certain to shed light on the nature of tissue specificity. Another approach, with which the present paper is concerned, makes use of naturally occurring variants. The existence of closely related strains or species that show alterations in patterns of tissue specificity should provide a model for identifying and studying the genetic elements that are responsible for the altered patterns and that are, therefore, critical in determining the tissue-specific phenotype.
al-Antitrypsin (AT),' one of the major proteinase inhibitors in mammalian plasma, is produced, as are most plasma proteins, predominantly in the liver. In inbred strains of mice (Mus musculus), such as BALB/cJ, AT mRNA accumulates to concentrations of about 6000 molecules/cell in liver and no more than 15-25 molecules/cell in kidney, spleen, or brain (6).' During studies of wild-derived mice, we observed that one species, Mus caroli, expresses abundant amounts of AT in the kidney as well as in liver. In this report we have characterized this novel expression of AT by comparing the liver and kidney AT mRNAs and proteins in laboratory inbred strains of M. musculus and in the wild species M. caroli. In addition, we include initial studies of the AT genes of these animals.

EXPERIMENTAL PROCEDURES
Animals-Inbred strains of M. musculus (C57BL/6J, DBA/2J) were purchased from the Jackson Laboratory, Bar Harbor, M E they were used at around 8 weeks of age. M. caroli mice were obtained from a random-bred colony maintained by Dr. Verne Chapman of this Institute; these mice were originally trapped in Thailand. A hybrid between M. caroli and M. musculus was produced and kindly supplied to us by Dr. J. Rossant of Brock University, St. Catherines, Ontario.
Culture of Liver and Kidney Slices-Liver (perfused with phosphate-buffered saline) and kidneys were excised and cut into cubes 1 mm3 in volume. The tissue minces were washed 3 times with Dulbecco's modified Eagle's medium containing 10% fetal calf serum (50 ml/g of tissue) and incubated in the same medium (20 ml/g of tissue) under an oxygen atmosphere in a shaking water bath at 37 "C. After 30 min, the medium was replaced by the same volume of serum-free medium containing 1/10 of the normal concentration of methionine (3 pg/ml) and 100 pCi/ml of ~-[~~S]methionine (1000 Ci/mmol). After 3 h, the medium was removed and centrifuged first for 5 min at 1000 X g and then for 60 min at 200,000 X g. The supernatant was freezedried following 24-h dialysis against 25 mM NHIHC03, 0.05 mM phenylmethylsulfonyl fluoride. To generate nonglycosylated proteins, kidney minces were incubated in medium containing 0.5 pg/ml of tunicamycin for 2 h prior to the labeling period. Inhibition of glycosylation did not significantly impair the synthesis and export of AT.
Analysis of Proteins-Proteins were separated by two-dimensional polyacrylamide gel electrophoresis (7) and the radioactive patterns were visualized by fluorography (8). Crossed immunoelectrophoresis was carried out as described by Weeke (9) using rabbit antiserum against M. musculus plasma AT (10) (generously provided by Dr. Jack Gauldie, McMaster University, Hamilton, Ontario, Canada) in the second dimension gel. Immunological identification of AT forms was achieved by mixing radioactive samples with nonlabeled carrier AT (urinary proteins of a testosterone-treated M. caroli female) followed by crossed immunoelectrophoresis; the radioactive precipitin lines were cut out, boiled in sodium dodecyl sulfate sample buffer * R. Hill, P. Shaw, P. Boyd, H. Baumann, and N. Hastie (1984) The abbreviation used is: AT, al-antitrypsin.
The proteinase-binding activity of AT was determined on crossed immunoelectrophoresis gels. After electrophoresis, the gels were washed for 2 h with phosphate-buffered saline, overlayed with buffer containing 12s1-laheled trypsin or chymotrypsin, and incubated for 30 min a t room temperature. Following exhaustive washing, the gels were dried and autoradiographed. This assay was specific for AT, as no radioactivity was associated with coprecipitated albumin.
Partial proteolytic digestions of proteins recovered from two-dimensional polyacrylamide gels were carried out with Staphylococcus nureus VU protease within a 15% polyacrylamide gel as described by

Ana1,vsis of RNA and DNA-Total
RNA was extracted by the guanidine-HCI procedure (13). In oitro translation was in a reconstituted cell-free system (14). For Northern blot analysis (15). 15 pg of RNA was fractionated on a 1.5% agarose gel containing 2.2 M formaldehyde. transferred to nitrocellulose, and hybridized to R2P-laheled p1796 DNA (generously provided by Dr. N. D. Hastie, Medical Research Council, Edinhurgh, United Kingdom). This cDNA-containing plasmid was originally isolated by Harth P t al. (6); in subsequent work, sequence analysis demonstrated that p1796 corresponds to rrl-antitrypsin mRNA.' DNA was isolated from nuclei of the indicated tissues (16). Restriction endonucleases were purchased commercially and used according to the instructions of the supplier. DNA fragments were fractionated in 1% agarose gels, transferred to nitrocellulose filters, and hybridized to R2P-laheled p1796 DNA (17)

A ) and kidnev ( H ) of a M . carol1
female were translated in a cell-free system and the products were separated hv two-dimensional gel electrophoresis. C , nn aliquot of the translation mixture from R was treated with A T antibodies and the immunoprecipitate was subjected to electrophoresis. Fluorogr~ms in A and R were exposed for 1U h, that in (' for 1 month. The following spots are indicated: a. actin; A T , (tl-antitrypsin; A(;f'-I and A(;l'-2, Crl-acid glycoprotein-1 and -2: A h . albumin; M('L', major urinary proteins.

FIG. 1. Levels of A T m R N A
in M. caroli a n d M . musculus kidney and liver. A , total kidney RNA was prepared from 8 individual M . caroli females, fractionated hy agarose gel electrophoresis. blotted onto nitrocellulose, and hybridized to "P-laheled plT96 DNA.
R, liver and kidney RNAs were prepared from four individual M. c a d i females or two individual M . musculus females (strain DRA/ 2.J). The RNAs were analyzed for AT mRNA as descrihed in A .

RESULTS
Expression of AT mRNA in M . caroli Kidnq-During analysis of gene expression in wild-derived stocks o f mice, which exhibit a more interesting and extensive arrav o f phenotypes compared to laboratory inbred strains of M . musculus. we observed that kidney mRNA from the species M. carnli e ncodes an abundant pol-ypeptide t h a t is absent from the trans-cq-Antitrypsin Expression lation products of kidney mRNAs from other mouse species. Since the new polypeptide exhibited a charge and molecular weight identical to the precursor form of AT, whose expression in mice had previously been shown to be liver-specific (6),2 we initiated a series of experiments aimed a t definitive identification of this unique expression of AT in M. caroli kidney.
Demonstration of the presence of A T mRNA in M. caroli kidney was facilitated by the availability of a cDNA probe, p1796, corresponding to M. musculus liver AT mRNA (6).2 Northern blot analysis (Fig. 1A) revealed that kidney RNA from several individual M. caroli females exhibited significant levels of a single RNA species that hybridizes to p1796 and that has a length of about 2.0 kilobases, which is identical to that reported for M. musculus (6). Although all animals contained detectable levels of this mRNA, there was substantial variation in its concentration from individual to individual. The source of this variability is unknown. The AT mRNA level in M. caroli kidney was compared to that in M. caroli liver, and to that in M. musculus liver and kidney (Fig. 1R). Abundant amounts of AT mRNA are pres- and urine ( H ) were collected from a testosterone-treated M. caroli female, fractionated hy two-dimensional gel electrophoresis, and stained with Coomassie Blue. Liver and kidneys from the same animal were placed into tissue culture, and newly synthesized proteins were metabolically labeled with "S-labeled methionine (see "Experimental Procedures''); media themselves or AT immunoprecipitates prepared from aliquots of the media were fractionated by two-dimensional electrophoresis and observed by fluorography. C , liver medium; D, kidney medium; E, immunoprecipitate of liver medium; F, immunoprecipitate of kidney medium. The following proteins are indicated: AT, nl-antitrypsin; Ab, albumin; AGP, nl-acid glycoprotein; MUP, major urinary proteins; T/, transferrin. The minor spots to the acidic side of the immunoprecipitated AT in E and F are forms of AT that have been processed hy a modifying activity in the rabhit antiserum preparation. 7 ent in M. musculus liver while none is detectable in kidney; this is consistent with the 400-fold difference previously reported for these two organs (6). The level of AT mRNA in M . caroli liver is similar to that in M . musculus liver; however, in contrast to M. musculus, M . caroli kidney expresses levels of AT mRNA that can be as high or higher than that in liver. Although kidneys from individual animals again show variable amounts of AT mRNA, livers from the same animals did not show such variation, indicating that the levels of this mRNA are independently regulated in the two organs. We conclude from the blot hybridizations that M . c a r d kidney can attain an AT mRNA concentration that is similar to that in M . musculus liver, which has been calculated to be 6 0 molecules/cell (6)  like that of all other mice examined, contains no detectable A T mRNA (6) (data not shown). Analysk of the AT Polypeptide Produced in Vitro and in Viuo-We determined whether M. caroli kidney mRNA encodes an AT polypeptide as in the liver. Cell-free translation products of M. caroli liver and kidney mRNAs were separated by two-dimensional polyacrylamide gel electrophoresis (Fig.   2, A and R). Among the products of both was a protein that has an apparent molecular weight of 42,000, a PI of 5.7, and that is recognized by antibodies to M. musculus AT (Fig. 2C). A proteolytic comparison of these precursor forms is presented below (Fig. 5 ) .
Since M. caroli kidney contains high concentrations of an mRNA that encodes AT, questions arise pertaining to the in Expression 1168 structural and functional properties as well as to the fate of AT synthesized in uiuo. Since the hepatic AT is a secreted protein, we analyzed the spectrum of proteins synthesized and released into the medium by M . caroli liver and kidney slices maintained in tissue culture. A T was identified by immunoprecipitation. The liver-derived form is heterogeneous (Fig. 3, C and E ) , spanning an apparent molecular weight range of 48,000 to 52,000 and pI range of 5.0 to 5.3. The kidney-derived form, on the other hand, differs from that of the liver by its more basis charge and slightly lower molecular weight (Fig. 3, D and F ) . Only the liver form of A T could he detected in serum (Fig. 3A) while the kidney form and traces of the liver form were observed in urine (Fig. 3 R ) . Since other plasma proteins (e.g. albumin and major urinary proteins) also appear in the urine (Fig. 3 H ) , we presume that the presence of the liver form of AT in urine represents normal "leakage" of proteins from the blood stream. These results suggest that A T in M. caroli serum and urine are synthesized in the liver and kidney, respectively.
T o determine with greater sensitivity whether any kidney-derived AT can he detected in the circulation, we performed crossed-immune electrophoretic analysis of plasma (Fig. 4 ) , which indicated the presence of only the faster migrating, liver-derived AT.
Thus, liver AT is secreted into the hloodstream while kidney AT appears to he exclusively excreted into the urine.
T o compare the structures of the various forms of AT in more detail, we performed proteol-ytic mapping (Fig. 5 ) . By this criterion, the AT polypeptide derived from in L lltro ' trans- c a d i female. Aliquots of 0.5 pI of plasma and 3 pl of urine were separated hy crossed immunoelectrophoresis. The second dimension contained 2.5% rahhit antihody to M . musculus AT. Confirmation that the more anodal peak corresponds to the liver form of AT while the more cathodal peak corresponds to the kidney form was made by cutting out the appropriate precipitin hands, eluting the protein, and examining the A T on two-dimensional polyacrylamide gels. sented by lonas 1-5 were treated with protease; those represented hv lanes 6-9 were untreated controls. The fikwre represents a composite of fluorogams exposed for 6 days (Ianc..~ 1 -5 ) or 3 days (lonm 6-9).
caroli kidney and liver are also indistinguishable from each other (lane 4 and 5 ) . The results taken together suggest that the ATs produced by M. caroli liver and kidney are identical. The electrophoretic differences between the processed forms of kidney and liver AT (see Fig. 3) are probably, due, therefore, to differences in post-translational modifications; most likely, the two tissues do not carry out identical secondary glycosylation reactions, such as sialylation.
AT from serum and urine specifically binds both trypsin and chymotrypsin (data not shown). Since these proteases are substrates for AT (18), it appears that AT from each source is functional.
Kidney n,-Antitrypsin Expression of Hybrid Mice-To gain information on the genetic nature of the interspecies difference in kidney AT expression, we have examined an interspe- various RNAs were fractionated on preparative two-dimensional acrylamide gels. Protein spots were recovered and subjected to partial proteolysis (lanes 1-5) as in Fig. 5; lanes 6-9 are undigested controls. Lanes contain the following proteins: lanes 1 and 6, AT of ICR/Ha liver; lanes 2 and 7, AT of M. caroli liver; lanes 3 and 8, AT of hybrid liver; lanes 4 and 9, AT of hybrid kidney; lane 5, actin of hybrid liver. Actin has been included to indicate that there is no spill over of actin into AT. Lunes I, 3. and 4 contain two digestions, which represent the lower and upper half of the AT spot from the preparative gel. were treated with EcoR1, blotted onto nitrocellulose. and hybridized to V-labeled p1796 DNA. Size markers, indicated in kilohases to the h f t of the figure, were HindIII-generated fragments of h DNA ( l a m I ).
cific hybrid between M. musculus strain Ha/ICR and M. caroli (kindly supplied by Dr. J. Rossant) (19). Liver and kidney RNAs from a single male hybrid were translated in vitro, and the products were fractionated on two-dimensional acrylamide gels. Fig. 6A shows that hybrid RNA from each tissue encodes AT, indicating that kidney AT mRNA expression in M. caroli is not a recessive trait. Partial proteol-ytic digestion can be used to distinguish M. musculus from M. caroli AT synthesized in vitro (Fig. 6 4 compare lanes 1 and 2) and enables determining relative expression of the two mRNAs in hybrid tissues. The results of such analysis indicate that hybrid liver RNA ( l a n e 3 ) encodes the M. musculus as well as the M. caroli forms; in contrast, hybrid kidney RNA ( l a n e 4 ) encodes exclusively the M. caroli form. Thus. in kidney, the hybrid produces only M. caroli AT mRNA indicating that the species difference in kidney AT mRNA expression is due to an allele-specific genetic element(s). This suggests, hut does not prove, that a cis-active regulatory element linked to the cY,-antitrypsin structural gene determines the pattern of tissue specificity.
Analysis of AT Genomic Sequences-We examined genomic DNA from M. caroli and M. musculuq using p1796 as a probe in Southern blots. Fig. 7 shows that restriction nuclease EcoRI cleaves the AT DNA of M. musculus kidney into 4 or 5 fragments, depending on the strain; this is consistent with the estimated 4-5 copies of A T genes in the M. (2). T o ascertain whether the patterns of AT expression relate to the methylation status of AT genes, we have compared the HpaII and MspI digestion patterns of DNA from various M. musculus and M. caroli tissue (Fig. 8). As judged by the extent of HpaII digestion, AT genes in M. musculus are hypomethylated in liver relative to kidney and brain DNA, paralleling the pattern of AT expression in these tissues. In M. caroli, AT sequences are hypomethylated in liver and kidney relative to brain, again paralleling the expression pattern. Within a species, no differences were observed in the MspI digestion patterns among these tissues. Thus, the tissue-specific pattern of AT gene expression correlates, in general, with the extent of methylation in both mouse species.

DISCUSSION
In this paper we have described and characterized an unusual expression pattern for AT mRNA in mice. Kidney mRNA from M. caroli, in contrast to other mice, encodes functional and immunologically active AT, which is excreted into the urine. M. caroli is unique in being the only mouse strain or species examined that exhibits this pattern of AT expression.
Although the liver has been shown to be a major site of AT synthesis in a variety of mammals (20-22), production in other tissues cannot rigorously be excluded since comprehensive tissue-specificity studies have not, for the most part, been undertaken. Such has been done, however, in M. musculus mice where it is clear that AT expression in the liver is several hundredfold higher than in other tissues, such as kidney, spleen, or brain (6). Thus, during evolution of the M u s genus, a major switch in the tissue specificity of AT expression has taken place. We have no indication as to what, if any, physiological role AT is playing in the kidney or urine. Since it does interact with trypsin and chymotrypsin, its function may be neutralization of urinary proteases.
Divergence of M. caroli from other M u s species occurred some 5-10 million years ago, as estimated from comparisons of DNA relatedness and protein polymorphism. Several parameters, such as the nature of satellite DNA sequences (23, 241, the extents of molecular and biochemical variation (25), and lack of interfertility (25), serve to emphasize the evolutionary distance between M . c a r d and M . musculus. However, the two species are similar enough to allow generation of viable hybrid animals (19, 26); thus, the two genomes are compatible within the same cell.
Our results (Fig. 6) indicate that a cis-active regulatory element may be responsible for the species difference in kidney AT mRNA expression; this element must be within or close proximity to the AT structural gene. Unfortunately, direct analysis of linkage between the AT structural gene and this putative regulatory element is impossible, owing to the infertility of the interspecific hybrids (19). That tissue specificity of gene expression is determined by genetic information closely linked to the appropriate structural gene has recently been suggested for the genes encoding K chain ( E ) , chymotrypsin ( 3 ) , and insulin (3). Current studies of the A T model are focusing on the isolation and characterization of the structural genes from both species and should lead to identification of the critical cis-active factor determining expression in kidney. The existence of two closely related mice having such a striking switch in the expression pattern, coupled with current gene-transfer technologies (3,4). provides a unique opportunity to isolate genetic elements that determine tissue specificity, to study their modes of action, and to understand their evolution.