Thyroid Hormone-dependent Differential Regulation of Multiple Arginase Genes during Amphibian Metamorphosis*

We have cloned three nonhepatic arginase genes in Xenopus laeuis. The deduced amino acid sequences of the three arginases are almost identical and share about 60% identity with mammalian as well as Xenopus liver arginase. Both the liver and nonhepatic arginase genes are activated early during embryogenesis. The liver arginase gene is strongly expressed in tadpole liver, but weakly in other tissues. In contrast, the nonhepatic arginase genes have the strongest expression in the tadpole tail. During metamorphosis, the liver and nonhepatic arginase genes show distinct regulation patterns. In the intestine, both types of arginase genes are activated during the remodeling period. In the tail, the liver arginase  gene is activated uring  tail  resorption, whereas  the  nonhepatic  ones  are  highly  expressed throughout all stages examined. Finally, in the hindlimb, the liver arginase is up-regulated slightly during development, whereas the nonhepatic ones have low levels of expression until the end of metamorphosis. During 3,5,3’-~-triiodothyronine (TJ-induced metamorphosis, the nonhepatic arginase genes are activated very quickly, whereas the liver arginase gene is a late T, response gene. These differential regulation patterns during normal and T,-induced metamorphosis suggest potential functions for the arginases during tissue remodeling.

Thyroid hormone plays a causative role in the regulation of amphibian metamorphosis and is believed to control metamorphosis by regulating gene expression through its nuclear receptors (Dodd and Dodd, 1976). Both thyroid hormone receptor (TR) CY and p genes have been cloned in amphibians and found to be expressed during metamorphosis Kawahara et al., 1991;Schneider and Galton, 1991;Helbing et al., 1992). In the presence of thyroid hormone, these receptors likely activate andor repress a set of early response genes at the transcriptional level. The products of these early response genes in turn regulate the genes further downstream, leading to the final transformations in different tissues.
Many genes have been found t o be regulated either directly or indirectly by thyroid hormone during metamorphosis (Shi, 1994). Among the earliest and best studied are the genes encoding the five urea cycle enzymes: carbomyl phosphate synthetase, ornithine transcarbamylase, argininosuccinate syn-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
A s aquatic tadpoles are transformed into terrestrial frogs, the animals change from ammonia excretion to mostly urea excretion, and the activities of these urea cycle enzymes are dramatically elevated in the liver (Cohen 1970;Weber, 1967). In Xenopus laevis the liver arginase (argL) gene has been shown to be up-regulated in the trunk region, which includes the liver, after prolonged treatment with thyroid hormone (Xu et al., 1993).
Like the other urea cycle enzymes, arginase has been detected in many tissues other than liver (Jackson et al., 1986;Glass and Knox, 1977;Herzfeld and Raper, 1976;Stewart and Caron, 1977;Spector et al., 1983;Skrzypek-Osiecka et ai., 1983). The expression of argL in these tissues could contribute in part to the wide distribution of the enzyme. More importantly, however, biochemical and immunological evidence from these and related studies in mammals suggests the existence of nonhepatic arginases. Up to now, none of the nonhepatic arginase genes have been cloned and their regulation and function remain unclear.
We report here for the first time the cloning of full-length cDNAs for three closely related nonhepatic arginase genes (argl, arg2, and arg3). The cDNAs were isolated from the intestine of X. laevis tadpoles induced to metamorphose by thyroid hormone. We show that like the liver arginase gene, the nonhepatic arginase genes are expressed in many frog tissues. However, compared with the argL gene, the argl-3 genes have distinct developmental, tissue, and thyroid hormone-dependent regulation patterns, suggesting different roles for argl-3 and argL in X. laeuis, especially during metamorphosis.
When indicated, the protein synthesis inhibitors cycloheximide and anisomycin were added at 20 and 25 pg/ml, respectively, for 13 h beginning l h before the addition of T,. This treatment inhibits protein synthesis in tadpole tissues by 99% (Kanamori and Brown, 1992). Total RNA was isolated from dissected organdtissues or whole animals as described (Shi and Brown, 1993). The synchronized embryos were produced by in vitro fertilization (Dimitrov et al., 1993).
cDNA Cloning-A cDNA fragment of about 300 base pairs in size was isolated previously which hybridized to a few mRNA species that were up-regulated by T, in the intestine (clone IU22 in Shi and Brown (1993)). This cDNA fragment was used to screen a Lambda Uni-Zap (Stratagene) cDNA library made of intestinal RNA from stage 52-54 tadpoles treated with 5 I" T, for 18 h.' Several cDNA clones were isolated. Three clones had different sizes of cDNA inserts that are close to the sizes of the mRNA species and were found to encode three highly related arginases. The sequences have been deposited into GenBank with the accession numbers of U08406, U08407, and U08408 for argl, arg2, and arg3, respectively (Fig. 1).
The abbreviation used is: T,, 3,5,3'-~-triiodothyronine.  r g l 4 ) . The numbers are relative to the first nucleotide of zArg2. The lengths of cDNAclones are 2.7,3.1, and 1.6 kilobase pairs for argl, arg2, and arg3, respectively. Each clone has an open reading frame encoding 360 amino acids (shaded area) with the initiation (ATG) and termination (TAG) codons as indicated. argl and arg2 are highly homologous except for a 400-nucleotide deletion (between nucleotides 2551 and 2975 of rArg2) and a few small deletions/insertions in the 3'-untranslated region of argl. The 5'and 3'-end of arg3 are different from arg2 and in addition there are a few small deletionshnsertions in the overlapping 5'-and 3"untranslated regions compared with arg2. All numbers are relative to the first nucleotide of drg2. The arrows indicate the boundaries of the coding regions or the positions where the sequences of the cDNAs diverge from each other. out as described . The probes used included the 300-base pair polymerase chain reaction fragment and the 1.6-kilobase pair arg3 cDNA, both of which hybridized to all three mRNA species, and the cDNA insert for argL (Xu et aZ., 1993). Under the conditions used, argL and a r g l 3 probes did not cross-hybridize. To show that equivalent amounts of total RNA were present, the same or duplicated filters were hybridized with a cDNA probe for ribosomal protein L8 (rpL8), a gene whose mRNA level is independent of treatment with T, or protein synthesis inhibitors and essentially constant in tadpoles (Shi and Liang, 1994).

RESULTS
Cloning and Sequence Analysis of Three Arginase cDNAs in X. Zaevis-To isolate genes that are involved in the remodeling of amphibian intestine during metamorphosis, we previously constructed polymerase chain reaction subtractive cDNA libraries using intestinal RNAs from control and thyroid hormone-treated tadpoles. A fragment of about 300 base pairs was found to hybridized to at least two different genes in X. laevis genome and several mRNA species that are highly expressed in the intestine at the climax of metamorphosis and regulated by T, (Shi and Brown, 1993). To determine the identities and study the function and regulation of the corresponding genes, the polymerase chain reaction fragment was used to screen a dT-primed cDNA library. Three cDNA clones with different cDNA insert sizes were isolated and found to encode proteins homologous to arginases from yeast to human (Figs. 1 and 2). The sizes ofthe cDNAclones, 1.6,2.7, and 3.1 kilobase pairs for arg3, argl, and arg2, respectively, are close to the sizes of the three prominent mRNA bands detected on Northern blots (Fig.  31, suggesting that the clones are full-length. The cDNA clones are highly homologous to each other, sharing over 92% identity between arg3 and argl or arg2 and over 99% identity between argl and arg2 in the overlapping regions ( Fig. 1 and data not shown). Although at first glance, it appeared that at least argl and arg2 could be derived from a single gene that was alternatively spliced in the 3'-untranslated region, there are several deletions and insertions of up to 13 nucleotides in the 3'-untranslated region in addition to the deletion of over 400 nucleotides in argl compared with arg2, suggesting that they are derived from different genes. All three cDNA clones have a canonical polyadenylation signal (AATAAA) 17 nucleotides upstream ofthe poly(A) site, supporting their derivation from different genes. This conclusion is also consistent with the fact that multiple arginase genes are present in Xenopus genome (Shi and Brown, 1993).
Each of the three clones has an open reading frame encoding a protein of 360 amino acids (Fig. 2). In-frame stop codons are present upstream of the initiation codon in both arg2 and arg3, suggesting that the open reading frames encode the full-length proteins in vivo. Like the cDNA sequences, the sequences of argl and arg2 are over 99% identical, whereas they share about 97% identity with arg3 (data not shown).
A number of arginase genes have been cloned from yeast to human. All the cloned vertebrate arginase genes encode livertype arginases. argl-3 are clearly not the liver-type arginases as they are more homologous to the mammalian liver arginases than to the Xenopus liver arginase (about 61% versus 57% identity, respectively) ( Fig. 2 and data not shown). As expected, the Xenopus liver arginase is slightly more homologous (-63-66% identity) than a r g l 3 (61%) to the mammalian proteins. The classification of thexenopus argL as a liver arginase is also supported by the fact that it is strongly expressed in liver of adult frogs, but very weakly in other tissues (Xu et al., 1993). All these vertebrate proteins share about 40% identity with the yeast arginase. Although the structure and functional domains are unknown for any arginase, several regions (underlined with hatched bars in Fig. 2) are conserved from yeast to human, suggesting their involvement in the catalytic activity and/or maintenance of the structure of the enzymes.
llssue-specific Expression ofArginase Genes-The identification of two types (liver versus nonhepatic) arginases prompted us to investigate their expression in detail. First we compared their expression in different tissues in tadpoles with or without 18-h T, treatment. As expected from our early study (Shi and Brown, 19931, argl-3 were expressed in the intestine of stage 52-54 tadpoles, and their expression was up-regulated by T, treatment ( Fig. 3A) (due to the high homology among argl-3, the cDNA probes used hybridized to all three mRNAs). In the tadpole liver, however, an extremely weak signal was observed in control tadpoles. T3 treatment slightly up-regulated the expression of these genes in the liver, but their expression was again much weaker than that in the intestine. In stark contrast, the Xenopus liver arginase gene (argL) was strongly expressed in the liver and was not affected by the T, treatment ( Fig. 3 A ) . No signal was detected in the intestine of either treated or control tadpoles at this stage.
At stage 60, the onset of metamorphic climax, a r g l 3 and argL again displayed very different tissue-specific expression patterns (Fig. 3B). a r g l 3 were strongly expressed in the tail, but weakly, especially arg3, in the intestine, hindlimb, or the trunk region. In contrast, argL was not detectable in the tail, but strongly expressed in the trunk region, which included the tadpole liver. These results indicate that argL and the nonhepatic arginase genes a r g I 3 have different tissue specificities in their expression and that the expression of argL, as expected, is strong in the liver.
Developmental Regulation of Xenopus Arginase Genes-The expression of arginase genes throughout embryonic and tadpole development was analyzed using total RNA from the entire embryos or tadpoles from stage 9 (at midblastula transition) to stage 62 (the climax of metamorphosis) (Fig. 4). argl and arg2 mRNAs were clearly detectable as early as stage 16/17 (neurula) and arg3 mRNA by stage 23/24 (early tailbud).  It should be pointed out that in this and many other Northern poles was found around stage 47/48. On the other hand, argL blots, argl and arg2 were not well resolved. It is, therefore, mRNA was present by stage loll1 at gastrulation and throughuncertain whether they have the same expression profiles. out tadpole development. Higher levels of argL expression were However, in the instances where they were resolved, the two found at stages 23/24 and 60-62. Similarly, Xu et al. (1993) mRNA species were clearly regulated identically (see Fig. 3 and reported the expression of argL mRNA in tadpoles at stage data not shown). Highest expression of argl-3 in whole tad-46/47 or later, although they did not detect any mRNA at stages 28-44. It is possible that this discrepancy was simply due to different exposure times or specific activities of the probes used as neither they nor we have determined the absolute expression levels for argL. In this regard, it is also worth pointing out that argl-3 could also be expressed as early as stage lO/ll, but the signal was too weak under our hybridization conditions. Thus both argL and a r g l 3 are expressed very early during embryogenesis, well before liver differentiation, suggesting a role other than that in the urea cycle.

Differential Regulation in Different Tissues during Metarnor-
phosis-The expression of arginase genes in the liver could suggest their participation in the urea cycles. However, both argL and nonhepatic a r g l 3 are also expressed in other tissues.
We were interested in the roles of arginases in tissue remodeling during metamorphosis. Therefore, we analyzed the arginase expression during metamorphosis in three very different tissues, the intestine, hindlimb, and tail (Fig. 5), which under-FIG. 5. Northern blots showing differential regulation of the liver and nonhepatic arginase genes in the intestine, tail, and hindlimb during metamorphosis. The most dramatic period of intestinal remodeling is around stages 60-62 and that of tail resorption is stage 62 and later (Nieuwkoop and Faber, 1956;Dodd and Dodd, 1976;McAvoy and Dixon, 1977). The tail is completely resorbed by stage 66, the end of metamorphosis. Hindlimb morphogenesis is completed around stages 56 and undergoes rapid growth in the remaining metamorphic period. Ten pg of total RNA was loaded per lane, except for stage 64 tail and stage 56 hindlimb, which contained only 5 pg. The same or duplicated blots were probed with an argl-3 probe, argL, or rpL8. The upper band detected with the rpL8 probe is likely to be a precursor of the mRNA, as it was not observed in oocytes where no transcription takes place (not shown). Slight variation does occur in the levels of rpL8 mRNA during development, especially for the precursor (Shi and Liang, 1994). goes a dramatic remodeling, de novo development, and total resorption, respectively (Dodd and Dodd, 1976;Smith-Gill and Carver, 1981;Yoshizato, 1989). The expression of argL and a r g l 3 was found to be regulated differently in all three tissues.
In the intestine, both argL and a r g l 3 mRNA levels were elevated during metamorphosis (stages 58-64) and decreased by the end of metamorphosis (stage 66). However, a noticeable increase in a r g l d expression and its subsequent down-regulation was found to be earlier, at stage 58 and stage 64, respectively, compared with those for argL (stage 60 and stage 66, respectively). In the tail, argl-3 were very highly expressed throughout metamorphosis, although slightly less mRNA was present at stages 58 and 64. In contrast, argL expression was low until stage 62, when it was up-regulated, coinciding with the tail resorption process (Nieuwkoop and Faber, 1956). Finally, in the hindlimb a r g l 3 were expressed at low levels until stage 66, the end of metamorphosis, when it was dramatically up-regulated. argL, on the other hand, maintained low levels of expression after stage 58 with very little expression a t stage 56. When the expression of the same gene was compared in different tissues, it was found that a r g l 3 had the highest levels of expression in tail at all stages and in hindlimb at stage 66. Coincident with these high levels of expression, the amount of arg3 mRNA relative to that for argl-2 was much higher in these tissues at these stages, suggesting the existence of a factor(s) in maintaining both the very high levels of a r g l 3 expression and the high ratio of arg3 to argl-2 mRNA levels. was probed with argl-3, argL, or rpL8. argl-2 were activated after 1-day treatment and reached a peak expression after 2-3 days in the intestine, and in the tail, a r g l 3 were all up-regulated by about 2-fold after 1-7 days. The activation of arg3 in the intestine is not apparent in this figure due to its low levels of expression. But it could also be up-regulated in the intestine at least after 2-3 days of treatment (not shown). In contrast, argL was up-regulated in both tissues only after 5-7 days. The signal for rpL8 varied slightly in some samples. It was likely due to unequal loading. However, it clearly does not affect the conclusions above.

Differential Regulation of argL and argl-3 by Thyroid
Hormone-Thyroid hormone controls amphibian metamorphosis, and the endogenous T, concentration is known to be high at the climax (stages 60-66) of metamorphosis (Leloup and Buscaglia, 1977). The high levels of argL and argl-3 mRNAs in the remodeling intestine and resorbing tail suggest that T, may influence arginase gene expression. Indeed, we have shown previously that argl-3 expression could be induced in the intestine by treating stage 52/54 tadpoles with exogenous T, (clone IU22, Shi and Brown, 1993). We have now extended this observation by showing that argl-3 genes could be similarly activated in the intestine of stage 50 (not shown) and stage 56 (Fig. 6) tadpoles by exogenous T, (5 nM) when the endogenous T, concentration was low (Leloup and Buscaglia, 1977). Maximum expression in the intestine was observed after 2-3 day T, treatment. In the tail exogenous T, could slightly up-regulate (2-3fold) the argl-3 mRNA levels after 1-7 day T, treatment ( Fig.  6 and data not shown).
In contrast to argl-3, the expression of the argL gene was not affected by T, in the intestine and tail a t stage 50 (not shown) or at stage 56 during the first 3 days of treatment (Fig.  6). Only after 5-7 days of T, treatment was there a fewfold up-regulation in argL mRNA level in both the intestine and tail. This slow response of argL gene to T, is in agreement with that reported by Xu et al. (19931, where it was shown that argL expression could be up-regulated in the middle portion of stage 51-53 tadpoles by T, with a latent period of 70 h. Under our treatment conditions, intestinal remodeling, including length reduction and epithelial folding, is known to occur after 2-3 days . Thus, the expression of both argL and argl-3 during T, treatment correlate with that during natural metamorphosis (Fig. 5).
The more dramatic activation and lower basal levels of argl-3 expression in the intestine enabled us to further investigate this regulation by T,. When stage 52-54 tadpoles were treated with T, for varying time periods, it was found that activation of the argl-3 gene was detectable as early as 4 h after T, addition, and the mRNA level continued to increase up to 48 h (Fig. 7). This fast kinetics of the activation suggest that T, may directly regulate argl-3 expression. To test this, tadpoles at stage 52-54 were treated with T, in the presence or absence of protein synthesis inhibitors, and argl-3 mRNA levels were determined in the intestine. Although argl-3 were again found to be activated after 12-h T, treatment in the absence of protein synthesis inhibitors, the results were less clear when these inhibitors were present due to their dramatic stabilizing effect on the mRNAs (Fig. 8). On a lighter exposure, however, it did appear that the mRNA levels were slightly higher in the intestine from tadpoles treated with both T, and the inhibitors compared with those from tadpoles treated with the inhibitors alone (Fig. 8). This resistance to protein synthesis inhibitors together with the fast kinetics argues that argl-3 genes are early T, response genes.

Multiple Arginase Genes With Different Tissue Specificities Exist in X.
Laevis-The three nonhepatic arginases reported here are highly homologous to each other, over 97% identity among them. In addition, they share about 60% identity with liver arginases, including the one from Xenopus. However, the nonhepatic Xenopus arginases are more similar to mammalian liver arginases than to the Xenopus liver arginase. Since there has been evidence that nonhepatic arginases exist in mammals (Jackson et al., 1986), these sequence similarities suggest an order of their emergence during evolution. A gene duplication before the separation of amphibians and mammals likely gave rise to the liver and nonhepatic arginases. arg3 and argl or arg2 were likely derived from a gene duplication in Xenopus and still more recently, as most of the Xenopus genome was duplicated to produce a pseudotetraploid organism, argl and arg2 evolved.
The liver arginase gene is, as expected, strongly expressed in tadpole liver but much more weakly in other tissues. Similar observations have been made in adult frogs (Xu et al., 1993). In contrast, the nonhepatic arginase (argl-3) genes are only weakly expressed in the tadpole liver. The highest levels of argl-3 mRNA are present in the tail and postmetamorphic hindlimb. Furthermore, compared with argl-2, arg3 expression is much lower except in the tail and postmetamorphic hindlimb. Although it is unknown whether such a regulation has any functional importance, as argl-3 are more than 97% identical, it is possible that the very high levels of arg3 in these tissues simply serves to provide sufficient amounts of arginases when needed (see below). It would be interesting, however, to determine the mechanisms of the differential regulation among these highly related genes as well as between liver and nonhepatic arginase genes.
Correlation of Arginase Gene Expression with Normal and T,-induced Metamorphosis-We were initially interested in the arginase genes because of their expression in the intestine a t the metamorphic climax and their up-regulation by T, (Clone IU22, Shi and Brown, 1993). The remodeling of the intestine begins around stage 58 (Marshall and Dixon, 1978;McAvoy and Dixon, 1977;Ishizuya-Oka and Shimozawa, 1987). The first and most dramatic event is the gradual reduction of the length by as much as 10-fold. At the cellular level, massive cell death occurs in the primary epithelium around stage 61. By the end of metamorphosis at stage 66, a complex frog intestine is formed, which consists of a multiply folded secondary epithelium with elaborative connective tissue and muscles, in sharp contrast to the tadpole intestine, which is basically a simple tubular structure of primary epithelium with little connective tissue or muscles.
We have shown that arg1-3 are activated around the onset of intestinal remodeling (stage 58) and relatively high levels are present during the period of epithelial replacement and connective tissue development (stages 60-62). Lower levels of expression persist in post-metamorphic intestine. In addition, the liver arginase gene is also up-regulated during this period, although its peak expression in the intestine is shifted toward a later period of intestinal development (stages 62-64). No strong correlation exists between the regulation of argl-3 expression and tail resorption or limb development (Nieuwkoop and Faber, 19561, even though very high levels of expression are present in the tail and much lower levels in the hindlimb except at the end of metamorphosis. On the other hand, argL is activated during tail resorption (stages 62-64) and hindlimb development (stages 58-66), although a t much lower levels in the limb.
Possible Roles of Arginases in Arginine Metabolism and Amino Acid Biosynthesis during Tissue Remodeling-Arginase is best known as the last enzyme of the urea cycle in the liver. It catalyzes the conversion of arginine to ornithine and urea (Jackson et al., 1986;Cohen, 1970). However, both the liver and nonhepatic arginase genes in X. laevis are expressed in many different tissues, suggesting potential functions outside the urea cycle. In fact, nonhepatic arginases have also been reported in mammals (Jackson et al., 1986), although none of these genes have been cloned. It has been proposed that the nonhepatic arginases are involved in the conversion of arginine to proline and glutamate as well as polyamine biosynthesis (Yip and Knox, 1972;Takiguchi et al., 1988;Tabor and Tabor, 1984). Although the hypothesis remains to be verified, the regulation of arginases genes during metamorphosis is, at least, consistent with it. This stems from the fact that proline and glutamate are abundant in collagens.
Collagens are major components of extracellular matrices such as the basement membrane and interstitial connective tissue (Weiss, 1984). Although type IV collagen is rich in basement membranes and type I in interstitial connective tissues, they share with other collagens some important sequence features. Collagens consist of three chains with (~1 )~ ~2 configuration for type I and IV collagens. These polypeptides are mostly Q helices. In the Q helical region, the protein consists of repeating triplets of Gly-X-Y. Proline is generally present at the X position and hydroxyproline at the Y position. Such a sequence organization results in unusually high proportions of glycine and proline in collagen molecules. In fact, an analysis of eight collagen chains, two for each of al(I), a2(1), al(IV), and a2(IV), showed that on the average, glycine represents 27.7% of the proteins and proline, 17.8% (data not shown). In addition, glutamate and glutamine are 4.4 and 3.6% of the total protein, respectively. Taking into account the high abundance of proline and glycine, the average proportion of the remaining amino acids is expected to be 2.27%. Thus, glutamate and glutamine are also over-represented in collagens. It is, therefore, conceivable that the demand for proline and glutamate, and the derivative of the latter, glutamine, may be high in collagen-rich tissues. As arginase can be involved in the biosynthesis of proline and glutamate from arginine, this could account for the high levels of argl-3 expression in the tail and postmetamorphic hindlimb. Similarly, in the metamorphosing intestine, the activation of argl-3 and argL could also be required for the biosynthesis of proline and glutamate since the connective tissue develops extensively and the new basement membrane is formed upon secondary epithelial cell differentiation. It is unclear why lower levels of argl-3 are present in the hindlimb before stage 66. It is possible that other arginases such as argL may be involved. Finally, the delayed expression of argL compared with argl-3 during intestinal remodeling and its activation during tail resorption suggest that argL could also be involved simply in arginine removal after cell degeneration as both tissues undergo extensive cell death. The cloning of both liver and nonhepatic arginases now makes it possible to study these functions directly.