Ablation of Arginylation in the Mouse N-End Rule Pathway: Loss of Fat, Higher Metabolic Rate, Damaged Spermatogenesis, and Neurological Perturbations

In the N-end rule pathway of protein degradation, the destabilizing activity of N-terminal Asp, Glu or (oxidized) Cys residues requires their conjugation to Arg, which is recognized directly by pathway's ubiquitin ligases. N-terminal arginylation is mediated by the Ate1 arginyltransferase, whose physiological substrates include the Rgs4, Rgs5 and Rgs16 regulators of G proteins. Here, we employed the Cre-lox technique to uncover new physiological functions of N-terminal arginylation in adult mice. We show that postnatal deletion of mouse Ate1 (its unconditional deletion is embryonic lethal) causes a rapid decrease of body weight and results in early death of ∼15% of Ate1-deficient mice. Despite being hyperphagic, the surviving Ate1-deficient mice contain little visceral fat. They also exhibit an increased metabolic rate, ectopic induction of the Ucp1 uncoupling protein in white fat, and are resistant to diet-induced obesity. In addition, Ate1-deficient mice have enlarged brains, an enhanced startle response, are strikingly hyperkinetic, and are prone to seizures and kyphosis. Ate1-deficient males are also infertile, owing to defects in Ate1−/− spermatocytes. The remarkably broad range of specific biological processes that are shown here to be perturbed by the loss of N-terminal arginylation will make possible the dissection of regulatory circuits that involve Ate1 and either its known substrates, such as Rgs4, Rgs5 and Rgs16, or those currently unknown.


Introduction
N-terminal arginylation of intracellular proteins by Arg-tRNAprotein transferase (R-transferase) is a part of the N-end rule pathway of protein degradation (Fig. 1A). In eukaryotes, this pathway is a part of the ubiquitin (Ub)-proteasome system. The Nend rule relates the in vivo half-life of a protein to the identity of its N-terminal residue (reviewed in [1,2,3,4]). Degradation signals (degrons) that can be targeted by the N-end rule pathway are of two distinct kinds: N-terminal degrons, called N-degrons, and internal (non-N-terminal) degrons [1,5]. The main determinant of an N-degron is a destabilizing N-terminal residue of a substrate protein (Fig. 1A). The other determinants of N-degron are a substrate's internal Lys residue (the site of formation of a poly-Ub chain) and a nearby unstructured region [6,7]. An N-degron is produced from a precursor, called a pre-N-degron, through a protease-mediated cleavage of a substrate that exposes a destabilizing N-terminal residue.
The N-end rule has a hierarchic structure (Fig. 1A). N-terminal Asn and Gln are tertiary destabilizing residues in that they function through their enzymatic deamidation, to yield the secondary destabilizing N-terminal residues Asp and Glu [8]. Destabilizing activity of N-terminal Asp and Glu requires their conjugation to Arg, one of the primary destabilizing residues, by the Ate1-encoded R-transferase [9,10,11,12]. In eukaryotes that produce nitric oxide (NO), R-transferase arginylates not only N-terminal Asp and Glu but also Cys, after its conversion to Cyssulfinate or Cys-sulfonate, in reactions that require NO and oxygen ( Fig. 1A) [11,13]. Alternative splicing of the mammalian Ate1 pre-mRNA produces isoforms of R-transferase, a metabolically unstable protein whose enzymatic activity and the in vivo half-life are down-regulated by heme [10,12,14]. E3 Ub ligases of the N-end rule pathway are called N-recognins. An N-recognin is an E3 that can recognize (target for polyubiquitylation) at least a subset of N-degrons (Fig. 1A) [1,4]. Some of substrate-binding sites of an N-recognin target N-degrons, while other sites of the same N-recognin are specific for structurally unrelated internal (non-Nterminal) degrons [15,16]. At least four N-recognins, Ubr1, Ubr2, Ubr4 and Ubr5, mediate the mammalian N-end rule pathway (Fig. 1A) [4,17].
The functions of the N-end rule pathway in eukaryotes include selective degradation of misfolded proteins; the sensing of heme, oxygen, nitric oxide (NO), and short peptides; the regulation of DNA repair and peptide import; the signaling by transmembrane receptors, through the NO/O 2 -controlled degradation of Gprotein regulators Rgs4, Rgs5 and Rgs16; the fidelity of chromosome segregation; regulation of apoptosis, meiosis, spermatogenesis, neurogenesis, and cardiovascular development; the functioning of specific organs, in particular the brain and the pancreas; and regulation of leaf senescence, seed germination, and other processes in plants ( [2,12,16,18,19,20], and refs. therein). A partial N-terminal arginylation of the apparently long-lived mammalian a-actin [21] suggests that arginylation of some proteins may not alter their in vivo half-lives.
To bypass the embryonic lethality of nonconditional Ate1 2/2 mice, we employed the Cre-lox technique [24]. As shown below, a systemic postnatal deletion of the sole active Ate1 flox allele in juvenile Ate1 flox/2 mice causes a rapid decrease of body weight and results in early death of ,15% of Ate1-deficient mice, with surviving mice attaining only ,70% of normal weight. This failure to thrive occurs despite higher than normal food intake by Ate1deficient mice. These mice contain little or no visceral fat, exhibit an increased metabolic rate, a decreased fasting blood glucose level, and an increased intestinal import and retention of amino acids and/or peptides. Ate1-deficient mice are also resistant to diet-induced obesity and exhibit ectopic induction of the Ucp1 uncoupling protein in white adipose tissue (WAT). In addition, Ate1-deficient mice have enlarged brains, an enhanced startle response, and are strikingly hyperkinetic. They often suffer from kyphosis, i.e., an excessive curvature of the upper back, and from frequent seizures as well. Ate1-deficient males are also infertile, owing to defects in meiotic Ate1 2/2 spermatocytes. The remarkably broad range of specific biological processes that are shown here to be perturbed by the loss of N-terminal arginylation will facilitate the dissection of regulatory circuits that involve Ate1 and either its known substrates, such as Rgs4, Rgs5 and Rgs16 [11,13], or those currently unknown.

Results
Ate1 flox/2 Mouse Strains and Production of Ate1 2/2 Mice Standard methods were employed to produce, initially, ATE flox/+ mouse strains in which a specific segment of Ate1 was ''floxed'', i.e., flanked by 34-bp loxP repeats ( Fig. 2C-E). The targeting vector contained ,14 kb of Ate1, including the exon 1A-exon 4 segment that encodes an essential part of R-transferase [9] (Fig. 2A). Our previous work has shown that the Ate1 promoter (P Ate1 ) is bidirectional, expressing both Ate1 and an oppositely oriented gene termed Dfa (divergent from Ate1), which overlaps with exon 1A of Ate1 ( Fig. 1B) ( [14]; C.S.B. and A.V., unpublished data). To minimize the possibility of perturbing the expression of Dfa, the ''floxed'' region of Ate1 encompassed exons 2-4, away from exon 1A (Fig. 1B). Our aim was to produce ATE flox/2 mouse strains that were ''poised'' to lose their remaining active ATE flox allele through the expression of Cre recombinase. To do so, heterozygous matings were carried out among the above ATE flox/+ mice, the previously constructed ATE +/2 mice [10], and a mouse strain that contained the CaggCreER gene, expressed from the ubiquitously active chimeric Cagg promoter [25]. CaggCreER encoded CreER, a fusion between Cre and a derivative of the mouse estrogen receptor ligand binding domain. CreER was functionally inactive (sequestered in the cytosol) but could be activated by intraperitoneal (IP) injections of tamoxifen (TM) [25]. Depending on configurations of their Ate1 alleles, the resulting mice, poised for the loss of Ate1, were termed Ate1 flox/2 ;CaggCreER or Ate1 flox/flox ;CaggCreER.
Using standard methods, we could demonstrate the presence of Ate1 flox/2 ;CaggCreER mice, at expected (Mendelian) frequencies, in the progeny of above matings. These mice expressed TM-inducible CreER recombinase and contained a single copy of Ate1 flox , the active Ate1 allele (Figs. 2 and 3A). The functional intactness of Ate1 flox was inferred from the fact that Ate1 flox/2 ;CaggCreER mice survived embryogenesis (in contrast to Ate1 2/2 mice [10]) and were phenotypically similar (in the absence of TM treatment) to Ate1 +/2 and Ate1 +/+ mice. To induce the Ate1 flox RAte1 2 conversion, ,1 month old Ate1 flox/2 ;CaggCreER mice and their Ate1 flox/+ ;CaggCreER (as well as Ate1 +/2 ;CaggCreER) littermates, used as controls, were treated with TM (see Materials and Methods for details, including the ages of TM-treated mice). Southern hybridization and PCRbased analyses of DNA from tissues of the resulting mice (sampled ,1 month after TM treatment) confirmed the TM-induced, Cremediated excision of the Ate1 flox allele in Ate1 flox/2 ;CaggCreER mice. The frequency of Ate1 flox RAte1 2 conversion was nearly 100% in the brain and kidney of these mice, but significantly lower in several other tissues (Figs. 2G, H and 3B).
We also used an affinity-purified antibody to mouse Ate1 [11] to carry out immunoblotting (IB) with extracts from brain, heart, kidney, liver, muscle, brown adipose tissue (BAT) and white adipose tissue (WAT) that were harvested up to 8 months after TM treatment of Ate1 flox/2 ;CaggCreER mice, versus identically TMtreated control littermates. No Ate1 could be detected by IB in several tissues of TM-treated Ate1 flox/2 ;CaggCreER mice, in contrast to readily detectable Ate1 in TM-treated control mice (Fig. 1C). The only significant exception was liver (Fig. 1C, lanes 7, 8; cf. lanes 5, 6 or lanes 11-14; see also below). One effect of Ate1 depletion in the mouse brain was a striking increase of Rgs4, a physiological Ate1 substrate (see Introduction) that down-regulates specific G proteins by acting as a GTPase-activating protein (GAP) (Fig. 1D). Whereas no Rgs4 could be detected in the Ate1containing brain (owing to degradation of Rgs4 by the N-end rule pathway [11]), an intense band of Rgs4 was present in the Ate1deficient brain, illustrating high penetrance of Ate1 deletion in the brain (Fig. 1D).
We also performed in vitro arginylation assays with extracts from several tissues of Ate1 flox/2 ;CaggCreER mice 21 days after TM treatment, versus extracts from identically treated Ate1 +/+ or ATE1 +/2 mice. The TM-induced decrease of arginylation activity in specific organs of Ate1 2/2 ;CaggCreER mice ranged from ,90% in the brain and kidney to ,60% in the liver (Fig. 3C-E). Although heterozygous Ate1 +/2 mice were phenotypically similar to their wild-type (Ate1 +/+ ) counterparts, we found that Ate1 +/2 mice grew slightly but consistently slower than Ate1 +/+ mice, and reached a lower average weight (Fig. 4B). In agreement with this mild but detectable haploinsufficiency of Ate1, the arginylation activity in extracts from, e.g., brains or hearts of Ate1 +/2 mice was significantly below its wild-type (Ate1 +/+ ) levels ( Fig. 3C), implying the absence of a compensatory (e.g., autoregulated) increase of Ate1 expression upon a decrease of Ate1 gene dosage.
In the entire cohort of TM-treated post-natal Ate1 flox/2 ;CaggCreER mice, 15% of them (18 of 119 mice) died over 42 days after TM treatment. Crucially, none of identically TM-treated control mice (Ate1 +/+ ;CaggCreER, Ate1 +/2 ;CaggCreER, or Ate1 flox/+ ;CaggCreER) died in the same time interval. The frequency of Ate1 2/2 ;CaggCreER mice  [10], in which the Ate1 exons 1b through 3 were replaced by a cassette encoding a promoterlacking, NLS-containing LacZ (NLS-bgal) (it was expressed from the endogenous P Ate1 promoter) and the Neo selection marker expressed from the phosphoglycerate kinase P PGK promoter (green rectangles). (B) A diagram of the 59 end of wild-type (wt) mouse Ate1, indicating approximate locations of exons 1a through 5. (C) The ,22.5 kb targeting construct containing a ,6 kb long-arm region of Ate1 homology (shown as a shaded rectangle on the left); a single loxP site (red triangle) upstream of Ate1 exon 2, a ''floxed''-hygromycin-resistance (hph) cassette, expressed from the P PGK promoter (blue arrow between two red triangles) downstream of Ate1 exon 4; a ,2 kb short-arm region of homology (an inclined shaded rectangle), and the HSV thymidine kinase (tk) negative-selection cassette expressed from the P HSV promoter (yellow arrow).  [10] unconditionally null Ate1 2 allele (denoted as ''null'') yields the 5.8 kb fragment. Both the wild-type Ate1 allele and the flox-on (Ate1 flox ) allele yield the 9.7 kB fragment, whereas the null flox-off (Ate1 2 ) allele yields the characteristic 3.8 kb fragment. The use of DNA probe D and EcoRI-digested DNA from specific tissues of tamoxifen (TM)-treated Ate1 flox/2 ;CaggCreER mice allowed approximate estimates of the levels of Cre-mediated recombination that produced the flox-off (Ate1 2 ) allele. For example, whereas no flox-on (Ate1 flox ) allele could be detected in the kidney and brain of Ate1 flox/2 ;CaggCreER mice after TM treatment (lanes 5, 6), approximately equal amounts of flox-on (Ate1 flox ) and flox-off (Ate1 2 ) alleles were present in the heart of TM-treated Ate1 flox/2 ; CaggCreER mice. Lanes 1-3, 1,000, 250, and 25 ng of EcoRI-digested wt mouse genomic DNA (from a tail biopsy), respectively. Lane 4, EcoRI-digested genomic DNA from the tail of a previously constructed [10] Ate1 +/2 mouse. Lanes 5-7, EcoRI-digested genomic DNA from the indicated tissues of TM-treated Ate1 flox/2 ;CaggCreER mice. Lane 8, same as lane 7, but from a TM-treated Ate1 flox/2 mouse (lacking the CaggCreER transgene). doi:10.1371/journal.pone.0007757.g002 succumbing upon the acquisition of Ate1 2/2 genotype was agedependent. Specifically, 46% of Ate1 2/2 ;CaggCreER mice younger than 30 days at the beginning of TM treatment died within 42 days after TM treatment. In contrast, only 12% of Ate1 2/2 ;CaggCreER mice died if they were older than 30 days (by up to 56 days) at the beginning of TM treatment. Those among Ate1 2/2 ;CaggCreER mice that survived for at least 42 days after TM treatment eventually resumed growth, but the rate of growth and their maximum weight were significantly below those parameters for identically TM-treated control mice ( Fig. 4A-C).
Among surviving Ate1 2/2 ;CaggCreER mice, 10% (8 of 80) developed patches of red hair among their normally black hair, in contrast to identically TM-treated Ate1-containing mice (data not shown), suggesting a misregulation of melanocytes in Ate1deficient mice. The liver, spleen, intrascapular brown adipose tissue (BAT), pancreas, and testis of Ate1 2/2 ;CaggCreER mice appeared normal and were of appropriate sizes (if the smaller size of these mice (Fig. 4C) was taken into account), whereas the brains, hearts and kidneys of these Ate1-deficient mice were disproportionately large, in comparison to those of Ate1-containing siblings (Fig. 5D). Intact brains of Ate1-deficient mice appeared swollen, in comparison to brains harvested, in parallel, from identically treated Ate1-containing siblings (Fig. 6A). In addition, Ate1deficient males were infertile, in agreement with defects in their testes ( Fig. 6F-I). Yet another abnormality of Ate1-deficient mice  [10] unconditionally null Ate1 2 allele, using primers CB156 and CB157 (Table 4). Lower panel: the 470 bp DNA fragment characteristic of the Cre-produced flox-off (Ate1 2 ) allele, with primers CB110 and CB157 (Table 4); and the 324 bp DNA fragment (control), amplified from the IL-2 gene using primers IMR42 and IMR43, in the same PCR reaction. (B) The Cre-mediated Ate1 flox RAte1 2 conversion, detected by PCR (as described in panel A) in genomic DNA isolated from the indicated tissues immediately after the fourth (daily) IP injection of tamoxifen in a 24-day old Ate1 flox/2 ;CaggCreER mouse. (C) Relative in vitro arginylation activity (cpm/reaction) in extracts of the indicated tissues from a wild type mouse (Ate1 +/+ ) (black bar), a heterozygous mouse (Ate1 +/2 ) (blue bar), and an Ate1 2/2 mouse (the latter mouse was initially Ate1 flox/2 ;CaggCreER) (red bar) from the same litter 76 days after TM treatment. A white bar on the right indicates the relative arginylation activity obtained with purified recombinant mouse Ate1 (denoted as ''rAte1'') that had been expressed in S. cerevisiae. Shown here are ''cpm/reaction'' after subtracting ''cpm/reaction'' in the null-control (''buffer alone'') sample. The control incorporation was approximately equal to that observed in extracts from spleen and thymus. In other words, the assay configured as described in this panel and in Materials and Methods was not sensitive enough to robustly detect the arginylation activity in extracts from spleen and thymus. (D) Relative in vitro arginylation activity (cpm/reaction) in the whole brain, cerebellum, and hippocampus harvested from wild type mice (Ate1 +/+ ; n = 3), heterozygous mice (Ate1 +/2 ; n = 3), and Ate1 2/2 mice (specifically, Ate1 flox/2 ;CaggCreER mice; n = 3) mice 40 days after TM treatment. Standard deviations are indicated. (E) Relative in vitro arginylation activity (cpm/reaction) in testis extracts from Ate1 +/+ mice (n = 3) and Ate1 2/2 mice (specifically, Ate1 flox/2 ;CaggCreER mice; n = 3) ,130 days after TM treatment. Standard deviations are indicated. doi:10.1371/journal.pone.0007757.g003 was their strikingly lower content of the peritoneal white adipose tissue (WAT), on average only 16% of WAT in Ate1-containing mice (Figs. 5D and 7A-C). These phenotypes are discussed below.

Spermatogenesis Defects and Infertility of Ate1-Deficient Male Mice
The marking of Ate1 2 allele with NLS-b-galactosidase (bgal) expressed from the P Ate1 promoter revealed high levels of Ate1 expression in the neural tube and other specific, often sharply delineated, regions of Ate1 +/2 embryos [10]. An earlier study detected high levels of Ate1 expression in spermatogonia (stem cells, located at the periphery of testis' seminiferous tubules), and possibly also in early meiotic spermatocytes of adult mice [26]. Male Ate1 2/2 ;CaggCreER mice that were produced by TM treatment (Fig. 4C) were found to be infertile in matings with Ate1-containing females, in contrast to identically TM-treated Ate1-containing males (data not shown). XGal staining of testis sections of NLS-bgal-marked Ate1 +/2 mice in the present work ( Fig. 6J) confirmed and extended the earlier evidence [26] for the pattern of Ate1 expression in testis. Whereas the lumens of seminiferous tubules in Ate1-containing testis were filled with inward-pointing sperm tails, the lumens of tubules in Ate1deficient testis contained few sperm cells, in a disorganized arrangement ( Fig. 6F-I), in agreement with the observed infertility of Ate1-deficient males.
To address the timing of requirement for Ate1 during spermatogenesis, we mated wild-type females with Ate1 flox/2 males that contained (instead of the CaggCreER gene) the PrpCreER gene (line 28.8) [27] or the PrmCre gene [28]. PrpCreER expresses TMinducible CreER from the Prp promoter, whose activity in testis is confined to spermatogonia and meiotic spermatocytes [27]. In contrast, PrmCre expresses the (unconditionally active) Cre recombinase from the protamine promoter, which is active at later stages of spermatogenesis, in (haploid) round and elongating and Ate1-deficient (n = 2; red curve) mice from the same litter as a function of time after tamoxifen (TM) treatment. Weights were measured at weekly intervals. Vertical bars indicate the ranges of measured weights. (B) Averaged growth curves for the indicated numbers of mice after TM treatment, plotted as a percentage of their weight immediately before TM treatment. Red, black and blue curves: Ate1 2/2 (n = 87), Ate1 +/+ (n = 55), and Ate1 +/2 (n = 66) mice. Red arrow indicates the time (,21 days) after TM treatment by which ,15% of Ate1-deficient mice have died while the rest of them began to gain weight. Note a slightly but clearly decreased weight of heterozygous (Ate1 +/2 ) mice (blue curve), in comparison to Ate1 +/+ mice (black curve) ,1 year after TM treatment. Error bars indicate standard deviations (SD). (C) Typical appearance of Ate1 2/2 versus wt mice (a smaller, leaner Ate1 2/2 mouse) ,1 year after TM-mediated ablation of Ate1. (D) Mean body lengths (6 SD) (from tip-of-nose to base-of-tail) between pairs of Ate1 2/2 (red bar) and Ate1 +/+ (black bar) mice. This comparison was derived from the data in Fig. 5C. Statistical analysis was performed using an unpaired t-test (p,0.08). doi:10.1371/journal.pone.0007757.g004 spermatids [28]. Three breeding pairs for each of two kinds of Ate1 flox/2 males (PrpCreER-based and PrmCre-based) and wild-type females were set up. 33% fewer litters and 50% fewer pups were produced with Ate1 flox/2 ;PrpCreER males, in comparison to Ate1 flox/2 ;PrmCre males (Table 1). (This substantial difference is expected to be even larger in a setting where an expressed Cre does not require a second, TMmediated step for activation, as is the case with TM-independent Cre expressed from the Prm promoter, but not with TM-inducible CreER, expressed from the Prp promoter.) Nearly equal numbers of the Ate1 flox (active) and Ate1 2 (inactive) alleles were present in the heterozygous progeny of matings that involved Ate1 flox/2 ;PrmCre males (13 versus 14 pups containing Ate1 flox versus Ate1 2 alleles, respectively). In contrast and most revealingly, only one Ate1 2 (inactive) allele but 12 Ate1 flox (active) alleles were present in the progeny of matings that involved Ate1 flox/2 ;PrpCreER males (Table 1). These findings suggest that the PrmCre-mediated inactivation of the (D) Comparison of tissue weights (as a percentage of total body weight (TBW)). Numbers in parentheses indicate the numbers of mice sampled and averaged for each tissue (Ate1-containing and Ate1-deficient). Brain (n = 43), liver (n = 28), heart (n = 17), kidney (n = 17), spleen (n = 16), white adipose tissue (WAT; n = 10), brown adipose tissue (BAT; n = 10), pancreas (n = 6), and testis (n = 8) from Ate1-containing (black bars) and Ate1-deficient mice (red bars). * = p,8610 215 ; ** = p,5610 25 ; and *** = p,0.003. Statistical analysis was performed using an unpaired t-test. Standard deviations are indicated. doi:10.1371/journal.pone.0007757.g005 comparison of representative brains harvested from an Ate1 +/+ and an Ate1 2/2 mouse, respectively, 134 days after tamoxifen (TM) treatment. Lower panel: brain weights expressed as percentages of total body weights in Ate1 +/+ (n = 41) and Ate1 2/2 (n = 40) mice. Horizontal bars and numbers indicate mean values. (B) Wet (0.4053 g versus 0.4608 g) and dry (0.1022 g versus 0.1119 g) weight components of the total mean brain weights (6SD) in Ate1 +/+ and Ate1 2/2 mice. (C) Total distance traveled (in meters), over 15 min, in an open field test among mice of different genotypes belonging to the same litter, 44 days after TM-treatment. Bar 1, Ate1 flox/+ ;CaggCreER mouse. Bar 2, Ate1 +/+ ;CaggCreER mouse. Bar 3, Ate1 +/+ mouse. Bar 4, Ate1 flox/2 ;CaggCreER mouse that was converted to Ate1 2/2 by TM treatment. Blue and red bars denote Ate1-containing and Ate1-deficient mice, respectively. (D) Same as in C but maximum lengths of single movements (in centimeters). (E) Same as in C but mean velocities (in cm/second) over 15 min. (F) Paraffin sections (4 mm) of testis showing cross-sections of seminiferous tubules in Ate1 +/+ testis stained with hematoxylin and eosin (1506 magnification). (G) Same as in F but Ate1 2/2 testis. Note that sperm tails in the lumens of Ate1 2/2 tubules are sparse in comparison to those in Ate1 +/+ testis. (H) Same as in F but at 6006 magnification. (I) Same as in G but at 6006 magnification. (J) XGal staining for bgal activity in a 10-mm section of Ate1 +/2 testis in which one copy of Ate1 was replaced by an ORF encoding NLS-b-galactosidase and expressed from the P Ate1 promoter (1006 magnification). (K) Immunoblotting analysis, using antibody to poly (ADP-ribose) polymerase (PARP), of testis extracts from an Ate1-containing (Ate1 flox/2 (+/2)) and an Ate1-deficient (Ate1 flox/2 ;CaggCreER (2/2)) mouse 16 days after TM treatment. Note the loss of the full-length length 116 kDa PARP and the presence of the 85 kDa PARP fragment (lane2). An asterisk denotes a protein crossreacting with anti-PARP antibody. doi:10.1371/journal.pone.0007757.g006 Ate1 flox allele, which occurs at a post-meiotic stage of spermatogenesis [28], takes place at a time when Ate1 is no longer essential for production of viable sperm cells, thus accounting for high frequency of the Ate1 2 allele in the progeny of matings that involve Ate1 flox/2 ;PrmCre males. In contrast, the PrpCreER-mediated inactivation of the Ate1 flox allele, which takes place in meiotic spermatocytes [27], clearly discriminated against the transmission of the Ate1 2 allele, in comparison to the Ate1 flox (active) allele, most likely because spermatocytes that became Ate1-deficient before they became haploid were sufficiently perturbed by the absence of arginylation to either Previous work demonstrated a defective assembly of synaptonemal complexes and massive apoptosis of spermatocytes in Ubr2 2/2 mice [26]. The Ate1 R-transferase acts upstream of Ubr2 and other Ub ligases of the N-end rule pathway (Fig. 1A). Given a role of Ate1 in spermatogenesis demonstrated in the present study, it is possible that the currently unknown N-end rule substrate(s) whose degradation is in down-regulated in Ubr2 2/2 spermatocytes is an Ate1 substrate. To assess the extent of apoptosis in Ate1deficient spermatocytes, we employed immunoblotting with antibody to poly(ADP-ribose)-polymerase (PARP), which is cleaved by caspases late in apoptosis. Anti-PARP antibody detected the (expected) 116 kDa full-length PARP in extracts from Ate1-containing mouse testis, but no 85-kDa PARP fragment, a marker of apoptosis (Fig. 6K, lane 1) [29]. In contrast, Ate1-deficient testis contained the 85-kDa fragment of PARP but virtually no full-length PARP, indicating extensive apoptosis in the absence of Ate1 (Fig. 6K, lane 2; cf. lane 1), in agreement with cytological and Ate1 2/2 male-infertility data ( Fig. 6F-I). The 85-kDa PARP fragment is expected to bear N-terminal Gly [29], which is not a substrate of the Ate1 R-transferase (Fig. 1A). Thus the absence of the 85-kDa PARP fragment in Ate1-containing testis (Fig. 6K, lane 1) signifies the lack of production of this fragment by caspases, rather than its degradation by the arginylation branch of the N-end rule pathway. Proteins that require N-terminal arginylation for their degradation and that are likely to be relevant to meiotic functions of Ate1 include Rec8 [30,31], a subunit of meiotic cohesin whose cleavage by separase is expected to produce an Ate1 substrate, similarly to the cleavage of Scc1/Rad21, the somatic counterpart of Rec8 (see Introduction).

Hyperkinesia, Seizures, and Enlarged Brains of Ate1-Deficient Mice
Most of Ate1 2/2 ;CaggCreER mice (96 of 180) were strikingly hyperactive (hyperkinetic) (Figs. 6C-E and 8A). Intact brains harvested from Ate1-deficient mice appeared swollen, in comparison to brains harvested, in parallel, from Ate1-containing siblings (Fig. 6A). While the average brain weight, as a percentage of total body weight (TBW), of Ate1-containing mice was 1.96%, that of Ate1 2/2 ;CaggCreER mice was 3.09% (Fig. 6A). In addition, there was a larger scatter of relative brain weights for Ate1-deficient mice, in comparison to identically TM-treated Ate1-containing controls. In particular, the brains of some Ate1-deficient mice reached 5% of TBW (Fig. 6A). Histological patterns of NLS-bgal [10] expressed from the P Ate1 promoter in the brains of Ate1 +/2 mice (data not shown) were in agreement with in situ hybridization data in the Allen Brain Atlas (http://www.brain-map.org/), in that Ate1 was expressed at varying but significant levels throughout the mouse brain, particularly in the hippocampus, dorsal thalamus, and cerebellum. No Ate1 protein could be detected in brain extracts of Ate1 2/2 ;CaggCreER mice, in contrast to extracts from wild-type or Ate1 +/2 brains (Fig. 1C, D). The virtually null Ate1 state of the brain in Ate1 2/2 ;CaggCreER mice was also indicated by a strong accumulation of Rgs4, a physiological substrate of Ate1 (see Introduction) (Fig. 1D).
We carried out cell proliferation assays with Ate1 2/2 ;CaggCreER mice (and controls), using 5-ethynyl-29-deoxyuridine (EdU). In examinations of EdU-labeled brain sections, we paid particular attention to regions such as the hippocampus and the periventricular zone of the lateral ventricles, where neurogenesis is known to occur. However, no differences in EdU incorporation between Ate1-deficient and Ate1-containing brains were observed (data not shown), consistent with a brain edema (fluid accumulation) being a significant cause of brain enlargement in Ate1-deficient mice. We also determined the water content of freshly isolated brains, by subtracting their ''dry weights'' (after freeze-drying) from their total weights. The average water content and dry weight of control (Ate1-containing) brains was 79.9% and 20.1%, respectively, versus 80.5% and 19.5%, respectively (p,0.03), for Ate1-deficient brains (Fig. 6B). Thus cerebral edema at least contributes to the observed differences in brain weight between Ate1-deficient and Ate1-containing mice. It remains to be determined whether an edema (owing, e.g., to an osmotic imbalance or inflammation) suffices to account for consistently observed Ate1-dependent differences in brain weights (Fig. 6A, B).
There was also a 10-fold higher propensity for seizures among Ate1-deficient mice. For example, during routine cage changes and handling of mice, ,3.1% of Ate1-deficient mice (38 of 1,232) versus ,0.3% of identically TM-treated Ate1-containing mice had tonic-clonic seizures. The skulls of Ate1-deficient mice appeared to be thinner, ''softer'' than the sculls of Ate1-containing mice. Although MRI analyses did not reveal statistically significant abnormalities in the shape or size of skulls in Ate1-deficient mice (Fig. 8B, C), the MRI data did not preclude the possibility that bone structure may be perturbed in the absence of Ate1. These issues remain to be addressed.
The neurological/behavioral abnormalities of Ate1-deficient mice included an enhanced startle response, a marker for increased anxiety in rodents. Specifically, the latency between stimulus and response (T max ) for Ate1-deficient mice was between 54% and 76% of the average latency for Ate1-containing controls, i.e., Ate1-deficient mice reacted significantly faster (Fig. 8D), thus exhibiting an enhanced startle response. The open field test is used to assess locomotor, exploratory and anxiety-like behavior in rodents. This test revealed a remarkably hyperkinetic behavior of Ate1-deficient mice (Figs. 6C-E and 8A), consistent with their enhanced startled response (Fig. 8D). The initial test involved a 15min comparison of movements of Ate1-deficient mice versus Ate1containing siblings of the same litter. An Ate1-deficient mouse traveled, during the test, a 3-fold greater distance than their (identically TM-treated) Ate1-containing counterpart (175.71 m versus 55.63 m, respectively) (Fig. 6C). The mean velocity of an Ate1 2/2 ;CaggCreER mouse was 19.5 cm/sec, in comparison to 7.0 cm/sec for a wild-type (Ate1 +/+ ) mouse, 6.2 cm/sec for an Ate1 +/+ ;CaggCreER mouse, and 5.5 cm/sec for an Ate1 flox/+ ;CaggC-reER mouse (Fig. 6E). To assess generality of this striking phenotype, we repeated the open field test with three Ate1-deficient mice at 10, 26, 38, and 82 days after TM treatment, in parallel with TM-treated Ate1-containing (control) mice. At 10 days after TM treatment, i.e., soon after the acquisition of the Ate1 2/2 genotype, the differences between distances travelled by Ate1-deficient versus Ate1-containing mice were small (Fig. 8A). However, by 26 days after TM treatment, there was a statistically significant difference between Ate1-deficient and Ate1-containing mice in regard to their locomotor activity (Fig. 8A). By 82 days after TM treatment, the locomotor activity of Ate1-deficient mice, in conjunction with their elevated overall anxiety, increased so much that the device in which the open field tests were performed became nearly impractical, as Ate1-deficient mice (in contrast to Ate-containing ones) kept jumping out of the testing box. Comparison of the response latency (T max ; recorded in msec) between Ate1-containing (n = 3; black bars) and Ate1-deficient mice (n = 3; red bars) to a 40-msec pulse of 120 dB (p120; p,0.3), a 40-msec pulse of 120 dB preceded by a pre-pulse of 5 dB (pp5; p,0.09), or a 40-msec pulse of 120 dB preceded by a pre-pulse of 15 dB (pp15; p,0.01). Statistical analysis was performed using an unpaired t-test. doi:10.1371/journal.pone.0007757.g008

Depletion of White Adipose Tissue in Ate1-Deficient Mice, and Their Resistance to Diet-Induced Obesity
To address the cause of a strikingly lower content of the peritoneal white adipose tissue (WAT) in Ate1-deficient mice, on average only 16% of WAT in Ate1-containing mice (Figs. 5D and 7A-C), we examined sections of intraperitoneal WAT. The average diameter of WAT adipocytes from Ate1-deficient mice was ,30% of the average diameter of such cells in identically TM-treated Ate1containing mice (25.567.4 mm versus 76.2616.2 mm, respectively) (Fig. 7D, E). Thus, at least the bulk of WAT decrease in Ate1deficient mice resulted from a decreased lipid content of individual adipocytes, rather from an extensive loss of adipocytes. Similar results were obtained with intrascapular brown adipose tissue (BAT) (Fig. 9A, B). The leanness of Ate1-deficient mice was particularly striking in view of their hyperphagy (see below).
We also asked whether the consistent difference in weight between Ate1-deficient and Ate1-containing mice on a standard ad libitum diet (Fig. 4A-C) could be reduced by an energy-rich, high-fat diet (HFD). At the end of the resulting 10-week test, the average weight of HFD-treated Ate1-containing mice was 152% of their starting weight (40.0 g versus 26.3 g). In contrast, the average weight of identically HFD-treated Ate1-deficient mice was only 122% (24.0 g versus 19.7 g) of their starting weight (Fig. 7F), indicating their relative resistance to diet-induced obesity. Yet Average core body temperatures of Ate1-containing (n = 8; black circles) versus Ate1-deficient (n = 11; red circles) mice during the first 3 weeks after TM treatment, in comparison to average core body temperatures of Ate1-containing (n = 54; black diamonds) versus Ate1-deficient (n = 36; red diamonds) mice beyond the first 3 weeks after TM treatment. (F) Core body temperature of individual Ate1-containing (black curves) and Ate1deficient (red curves) mice, recorded at 30-min intervals after placing mice in a room at 4uC. Mice were removed from the cold room after 6 hr or when their core body temperature fell below 28uC. doi:10.1371/journal.pone.0007757.g009 another phenotype of Ate1-deficient mice, observed during their initial loss of weight after TM treatment (Fig. 4A, B), was their lower core body temperature, on average 35.1uC, in comparison to identically TM-treated Ate1-containing control mice, whose average core body temperature was 36.0uC during the same time, in the absence of weight loss (Fig. 9E). After the early deaths of ,15% of Ate1 2/2 ;CaggCreER mice (Fig. 4A, B), the average temperature of surviving mice (36.6uC) was not significantly different from that of Ate1-containing control mice (36.7uC) (Fig. 9E). As one would expect from their depletion of WAT ( Fig. 7A-C), Ate1-deficient mice were strongly hypersensitive to cold (Fig. 9F).
Ucp1 is a proton carrier in the mitochondrial inner membrane that mediates a partial uncoupling of oxidative phosphorylation from ATP synthesis, an alteration that can increase heat production and thereby regulate body temperature and energy homeostasis. Although Ucp1 is normally expressed in BAT but not in WAT, several mouse mutants other than Ate1 2/2 that are resistant to diet-induced obesity have been shown to ectopically express Ucp1 in WAT [32,33]. Using RT-PCR and immunoblotting with anti-Ucp1 antibody, we found that the levels of Ucp1 mRNA and Ucp1 protein in BAT did not change significantly between Ate1-deficient and Ate1-containing mice (Fig. 7G-I).
Remarkably, however, the levels of both Ucp1 mRNA and Ucp1 were strongly increased in WAT of Ate1-deficient mice (Fig. 7G-I). A Ucp1-Ate1 connection revealed by these findings adds a new dimension to the understanding of Ucp1 regulation ( [32] and refs. therein), and may also provide an experimental route to identifying a relevant circuit that involves Ate1.

Increased Metabolic Rate in Ate1-Deficient Mice
During the week prior to TM treatment, Ate1 flox/2 ;CaggCreER and control (Ate1 flox/+ ;CaggCreER) mice (at that point, both strains contained Ate1) consumed 0.63 and 0.62 kcal of standard chow per gram of body weight per day, respectively (Fig. 10C). Within a week after TM treatment the now Ate1-deficient Ate1 2/2 ;CaggC-reER mice increased their food consumption on average to 125% of identically TM-treated Ate1-containing mice (Fig. 10C). This pattern of significant hyperphagia of Ate1-deficient mice continued for the duration of this study, i.e., up to ,8 months, with regular measurements for 6 weeks following TM treatment and intermittent comparisons afterwards (Fig. 10C). Thus, despite their initial decline of weight shortly after TM treatment and the early death of ,15% of Ate1-deficient mice, and despite their subsequent failure to gain, on average, more than ,63% and ,69% of the weights of Ate1 +/+ and Ate1 +/2 mice, respectively, the Ate1-deficient mice consumed significantly more food than their Ate1-containing counterparts (Fig. 10C). To address their patterns of glucose utilization, we fasted these mice for 16 hr and measured blood glucose before after administering a 50-mg (0.2 ml) bolus of glucose by gavage. The kinetics of rise and fall of blood glucose levels under these conditions was similar for Ate1-deficient and Ate1-containing mice (Fig. 10A). Ate1-deficient mice had lower fasting glucose levels than Ate1-containing mice (88.6 mg/dl versus 125.3 mg/dl, respectively; p,0.04), and also lower glucose levels 6 hr after administration of glucose (80.9 mg/dl versus 109.7 mg/dl, respectively; p,0.04), consistent with the (expected) higher energy expenditure of Ate1-deficient mice, and suggesting normal sensitivity of these mice to insulin (Fig. 10B). There were no other significant differences in blood composition (as well as urine composition) between Ate1-containing and Ate1-deficient mice (Tables 2 and 3).
To measure metabolic rate, we employed indirect calorimetry (see Materials and Methods), determining O 2 consumption and CO 2 production by mice under resting conditions. The metabolic rate (resting metabolic rate, RMR) of Ate1-deficient mice was indeed higher than normal: they consumed on average 46.12 ml of O 2 per kg per min, versus 29.3 ml of O 2 per kg per min for Ate1-containing mice (Fig. 10E). In contrast, the respiratory exchange ratio, RER (the ratio of CO 2 eliminated from the lungs to O 2 taken into the lungs), a parameter that depends on a preferred source of fuel (e.g., carbohydrates versus fat), was similar for Ate1-deficient and Ate1-containing mice: 0.75 and 0.76, respectively (Fig. 10F).
The S. cerevisiae N-end rule pathway regulates the import of short peptides through the conditional degradation of Cup9, the import's repressor [34]. It is likely (but remains to be verified) that the N-end rule pathway regulates the transmembrane traffic of peptides in mammals as well. To address the possibility that significantly lower weights (despite hyperphagia) of Ate1-deficient mice might stem, at least in part, from an impaired ability to import peptides and/or amino acids from their gastrointestinal (GI) tract, we labeled E. coli with a mixture of 14 C-amino acids and isolated a 14 C-protein fraction that was essentially free of nucleic acids, fatty acids, lipids and carbohydrates (see Materials and Methods). Ate1-deficient and Ate1-containing mice were gavaged with a bolus of this 14 C-protein preparation, followed by measurements of 14 C in several organs of these mice (and in their feces) as a function of time post-gavage. Ate1-deficient mice passed less 14 C in feces than Ate1-containing mice (Fig. 9D). Moreover, Ate1-deficient mice accumulated more of 14 C in their brains, livers, spleens, kidneys and hearts than Ate1-containing mice (Figs. 9C and 10D). Irrespective of mechanistic causes involved (they remain to be understood), higher than wild-type levels of protein-derived 14 C delivered to tissues of Ate1-deficient mice indicated the absence of significant defects in their transport of peptides and/or amino acids from GI tract.
Given the metabolic and behavioral abnormalities of Ate1deficient mice (Figs. 6C-E, 8A, D and 10E, F), we also examined them for expression of neuropeptides. As we would be interested, at present, only in strong differences, a semiquantitative RT-PCR was employed. Using total RNA from hypothalami of Ate1deficient versus Ate1-containing mice, we found no consistent differences between these mice in regard to the levels of hypothalamic mRNAs that encoded the agouti-related protein (AgRP) and the neural peptide Y (NPY) (Fig. 10G). Strikingly, however, there was a consistent and strong decrease of expression, in Ate1-deficient mice, of mRNA encoding proopiomelanocortin (POMC) (Fig. 10G). POMC is a precursor of several neurohormones with broad systemic and brain-specific functions ( [35] and refs. therein). These functions include a role in melanocyte regulation (a process that is likely to be perturbed in Ate1-deficient mice; see above) and a down-regulation of food intake (the observed deficiency in POMC is consistent with hyperphagia of Ate1-deficient mice (Fig. 10C, G)). Similarly to a connection between Ate1 and the Ucp1 uncoupling protein (Fig. 7G-I), our finding of a link between N-terminal arginylation and the expression of POMC is likely to provide an experimental route to identifying the relevant Ate1-dependent circuit.

Discussion
A cell is alive owing to a cell-wide dynamic network of structurally or functionally interacting biopolymers. Some parts of this network can be sufficiently insulated, through their design, to be considered, in the first approximation, as distinct circuits. The N-end rule pathway is one such circuit. Its enzymes receive as their input specific degron-bearing proteins and convert them, through deamidation, arginylation, polyubiquitylation and processive degradation, into an output of proteolysis-derived short peptides (Fig. 1A). The rate and selectivity of the proteasomemediated protein degradation by the N-end rule pathway are modulated by physiological effectors, including specific phosphokinases, short peptides, redox, heme and nitric oxide (see Introduction). Some of N-end rule substrates are produced by proteases that include MetAPs, separases, caspases and calpains. These and other nonprocessive proteases, which function as upstream components of the N-end rule pathway, have in common their ability to convert, through a cleavage, a pro-Ndegron into an N-degron. Figure 10. Energy balance and metabolic rate in Ate1-deficient mice. (A) Glucose tolerance test. Glucose concentration (mg/dL) in whole blood of Ate1-containing mice (n = 15; black curve) and Ate1-deficient mice (n = 11; red curve), at different times after a bolus of glucose by gavage, following a 16-hr fast. Glucose was administered at time zero. Error bars indicate 6SD. (B) Fasting blood glucose levels. Average blood glucose levels (mg/dL) in Ate1-containing mice (n = 15; black bar) and Ate1-deficient mice (n = 11; red bar), with measurements shortly before glucose gavage (after a 16-hr fast) and 6 hr after the gavage in A. Standard deviations are indicated. Statistical analysis was performed using an unpaired t-test (p,0.04). (C) Average daily energy consumption (kcal/gm of body weight) for Ate1-containing mice (n = 5; black curve) and Ate1-deficient mice (n = 3; red curve), with measurements from 1 week prior to tamoxifen (TM) treatment. Vertical arrow indicates the beginning of a 5-day TM treatment. Error bars indicate 6SD. (D) Relative efficiencies of the import of 14 C-amino acids and 14 C-peptides from gastrointestinal tract in Ate1-containing mice (black bars) versus Ate1-deficient mice (red bars). Shown here are representative comparisons of the retention of 14 C (in cpm/gm) in the brains, livers, spleens, kidneys, and hearts of indicated mice 16 days after gavage with a single bolus of 14 C-labeled proteins (see Materials and Methods). Mice were gavaged 26 days after TM treatment. (E) Comparison of resting metabolic rate (RMR) (measured in O 2 (ml) consumed per kg of body weight per min) for Ate1-containing mice (n = 6; black bar) versus Ate1-deficient mice (n = 6; red bar). Standard deviations are indicated in E and F. Statistical analysis was performed using an unpaired t-test (p,0.008). (F) Comparison of the respiratory exchange ratio (RER), measured as CO 2 (in ml) per ml of O 2 , for Ate1-containing mice (n = 6; black bar) and Ate1-deficient mice (n = 6; red bar) mice. No statistically significant difference in RER was observed. (G) RT-PCR analyses of AgRP, MCH, HPY, and POMC mRNA levels in the hypothalami of TM-treated Ate1-containing mice (Sets 1 and 4) versus Ate1-deficient mice (Sets 2 and 3). Set 1, Ate1 flox/flox (+/+); Set 2, Ate1 flox/flox ;CaggCreER (2/2); Set 3, Ate1 flox/flox ;CaggCreER (2/2); Set 4, Ate1 flox/+ ;CaggCreER (+/2). In sets 1 and 2, hypothalami were isolated 93 days after TM treatment. In sets 3 and 4 hypothalami were isolated ,1 year after TM treatment. Sloping triangles indicate decreasing inputs (by 2-fold) of total RNA. doi:10.1371/journal.pone.0007757.g010 The present study expanded the earlier understanding of the Ate1 R-transferase (Fig. 1A) by making possible a postnatal inactivation of mouse Ate1. (Unconditional deletion of Ate1 results in embryonic lethality [10].) Described and discussed in Results is a large set of defects, some of them quite striking, in juvenile and adult mice upon the postnatal inactivation of Ate1 and the resulting loss of N-terminal arginylation (Figs. 1C, D). The initial abnormality is a rapid decrease of body weight and early death of ,15% of Ate1-deficient mice, with surviving mice attaining, gradually, only ,70% of the weight of wild-type mice identically treated with tamoxifen (TM) (Fig. 4A-C). Both ''partial'' lethality and the transiency of acute crisis, over ,3 weeks after TM treatment (red arrow in Fig. 4B), remain to be understood in molecular terms. This crisis and subsequent failure to thrive occur despite higher than normal food intake by Ate1-deficient mice (Fig. 10C). These mice contain little or no visceral fat (Figs. 5D and 7A-E), and exhibit an increased metabolic rate (Fig. 10E), resistance to diet-induced obesity (Fig. 7F), enlarged brains (Figs. 5D and 6A), kyphosis (Fig. 5B), a striking hyperkinesia (Figs. 6C-E and 8A), and male sterility (Fig. 6F-K).
Owing to current constraints of the Cre-lox technology ([25] and refs. therein), the extent of Ate1 inactivation, while nearly 100% in some mouse tissues, was variable in others. The TMinduced Ate1 2/2 ;CaggCreER mouse strains are thus mosaics of Ate1 flox/2 and Ate1 2/2 cells, where Ate1 2/2 cells are a great majority in most organs, such as the brain, but even there do not reach 100% of all cells (see Results). The initial weight loss upon the TM-induced conversion of Ate1 flox/2 mice to Ate1 2/2 mice was accompanied by death of ,15% of Ate1 2/2 mice (Fig. 4A, B). Such a ''partially'' lethal phenotype suggests that an adult-onset Ate1 2/2 genotype in all cells (as distinguished from most cells) of a mouse might be incompatible with viability, similarly to the  embryonic lethality of unconditional Ate1 2/2 mice. Thus, paradoxically, the discovery, in the present study, of specific Ate1-linked defects in adult mice might have been made possible by incomplete penetrance of the Cre-induced conversion of Ate1 flox to the Ate1 2 allele. The set of definitively identified mammalian N-end rule substrates that involve N-terminal arginylation consists, at present, of fewer than 10 proteins. They subserve different functions, from chromosome segregation to control of apoptosis and regulation of G proteins (see Introduction). This set is the tip of the iceberg, as several considerations [2], in addition to our findings above, strongly suggest a larger number of physiological Ate1 substrates. Given this complexity, the specific and often striking phenotypes of Ate1-deficient mice that were discovered in the present work will be of major assistance in deciphering the underlying Ate1 circuits.

Animal Care and Treatments
All animal care and procedures in the present study were conducted according to the relevant NIH guidelines, and were approved (Protocol #1328) by the Institutional Animal Care and Use Committee, the Office of Laboratory Animal Research (OLAR) at the California Institute of Technology, where the entire present study was carried out. Mice were housed at ,22uC, at a pathogen-free (barrier) facility, using a 12 hr light/12 hr dark cycle, with Laboratory Rodent Diet 5001 (PMI International, Richmond, IN) ad libitum. Mice aged between 3 and 8 weeks were treated with tamoxifen (TM) (Sigma) (2 mg in 0.2 ml sesame oil) by daily intraperitoneal (IP) injections over 5 days. Mice were weighed weekly, starting 3 days before the first TM treatment. For a high-fat diet (HFD) study, mice were fed ad libitum a diet containing 35.5% fat (BioServe, Frenchtown, NJ), and were weighed weekly.

Construction of Ate1 flox/2 ;CaggCreER Mouse Strains
Mouse genomic DNA encoding Ate1 was isolated from a BAC library [10]. Two pBluescript-based plasmids were used to construct the targeting vector. In one insert, a ,12 kb HindIII fragment contained Ate1 exons 1a, 1b, 2, and 3 as well as ,1.9 kb of DNA 39 of exon 3 (ending just before exon 4). In the other insert, a ,2.9 kb fragment contained Ate1 exons 4 and 5. The entire ,12 kb HindIII fragment and a part of the ,2.9 kb fragment were modified as described below and assembled into a final ,22.5 kb targeting vector consisting of the following parts ( Fig. 2C): (i) pBR322 backbone (New England Biolabs, Ipswich, MA); (ii) a ,6.3 kb ''long arm'' of Ate1 homology containing the Ate1 exon 1a, the bidirectional P Ate1 promoter [14], and exon 1b; (iii) A single loxP site ,300 bp upstream of Ate1 exon 2; (iv) a ,2 kb fragment that contains, 50 bp downstream of Ate1 exon 4, a ''floxed'' Hph (hygromycin) antibiotic-resistance marker, expressed from the P PGK promoter [36]; (v) a ,1.2 kb ''short arm'' of Ate1 homology that spans most of the intron between exons 4 and 5; (vi) a gene encoding HSV-TK (herpes simplex virus thymidine kinase), expressed from the P PGK promoter. The targeting vector was linearized with BamHI and electroporated into CJ7 embryonic stem (ES) cells (a gift from Dr. Thomas Gridley, formerly of Jackson Laboratories, Bar Harbor, ME). ES cells were grown in DMEM supplemented with 15% fetal bovine serum (FBS), 0.1 mM non-essential amino acids, 0.1 mM b-mercaptoethanol, 2 mM glutamine, 100 U/ml penicillin, 0.1 mg/ml streptomycin, 1 mM pyruvate, and 1,000 U/ml leukemia inhibitory factor (LIF) [37], using a feeder layer of hygromycin-resistant mouse primary fibroblasts that had been treated with 10 mg/ml mitomycin C for 3 hr at 37uC. Selection with hygromycin (at 0.2 mg/ml) and 1-(29-deoxy, 29-fluoro-b-D-arabinofuranosyl)-5iodouracil (FIAU; at 0.4 mM) was started 24 hr after electroporation. Correctly targeted ES cell clones that contained ''tri-loxed'' Ate1 allele (Figs. 2C, D) were identified using Southern hybridization and PCR. Southern DNA probes and positions of primers for PCR are indicated in Fig. 2.
Two correctly targeted, independently produced ES cell lines that had apparently normal karyotypes were injected into 3.5-dayspostcoitum C57BL/6J blastocysts and implanted into pseudopregnant females. The resulting male chimeric offspring were mated with C57BL/6J females. In some of the progeny, ''floxed'' ES cells became a part of germ line. Standard mating techniques [36,38] were then used to produce, initially, mouse strains that contained a ''tri-lox'' Ate1 configuration, in that they also contained the floxed positive-selection P PGK -hph cassette (Fig. 2D). This DNA segment was removed by mating Ate1-tri-lox heterozygotes with EIIa-Cre mice that expressed Cre recombinase only in early, pre-implantation blastocysts [39,40,41]. Owing to the presence of three loxP sites at the initial floxed Ate1 locus, F1 progeny from this cross were mosaic, i.e., their tissues, including germ line, contained varying configurations of retained loxP sites, depending on specific patterns of Cre-mediated recombination ( Fig. 2A-F). To isolate a mouse strain with the desired configuration of (retained) loxP sites (Fig. 2E), the above F1 mosaic mice were mated to wild-type C57BL/6 mice. This produced, among other progeny, a strain that lacked the P PGK -Hph cassette and had the desired Ate1 flox/+ genotype, in the (mixed) C75BL/6J-129SvEv background.

Southern Hybridization and PCR
Total genomic DNA was isolated from ES cells by washing them twice with phosphate-buffered saline PBS, followed by an overnight incubation at 50uC in 10 mM EDTA, 10 mM NaCl, 0.5% Sarcosyl,10 mM Tris-HCl (pH 7.5) containing Proteinase K at 0.2 mg/ml. Thereafter an equal volume of 75 mM NaCl in 100% ethanol was added. Precipitated genomic DNA was then gently washed twice with 70% ethanol and resuspended in T 10 E 0.1 buffer (10 mM Tris (pH 8.0), 0.1 mM EDTA). Total genomic DNA was isolated from mouse tails or other tissues by overnight incubation at 55uC, with constant rotation, in 5 mM EDTA, 0.2 M NaCl, 0.3% SDS, 0.1 M Tris (pH 8.5) containing Proteinase K at 0.4 mg/ml. follows. A two-primer PCR using the CB156F and CB157R primers (Table 4) was employed to produce and detect a 512 bp fragment of the Ate1 flox (''floxON'', active) allele as well as a 472 bp fragment of the wild-type Ate1 + allele. A four-primer PCR using the CB110F, CB157R, OIMR0042, and OIM0043 primers Z (Table 4) was employed to produce and detect a 470 bp fragment of the Ate1 floxderived Ate1 2 allele (''floxOFF'') as well as a (control) 324 bp fragment of the Il-2 gene. A three-primer PCR using the AK49, YT641, and AK83-CBfix (Table 4) was employed to detect both a 300 bp fragment of the unconditional Ate1 2 allele [10] and a 560-bp fragment of the wild-type Ate1 + allele. A four-primer PCR using the Cre-1, Cre-2, CB159R, and CB160F primers (Table 4) was employed to detect both a 320 bp fragment of the CaggCreER transgene as well as a 1,060 bp fragment of the wild-type Ate1 + allele. All PCR reactions except for those to detect the CaggCreER transgene were carried out using HotStar Taq DNA polymerase, standard buffer conditions (Qiagen, Valencia, CA), 35 cycles of template denaturation for 30 seconds at 95uC, followed by primer annealing for 30 seconds at 60uC and primer extension for 1 minute at 72uC. PCR reactions for detecting CaggCreER were carried out using 30 cycles of template denaturation for 30 seconds at 95uC, followed by primer annealing for 30 seconds at 58uC and primer extension for 45 seconds at 72uC.

Northern and RT-PCR Analyses of RNA
Total RNA was isolated from various mouse tissues using the RNeasy Protect Mini Kit (Qiagen). Tissue disruption and homogenization were done in Buffer RLT and the MP FastPrep-24 instrument with Lysing Matrix D (MP Biomedicals, Solon, OH) for 2 runs at 6.5 m/s. First-strand cDNA was primed with oligo-dT using the SuperScript III First-Strand Synthesis System (Invitrogen, Carlsbad, CA) and PCR was carried out using primers cited in legends to the corresponding figures and in Table 4.

Tissue Extracts and Immunoblotting
Various mouse tissues were harvested and lysed in ''Tissue Lysis Buffer'' (10% glycerol, 0.05% NP40, 0.15 M NaCl, 2 mM EDTA, 1 mM dithiothreitol (DTT) 1 mM phenylmethylsulfonyl fluoride PMSF 50 mM HEPES, pH 7.5) plus freshly dissolved ''Complete EDTA-Free Protease Inhibitors'' (Roche), using the MP FastPrep-24 instrument and Lysing Matrix D (MP Biomedicals, with 2 or 3 runs at 6.5 m/s for 25 sec each, and with 5-min incubations on ice between the runs. The lysates were centrifuged at 10,000g for 20 min at 4uC. The supernatants were fractionated by SDS-12.5% PAGE, transferred to Immobilon-P PVDF membranes (Millipore, Billerica, MA), and analyzed by immunoblotting (IB) with antibodies indicated in specific figures. Immunoblots were visualized using SuperSignal West Pico or SuperSignal West Dura reagents (Thermo Scientific, Rockford, IL) according to the manufacturer's instructions.

Other Analyses of Mouse Tissues
Specific mouse tissues were dissected immediately after euthanasia by CO 2 inhalation. The tissues washed with PBS, blotted dry on Kimwipes, and weighed (wet). For dry-weight measurements, mouse brains were dissected intact, washed in PBS, blotted dry on Kimwipes, weighed wet, then incubated overnight in acetone. After acetone incubation, individual brains were lyophilized until their (dry) weight no longer decreased.
For routine histological examinations, tissues or organs were fixed in Bouin's solution or in 4% formaldehyde, using standard procedures [37]. Fixed samples were embedded in paraffin, sectioned, and stained with hematoxylin and eosin. To stain for LacZ (NLS-bgal), dissected tissues or organs were fixed in LacZfix (0.2% glutaraldehyde, 5 mM EGTA, 0.1 M MgCl 2 in PBS (pH 7.3)) for 4 hr, rinsed twice with PBS, dehydrated overnight at 4uC in 30% sucrose, 2 mM MgCl 2 in PBS, and embedded and frozen in Tissue-Tek O.C.T. Compound (Sakura Finetek USA, Inc. Torrance, CA). Cryosections (prepared using a Tissue Tek Microtome/Cryostat model 4553) were mounted onto glass slides, fixed in LacZfix for 10 min at RT, washed 3 times in LacZWash (0.02% NP40, 01% Na-deoxycholate, 2 mM MgCl 2 in PBS), and stained overnight at 37uC with LacZ stain (LacZWash containing 1 mg/ml XGal, 5 mM K 4 Fe(CN) 6 and 5 mM K 3 Fe(CN) 6 ). Stained sections were washed with PBS and mounted with Permount for light microscopy. Apoptosis was assessed by TUNEL, a nuclear DNA fragmentation assay, using a TUNEL kit (Roche, Indianapolis, IN), fluorescein-dUTP, and manufacturer's instructions. Cell proliferation was assayed using the Click-It Edu Cell Proliferation kit (Invitrogen).

Blood and Urine Analyses
Blood (,0.6 ml per mouse) was withdrawn by cardiac puncture and transferred into BD Microtainer SST tubes (BD, Franklin Lakes, NJ). The serum fraction was prepared by centrifugation in a microcentrifuge after clotting occurred, immediately frozen in liquid N 2 and stored at 280uC. The levels of glucose, cholesterol, sodium, potassium, chloride, calcium, phosphorus, blood urea nitrogen, creatine, total protein, albumin, total bilirubin, aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase, c-glutamyltransferase (gamma gt), as well as the T3 and T4 hormones were determined by Phoenix Central Laboratories (Everett, WA).
Urine was obtained by placing the external urethra over a test tube. Urine samples collected from Ate1-deficient mice were pooled and compared with pooled urine from Ate1-containing mice. The levels of glucose, bilirubin, ketones, blood, protein, urobilinogen, nitrite, leukocytes, as well as the pH and specific gravity were determined using the Multistix 10 SG Reagent Strips (Bayer, Tarry Town, NY).

Measurements of Body Temperature and Cold Sensitivity
Mice (housed at one mouse per cage, without bedding but with food and water) were exposed to a 4uC environment for up to 6 hr Their core body temperature was monitored every 30 min via a rectal probe digital thermometer (Thermalert TH-8; Physitemp Instruments., Clifton, NJ).
Measurements of 14 C-Protein Uptake from Gastrointestinal Tract E. coli DH5a cells were grown in minimal M9 media supplemented with 0.5% glucose and 50 mCi of 14 C-amino acids (derived from 14 C-protein hydrolysate (Amersham)) until the incorporation of ,70% of the added 14 C amino acids. Cells were lysed by one freeze/thaw cycle in PBS containing 1 mg/ml lysozyme and incubated with RNase H and DNase I for 45 min at 37uC. A crude protein fraction was isolated by precipitation with cold 10% TCA (CCl 3 COOH), and the pellet was washed with icecold acetone. The pellet was redissolved in PBS, with a brief sonication to facilitate solubilization. 0.2 ml of the resulting sample, containing 260,000 cpm of 14 C-labeled E. coli proteins was fed to a mouse by oral gavage. Urine and feces was collected at various times post-gavage. Total 14 C was measured, using a scintillation spectrometer, in feces, urine, and (eventually) in mouse tissue samples that were collected either 48 hr or 15 days post-gavage.

Measurements of Glucose Uptake
For glucose analyses, mice were fasted for 24 hr, then gavaged with 50 mg glucose in 0.2 ml of water. Blood was collected through the lateral tail vein at 15, 30, 60, 90, 120, and 360 min post-gavage. Blood glucose levels were determined using the OneTouch UltraMini Blood Glucose Monitoring System (Life-Scan, Johnson and Johnson, Milpitas, CA).

Metabolic Analyses
The resting metabolic rate was determined at the Mouse Physiology Laboratory in the Department of Physiology at the Geffen School of Medicine, UCLA using indirect calorimetry as previously described [43,44]. Single mice were placed into a custom-made enclosed plexiglas chamber (25 cm612 cm67.5 cm, with 4 room air intake vents and one outflow port) and allowed to come to rest over a period of 30 min to 2 hr. Outflow of expired gases was sampled by the gas analyzer and recorded using a computerized acquisition system during a 30-min resting interval.

MRI Analyses of Mouse Brains
The procedures used were essentially the same as previously described [45]. Mice were given an IP injection of 40 mg/kg of Na-pentobarbital (Nembutal, Hospira, Inc., Lake Forest, IL). Once fully anesthetized, mice were transcardially perfused with 4% formaldehyde in PBS. MRI analysis was performed by the Caltech Brain Imaging Center. Briefly, mouse heads were excised and postfixed in 4% formaldehyde/PBS overnight. Hair and skin were removed from fixed heads, which were then soaked in 5 mM Gadolinium-based MR contrast agent (Prohance Bracco Diagnostics, Durham, NC) for 10 days, to decrease the intrinsic tissue relaxation rates and improve the MR acquisition efficiency. A gradient echo sequence (TE/TR = 8 msec/50 msec, 16 averages) was used to acquire 3D data sets of the mice heads, using a Bruker 7T Biospec animal magnet system. Images were reconstructed with an isotropic resolution of ,90 mm and analyzed using Brainsuite 2 software [46].

Open-Field and Startle Response Tests
For the open-field activity measurements, individual mice were placed into a square chamber (50 by 50 cm). Movements along the x and y axes were tracked and analyzed using Ethovision software (Noldus, Leesburg, VA) over 15-min intervals. Startle response tests were carried out essentially as described previously [47].