Biochemical Studies of Zmpste24 -deficient Mice*

Genetic studies in Saccharomyces cerevisiae identified two genes, STE24 and RCE1 , involved in cleaving the three carboxyl-terminal amino acids from isoprenylated proteins that terminate with a CAAX sequence motif. Ste24p cleaves the carboxyl-terminal “- AAX” from the yeast mating pheromone a-factor, whereas Rce1p cleaves the - AAX from both a-factor and Ras2p. Ste24p also cleaves the amino terminus of a-factor. The mouse genome contains orthologues for both yeast RCE1 and STE24 . We previously demonstrated, with a gene-knock-out experiment, that mouse Rce1 is essential for development and that Rce1 is entirely responsible for the carboxyl-terminal proteolytic processing of the mouse Ras proteins. In this study, we cloned mouse Zmpste24 , the orthologue for yeast STE24 and showed that it could promote a-factor production when expressed in yeast. Then, to assess the importance of Zmpste24 in development, we generated Zmpste24 -deficient mice. Unlike the Rce1 knockout mice, Zmpste24 -deficient mice survived development and were fertile. Since no natural substrates

Proteins that terminate in a carboxyl-terminal CAAX motif 1 undergo three sequential enzymatic processing events, farnesylation or geranylgeranylation of the cysteine, endoproteolytic release of the last three amino acid residues of the protein (i.e. removal of the -AAX), and methylation of the new carboxyl terminus of the protein by isoprenylcysteine carboxyl methyltransferase (1,2). The yeast genes responsible for the farnesylation and methylation steps were identified more than a decade ago (3,4), but the identification of the genes responsible for the middle processing step, the endoprotease step, remained elusive for years (2). Ultimately, however, Boyartchuk and co-workers (5) applied a novel genetic selection scheme and identified two yeast genes, RCE1 and STE24 (AFC1), involved in the carboxyl-terminal endoproteolytic processing of isoprenylated CAAX proteins. Rce1p is a protease involved in the carboxyl-terminal processing of both a-factor and the yeast Ras protein, Ras2p. Ste24p (Afc1p), a zinc metalloprotease, lacked activity against Ras2p but did process a-factor. Haploid MATa yeast lacking both RCE1 and STE24 (ste24⌬rce1⌬) grew normally but were unable to produce mature a-factor and therefore were sterile (5). Interestingly, Rce1p and Ste24p exhibited subtle differences in substrate specificities (5)(6)(7). Both proteins were capable of cleaving the carboxyl terminus of wild-type a-factor, which terminates in CVIA. However, only Ste24p processed an a-factor mutant terminating in CAMQ, and only Rce1p processed an a-factor mutant terminating in CTLM. 2 The identification of Ste24p as a CAAX endoprotease was initially surprising because Ste24p had also been described as a protease that cleaves the amino terminus of the a-factor precursor protein (8), and it seemed improbable that a single protease would recognize and cleave a single protein at two completely different sites. Subsequent studies, however, established that Ste24p does indeed play dual roles in a-factor processing (6,9). In addition to being a CAAX endoprotease, Ste24p removes the seven amino-terminal amino acids from the a-factor precursor. Ste24 is, however, only the first of two proteases that cleave the amino terminus of a-factor. After the Ste24pmediated cleavage step, an additional cleavage by Axl1p releases mature biologically active a-factor.
When the two CAAX endoproteases were identified in yeast, apparent orthologues already existed in mammalian cDNA data bases. Tam and co-workers (9) obtained a full-length cDNA for human ZMPSTE24, the orthologue for yeast STE24, and showed that its amino acid sequence was 36% identical to the yeast protein. No a-factor orthologue has been discovered in mammals, and no natural substrates for human ZMPSTE24 have been identified. However, the human enzyme faithfully carried out both the amino-terminal and carboxyl-terminal processing of a surrogate substrate, yeast a-factor, and restored a-factor production in MATa ste24⌬rce1⌬ yeast (9,10). ʈ Contributed equally to the results of this work. § § To whom reprint requests should be addressed: Gladstone Institute of Cardiovascular Disease, P. O. Box 419100, San Francisco, CA 94141-9100. Tel.: 415-695-3774; Fax: 415-285-5632; E-mail: syoung@ gladstone.ucsf.edu. 1 The abbreviations used are: CAAX motif, carboxyl-terminal CAAX motif is a cysteine (C), a pair of aliphatic amino acids (A), and then a final amino acid residue (X); bp, base pair; PCR, polymerase chain reaction; BAC, bacterial artificial chromosome; ES, embryonic stem; kb, kilobase pair(s); PAGE, polyacrylamide gel electrophoresis.
Our laboratory has sought to define the physiologic importance of the mammalian CAAX endoproteases, as well as the CAAX methyltransferase, isoprenylcysteine carboxyl methyltransferase (ICMT). We cloned and expressed cDNAs for human and mouse Rce1 and showed that the mammalian Rce1 proteins processed the Ras proteins as well as several other CAAX proteins (11,12). Rce1 knockout mice died during embryonic development (11). However, studies with Rce1-deficient embryos and embryonic fibroblasts showed that Rce1 is essential for Ras processing (11). Like Rce1, ICMT is essential for the processing of mammalian CAAX proteins and also for survival during embryonic development (13).
In the current study, we have turned our attention to cloning mouse Zmpste24 and producing Zmpste24-deficient mice. When we initiated this project, we had several physiologic and biochemical issues in mind. First, we wanted to gauge the physiologic importance of Zmpste24 in mammals. Would Zmpste24 be required for embryonic development, as was the case for Rce1 and ICMT? If Zmpste24-deficient mice survived, would they be fertile? From the biochemical perspective, we wanted to determine whether mouse Zmpste24 would be capable of carrying out the carboxyl-terminal processing of a-factor and, if so, whether the mouse enzyme would manifest the same peculiar specificities as the yeast enzyme (e.g. the ability to cleave an a-factor mutant terminating in CAMQ). We also wanted to determine whether the mouse enzyme would cleave the amino terminus of yeast a-factor and, if so, whether this enzymatic activity would be present in detectable levels in the tissues of wild-type mice. Finally, we wanted to determine if mouse Zmpste24 was the only mammalian enzyme capable of carrying out the amino-terminal processing of a-factor, or whether mammalian cells might contain more than one enzyme with that activity. Our current studies provide insights into each of these issues.

EXPERIMENTAL PROCEDURES
A Mouse Zmpste24 cDNA Clone-A XREFdb query (14) of Gen-Bank TM sequences with yeast STE24 sequences identified a human expressed sequence tag with sequence similarities (I.M.A.G.E. clone 284572, Research Genetics, Huntsville, AL). Oligonucleotide primers 5Ј-GAGTTTCAAGCTGATGCATTTGC-3Ј and 5Ј-GAGTTTCAAGCTGA-TGCATTGC-3Ј derived from the human sequence were used to amplify a 382-base pair (bp) fragment (representing sequences from Zmpste24 exons 9 and 10) from a mouse liver cDNA library. The 5Ј portion of the mouse Zmpste24 cDNA was generated by 5Ј rapid amplification of cDNA ends with oligonucleotide 5Ј-AGTCTTTAGCCTTCCCAAGTTTC-TTGGC-3Ј (corresponding to Zmpste24 amino acids 422-430) and the Marathon-Ready mouse liver cDNA kit (CLONTECH, Palo Alto, CA). Oligonucleotides 5Ј-ACTCTCCGAGGGACGCGTGT-3Ј and 5Ј-CTTAA-GAGCATCCAGTCATG-3Ј were used to amplify a mouse Zmpste24 cDNA. The cDNA was cloned into a TOPO TA cloning vector (Invitrogen, Carlsblad, CA), and both strands were sequenced. Predicted transmembrane domains were determined with the TMHMM transmembrane domain analysis program (genome.cbs.dtu.dk/services/TMHMM-1.0/). Amino acid sequences from different species were aligned with Mac-Vector 6.5.
Expression of Zmpste24 in different mouse tissues was assessed by hybridizing a 32 P-labeled Zmpste24 cDNA probe to a mouse Multiple Choice Northern blot (OriGene Technologies, Rockville, MD) and to a mouse multiple-tissue poly(A) ϩ RNA blot (CLONTECH). Hybridization and washing were performed as described previously (12). The blots were exposed to x-ray film for 72 h at Ϫ80°C.
Documenting That Mouse Zmpste24 Processes Yeast a-Factor-To test the expression vectors (pMB4, pMB5, pMB6, pMB7, pMB8, and pMB9), each was transformed into MATa ste24⌬rce1⌬ yeast (JRY5317). The ability of each expression vector to restore a-factor production was assessed with a standard pheromone diffusion (halo) assay (7). The principle underlying the halo assay is that the production of mature a-factor by a MATa yeast strain results in a zone of growth inhibition (halo) on a lawn of MAT␣ sst2 cells.
Analyzing the Substrate Specificity of Mammalian Endoproteases with Halo Assays-The coding regions of human Rce1 and mouse Zmp-ste24 were amplified by PCR, as described earlier, and cloned into pACA1 (5). Site-directed mutagenesis was used to change the CVIA CAAX box in the a-factor gene (MFA1) to CAMQ or CTLM. The mutant a-factor plasmids were transformed into a MATa ste24⌬rce1⌬mfa1⌬mfa2⌬ strain (JRY5463) (6). Yeast strains expressing the a-factor mutants and either human RCE1 or mouse Zmpste24 were spotted onto a lawn MAT␣ sst2 cells (JRY3443) and allowed to grow for 2 days at 30°C. The production of mature a-factor from a mutant a-factor construct was evident from a halo in the lawn of MAT␣ sst2 cells.
Analyzing the Substrate Specificity of Mammalian Endoproteases with a Coupled Proteolysis/Methylation Assay-The ability of yeast and mouse endoproteases to process 15-mer a-factor peptides terminating in CAMQ or CVIA was assessed in a coupled proteolysis/methylation assay (5,11,12). A wild-type CVIA a-factor peptide (YIIKGVFWDPA-[farnesyl-C]VIA) and a mutant CAMQ a-factor peptide (YIIKGVFWD-PA[farnesyl-C]AMQ) were purchased from California Peptide Research (Napa, CA). Membrane fractions were prepared from ste24⌬rce1⌬ yeast (JRY5317) that had been transformed with pMB4, pMB5, pMB6, pMB7, or pMB9. Yeast membranes (100 g) were mixed with 4 nmol of peptide and 10 M S-adenosyl-L-[methyl-14 C]methionine (55 Ci/mol, Amersham Pharmacia Biotech, Piscataway, NJ) and incubated for 120 min at 37°C. The methylation reaction was stopped by adding 50 l of 1.0 M NaOH containing 0.1% SDS. The reaction mixture (90 l) was spotted onto a pleated 2 ϫ 8-cm filter paper wedged in the neck of a 20-ml scintillation vial containing 5 ml of scintillation fluid (ScintiSafe Econo 1, Fisher). The vials were capped and incubated at room temperature for 5 h to allow the [ 14 C]methanol (formed by base-hydrolysis of methyl esters) to diffuse into the scintillation fluid (20 -22).
Construction of a Zmpste24 Gene-Targeting Vector-The screening of a strain 129/Sv bacterial artificial chromosome (BAC) library (Research Genetics) with the 382-bp cDNA fragment resulted in the identification a BAC clone (223 G6) that contained the 3Ј portion of mouse Zmpste24. A 9.4-kb SpeI fragment containing Zmpste24 exons 7-9 was subcloned into pBSSK (Stratagene, La Jolla, CA). These sequences were used to construct a sequence-replacement gene-targeting vector in pKSLoxPNT (23). pKSLoxPNT contains polylinker cloning sites, a thymidine kinase (tk) gene, and a 3-phosphoglycerate kinase-neomycin resistance (PGKneo) cassette flanked by loxP sites. The long arm (4.6 kb of intron sequences 5Ј to exon 8) was amplified from BAC DNA with oligonucleotides 5Ј-CAAGTGATTCGAGGCTAGCCTGGTCTACG-3Ј and 5Ј-GGCC-TTGACCCATCTAATCAAACAACCAGAC-3Ј and the Expand Long Template PCR System (Roche Molecular Biochemicals, Indianapolis, IN). The PCR fragment was cloned into the EcoRI site of pKSLoxPNT. The short arm (2.6 kb of intron sequences 3Ј to exon 8) was amplified with oligonucleotides 5Ј-ACGACGGCGGCCGCGAAACTGGCCTGGT-GTTCACTATGTAGC-3Ј and 5Ј-AGCAGCGCGGCCGCGAGGGCACA-ACACTGTATGTTACCAGG-3Ј and cloned into the NotI site. The integrity of the vector was verified by restriction mapping and DNA sequencing. The vector was linearized with XhoI before the electroporation of mouse embryonic stem (ES) cells.
Two targeted clones, each with a single neo integration, were used to generate Zmpste24-deficient mice (25). All mice had a mixed genetic background (ϳ50% C57BL/6 and ϳ50% 129/Sv). The mice were weaned at 21 days of age, housed in a barrier facility with 12-h light/dark cycle, and fed a chow diet containing 4.5% fat (Ralston Purina, St. Louis, MO). All mice described herein were examined at 8 -12 weeks of age. Zmpste24Ϫ/Ϫ, Zmpste24ϩ/Ϫ, and Zmpste24ϩ/ϩ embryonic fibroblasts were prepared from 13.5-day mouse embryos (11).
Production of Chimeric Mice from Zmpste24Ϫ/ϪES Cells-Homozygous knockout cells (Zmpste24Ϫ/Ϫ) were isolated as described by Mortensen et al. (26). Two lines of Zmpste24Ϫ/Ϫ ES cells were injected into C57BL/6 blastocysts to generate chimeric mice. At 8 weeks of age, the chimeric mice were sacrificed, and genomic DNA from multiple tissues was analyzed with Southern blots with the 5Ј probe.
Bioassays to Assess Carboxyl-terminal Processing of an a-Factor Peptide by Membranes from Mouse Cells and Tissues-The ability of Zmpste24Ϫ/Ϫ membranes to cleave the carboxyl terminus of an a-factor peptide, generating mature a-factor, was assessed with a bioassay. 3 Membranes from tissues of Zmpste24ϩ/ϩ and Zmpste24Ϫ/Ϫ mice and from Zmpste24ϩ/ϩ and Zmpste24Ϫ/Ϫ cells were mechanically disrupted by sonication (Branson sonifier, 10 cycles, output ϭ 5, duty cycle ϭ 50%) in an ice-cold hypotonic lysis buffer (10 mM Tris-HCl, 1 mM MgCl 2 , 1 M dithiothreitol, pH 7.4 containing protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 2 g/ml aprotinin, and 3 g/ml leupeptin, chymostatin, and pepstatin). The samples were then homogenized on ice with a Dounce homogenizer, resuspended in 0.15 M NaCl, and then subjected to centrifugation at 4°C for 3000 ϫ g for 10 min. The pellet was discarded and the supernatant fluid was spun at 100,000 ϫ g for 30 min at 4°C. The supernatant fluid was discarded and the pellet, representing the membrane fraction, was resuspended in ice-cold buffer (50 mM Tris-HCl, pH 7.4, 5 mM MgCl 2 , 1 mM EDTA, 100 mM NaCl, and protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 2 g/ml aprotinin, and 3 g/ml leupeptin, chymostatin, and pepstatin)) at concentrations of 3-14 mg/ml and then frozen at Ϫ80°C. A 100 M solution of the CAMQ peptide was prepared in methanol. Membranes from ste24⌬rce1⌬ yeast that expressed STE14 from a 2-m plasmid (SM3614/pSM1317) were prepared as a source of the prenylprotein methyltransferase (15). Each carboxyl-terminal processing reaction contained the a-factor peptide (0.75 M), membranes from mouse tissues (75 g/ml), ste24⌬rce1⌬/Ste14p yeast membranes (0.1 mg/ml) in a buffer containing 75 mM HEPES, 75 mM NaCl, 15 mM EDTA, pH 7.5, and 20 M S-adenosyl-L-methionine. The reactions were assembled in 96-well polystyrene plates in a total volume of 25 l. After a 2-h incubation at 30°C, the reactions were terminated by adding copper acetate (final concentration, 2 mM). The production of mature a-factor was detected by plating dilutions of the reaction mixture on a lawn of MAT␣ sst2 cells (SM1086) and allowing the cells to grow for 2 days at 30°C (27, 28). 3 Mature a-factor results in a zone of growth inhibition in the lawn of cells.
Assessing the Ability of Mouse Membranes to Carry Out the Aminoterminal Processing of P1 a-Factor-Yeast membranes containing radiolabeled P1 a-factor were prepared as previously described (10). P1 a-factor is fully COOH-terminal modified (i.e. isoprenylated, cleaved, and carboxyl methylated) (29). In a wild-type strain, COOH-terminal processing of a-factor occurs first and is followed by NH 2 -terminal processing (29). Isoprenylation is required (29) for the production of P1 a-factor, but blocking either methylation (30) or CAAX processing 4 does not impede NH 2 -terminal processing. Briefly, a yeast strain lacking STE24 but expressing MFA1 from a high-copy plasmid (SM3103/ pSM219) was radiolabeled with [ 35 S]cysteine, and membranes containing P1 a-factor were isolated (10). Proteolysis reactions (typically 50 l) were assembled on ice in polystyrene 96-well plates. Each reaction contained the metabolically labeled yeast membranes (0.5 mg/ml) and membranes (0.5 mg/ml) from Zmpste24ϩ/ϩ or Zmpste24Ϫ/Ϫ cells or tissues in a buffer containing 100 mM HEPES and 100 mM NaCl, pH 7.5. Some of the reaction mixtures contained 2 mM 1,10-phenanthroline, a zinc chelator. After a 2-h incubation at 30°C, reactions were terminated by adding 1% SDS and heating the samples to 100°C for 3 min. The a-factor intermediates were immunoprecipitated with an a-factor-specific antiserum. Each immunoprecipitate was then size-fractionated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and analyzed with a PhosphorImager (10,29).
We also incubated whole cell extracts from Rce1ϩ/ϩ and Rce1Ϫ/Ϫ fibroblasts in the presence of yeast membranes containing high levels of mouse Rce1 or yeast membranes containing high levels of mouse Zmp-ste24. All incubations contained 10 M S-adenosyl-L-[methyl- 14  base hydrolysis/methanol diffusion assay. All incubations were performed in duplicate.

RESULTS
A Mouse Zmpste24 cDNA-DNA sequencing of a mouse Zmpste24 cDNA revealed that the mouse enzyme, like the human enzyme, contains 475 amino acids (Fig. 1). The mouse and human amino acid sequences were 93% identical. Both the mouse and human sequences were 36% identical to the S. cerevisiae Ste24p sequence, with the highest levels of similarity within the last third of the molecule. Mouse Zmpste24, like the human and yeast enzymes, has a HEXXH motif (residues 335-339) that is characteristic of zinc metalloproteinases (5,8,9) and multiple membrane-spanning domains (Fig. 1). Mouse Zmpste24 is expressed in multiple tissues, with the highest levels of expression in the liver and kidney (Fig. 2).
Functional Studies of Mouse Zmpste24 in Yeast-To assess the functional integrity of mouse Zmpste24, a yeast plasmid encoding the Zmpste24 cDNA was transformed into ste24⌬rce1⌬ yeast. The transformed cells were tested for the ability to produce a-factor with an established bioassay that utilizes a lawn of MAT␣ sst2 cells. Nontransformed ste24⌬rce1⌬ yeast do not produce a zone of growth inhibition (halo) on the lawn of MAT␣ sst2 cells since they cannot form mature a-factor. The plasmid encoding mouse Zmpste24 restored a-factor production, as judged by the halo surrounding the colony (Fig. 3A). Mouse Rce1 also restored a-factor production (Fig. 3A). In anticipation of performing a gene-targeting experiment in mice, we sought to establish that the removal of the 104-bp exon 8 of Zmpste24 (encoding amino acids 318 -354, containing the HEXXH motif) would inactivate the gene product. We therefore transformed the ste24⌬rce1⌬ yeast with a mutant mouse Zmpste24 cDNA lacking exon 8 sequences. The mutant cDNA did not restore a-factor production (Fig. 3A).
Boyartchuk and co-workers (6) demonstrated that yeast Ste24p and Rce1p have distinct specificities for a-factor precursors with different CAAX sequences. Both yeast Ste24p and yeast Rce1p processed the wild-type a-factor precursor, which terminates in CVIA (5, 6). Ste24p was uniquely capable of processing mutant a-factor precursor molecules terminating in CAMQ, while Rce1p was uniquely capable of cleaving a mutant a-factor terminating in CTLM. Interestingly, we found that the mammalian enzymes exhibited the same specificities. Both mouse Rce1 and mouse Zmpste24 were capable of processing the wild-type a-factor terminating in CVIA (Fig. 3A). However, only RCE1 processed a mutant a-factor terminating in CTLM, and only Zmpste24 processed a mutant a-factor terminating in CAMQ (Fig. 3B).
The fact that both yeast Ste24p and mouse Zmpste24 can cleave a CAMQ a-factor was confirmed with coupled endoproteolysis/methylation assays (Fig. 3C). Membranes from ste24⌬rce1⌬ yeast that overexpressed yeast STE24 or mouse Zmpste24 cleaved and methylated a CAMQ a-factor peptide.

FIG. 2. Northern blot showing
Zmpste24 expression in multiple mouse tissues. A 32 P-labeled Zmpste24 cDNA probe was hybridized to a mouse Multiple Choice Northern blot (OriGene). Loading of samples was normalized to ␤-actin. The blot was exposed to x-ray film for 72 h at Ϫ80°C. A similar pattern was observed with a multiple tissue poly(A) ϩ RNA blot from CLONTECH (not shown).

FIG. 1. The mouse Zmpste24 amino acid sequence, aligned with the human and yeast sequences.
A conserved HEXXH zinc-binding motif is present in each of the three sequences (residues 335-339 of the mouse and human sequence). The mouse sequence is predicted to contain multiple transmembrane domains (residues 19 -37, 80 -102, 121-143, 171-189, 198 -216, 349 -367, and 386 -408). The yeast and human sequences also are predicted to contain multiple transmembrane domains (2). GenBank TM for mouse Zmpste24 cDNA sequence: AY029194.
Yeast membranes overexpressing the mutant mouse Zmpste24 (lacking the HEXXH motif) did not methylate the peptide, nor did yeast membranes that overexpressed yeast Rce1p or mouse Rce1. In contrast, membranes from yeast expressing mouse Zmp-ste24, yeast STE24, mouse Rce1, and yeast RCE1 cleaved and then methylated the wild-type CVIA a-factor peptide (Fig. 3D).
Zmpste24 Knockout Mice-A gene-targeting vector designed to replace exon 8 of Zmpste24 with a neo was used to inactivate mouse Zmpste24 (Fig. 4A). Approximately 1 in 30 drug-resistant clones was targeted (Fig. 4B). Two ES cell lines were used to generate Zmpste24ϩ/Ϫ mice, which were bred to generate Zmpste24Ϫ/Ϫ mice (Fig. 4C). Zmpste24Ϫ/Ϫ mice (produced from both targeted ES cell clones) were born at the expected mendelian frequency and were viable and fertile. To assess the relative capacities of wild-type cells and Zmpste24Ϫ/Ϫ cells to contribute to the formation of different mouse tissues, chimeric mice were generated with Zmpste24Ϫ/Ϫ ES cells (Fig. 4D). A total of 12 high-percentage (Ͼ90%) chimeric mice were obtained. As judged by Southern blot analysis, the Zmpste24deficient ES cells contributed robustly to the formation of each tissue that was analyzed (heart, lung, liver, spleen, kidney, adipose, brain, intestine, skin, testes, and skeletal muscle) (not shown).
Assessing Zmpste24 Enzymatic Activity Levels-We next sought to determine if the targeted mutation in Zmpste24Ϫ/Ϫ mice was associated with detectable changes in enzymatic activity levels. Since no natural substrates for mammalian Zmpste24 have been described, we tested the possibility that the membranes from Zmpste24Ϫ/Ϫ cells and tissues would be deficient in their capacity to process a surrogate substrate, yeast a-factor. In the first series of experiments, we incubated Zmpste24Ϫ/Ϫ and Zmpste24ϩ/ϩ membranes with the CAMQ a-factor peptide in the presence of S-adenosyl-L-[methyl-14 C]methionine, and then assessed base-labile methylation of the peptide. Similar levels of methylation were observed with both Zmpste24Ϫ/Ϫ and Zmpste24ϩ/ϩ membranes (Zmp-ste24ϩ/ϩ, 0.82 Ϯ 0.02 pmol/mg/min; Zmpste24Ϫ/Ϫ 0.78 Ϯ 0.02 pmol/mg/min), and the methylation was sensitive to 1,10-phenanthroline (data not shown). This result was clearly different from what we observed with the coupled endoprotease/methylation assays involving yeast membranes; Zmpste24-containing yeast membranes processed the CAMQ a-factor peptide significantly more than membranes lacking Zmpste24 (see Fig.  3C).
We suspected that the failure to find differences in methylation of the CAMQ a-factor peptide by Zmpste24Ϫ/Ϫ and Zmpste24ϩ/ϩ membranes was due to the presence of a membrane-bound, phenanthroline-sensitive exoprotease in the mouse membranes, which would be expected to generate farnesylcysteines or short farnesylcysteine-containing peptides. The latter substances would be readily methylated and would increase the background in the experiment, making it impossible to detect potential differences between the ability of Zmpste24Ϫ/Ϫ and Zmpste24ϩ/ϩ membranes to carry out the carboxyl-terminal endoproteolytic processing of a-factor. To circumvent this problem, we decided to perform a more specific assay, a yeast bioassay for fully processed mature a-factor. In this assay, we tested whether membranes from Zmpste24-deficient cells and tissues would be defective in producing mature a-factor from the 15-mer CAMQ a-factor peptide (i.e. defective in cleaving the -AMQ from the peptide, making it a substrate for carboxyl methylation). To address this issue, the CAMQ a-factor peptide was mixed with mouse liver membranes for 2 h, and the production of a-factor was assessed by assessing the ability of mature a-factor to inhibit the growth of MAT␣ sst2 yeast. Notably, membranes from livers of Zmpste24Ϫ/Ϫ mice were 4 -8 times less efficient in producing mature a-factor (Fig. 5A). Similar results were observed in other tissues in an independent experiment (Fig. 5B). FIG. 3. Yeast halo assays demonstrating that mouse Zmpste24 complements the defect in a-factor production in ste24⌬rce1⌬ yeast. A, halo assay demonstrating that the expression of yeast RCE1 (from plasmid pMB6), mouse Rce1 (pMB7), yeast STE24 (pMB5), and mouse Zmpste24 (pMB4) restores the production of a-factor in ste24⌬rce1⌬ yeast (JRY5317 ste24⌬::HIS3 rce1⌬::TRP1). A mutant mouse Zmpste24 construct lacking exon 8 sequences (which encode the HEXXH motif) did not restore a-factor production. B, halo assay demonstrating that yeast expressing mouse Zmpste24 process a mutant a-factor terminating in CAMQ (halo marked by arrows) but not a mutant terminating in CTLM (absent halo). Yeast expressing mouse Rce1 process a mutant a-factor terminating in CTLM (halo marked by arrows) but not a mutant terminating in CAMQ (absent halo). The halos in Panel B are smaller than in Panel A because detergents were not included in the agar in Panel B. C, coupled endoproteolysis/methylation assay demonstrating that the CAMQ a-factor peptide is cleaved and methylated by membranes from ste24⌬rce1⌬ yeast (JRY5317) that overexpress mouse Zmpste24 or yeast STE24 but not by membranes that overexpress yeast RCE1, mouse Rce1, or the mutant mouse Zmpste24. D, coupled endoproteolysis/methylation assay demonstrating that the CVIA a-factor peptide is cleaved and methylated by membranes from ste24⌬rce1⌬ yeast that express mouse Zmpste24, yeast STE24, mouse Rce1, and yeast RCE1, but not by the mutant mouse Zmpste24.
We also examined the possibility that membranes from Zmpste24Ϫ/Ϫ cells and tissues would be deficient in their ability to cleave the amino terminus of an a-factor intermediate. In yeast, both in vivo and in vitro, Ste24p cleaves the seven amino-terminal amino acids from P1 (farnesylated a-factor precursor), generating a shorter intermediate, P2, which can be resolved from P1 by SDS-PAGE (9, 10, 29). We produced a metabolically labeled P1 intermediate in ste24⌬ yeast, then tested the ability of Zmpste24Ϫ/Ϫ and Zmpste24ϩ/ϩ membranes to cleave the amino terminus from P1, converting it to P2 (Fig. 6). Membranes from Zmpste24ϩ/ϩ ES cells and fibroblasts converted P1 to P2, and this activity was blocked by 1,10-phenanthroline (Fig. 6A). Membranes from Zmpste24Ϫ/Ϫ ES cells and fibroblasts lacked this activity (Fig. 6A). Similarly, membranes from the liver, heart, and skeletal muscle of Zmpste24ϩ/ϩ mice converted P1 to P2, whereas membranes from Zmpste24Ϫ/Ϫ mice did not (Fig. 6B).
Testing for the Accumulation of Substrates in Rce1Ϫ/Ϫ and Zmpste24Ϫ/Ϫ Cells-Although no natural Zmpste24 CAAX protein substrates have been identified, the high degree of sequence conservation in the enzyme suggests that natural substrates must exist. We hypothesized that significant amounts of CAAX protein substrates might accumulate in both Rce1-deficient cells and Zmpste24-deficient cells. To test that possibility, whole cell extracts from primary fibroblasts (Zmpste24ϩ/ϩ, Zmpste24Ϫ/Ϫ, Rce1ϩ/ϩ, and Rce1Ϫ/Ϫ) were incubated with S-adenosyl-L-[methyl-14 C]methionine and membranes containing high levels of Ste14p, mouse Rce1, and mouse Zmpste24. As a control, we examined the accumulation of protein substrates in whole cell extracts from ICMTϩ/ϩ and ICMTϪ/Ϫ fibroblasts (13). The accumulation of methylatable substrates in each extract was gauged with a base hydrolysis/ methanol diffusion assay. As expected, methylatable substrates accumulated in ICMTϪ/Ϫ cells (Fig. 7A). Substrates also accumulated in Rce1Ϫ/Ϫ cells (Fig. 7A). However, no such accumulation was detectable in Zmpste24Ϫ/Ϫ cells (Fig. 7A). Similarly, no accumulation of methylatable substrates was detected in extracts of Zmpste24Ϫ/Ϫ livers, kidneys, or hearts (not shown).
In a separate experiment, we asked whether Zmpste24 would be capable of processing some or all of the substrates that accumulate in Rce1-deficient cells. Whole cell extracts from Rce1ϩ/ϩ and Rce1Ϫ/Ϫ fibroblasts were incubated with Ste14p, S-adenosyl-L-[methyl-14 C]methionine, and either mouse Rce1 or mouse Zmpste24. Again, we noted increased methylation when Rce1 was added to extracts from Rce1-deficient cells, reflecting the accumulation of Rce1 substrates in those cells (Fig. 7B). No such increase was observed when Zmpste24 was added to the Rce1-deficient cell extracts (Fig.  7B), indicating that Zmpste24 has little ability to process Rce1 substrates. DISCUSSION We have sought to define the physiologic importance of the two endoproteases and the methyltransferase involved in the "post-isoprenylation" processing of proteins containing a CAAX motif (2). We have shown that ICMT and Rce1 are essential for the processing of the Ras proteins and that both are required for embryonic development. The involvement of Rce1 and ICMT in Ras protein processing was not particularly surprising, given that the corresponding yeast genes, RCE1 and FIG. 5. Reduced carboxyl-terminal processing of a CAMQ afactor peptide by membranes prepared from livers of Zmpste24؊/؊ mice. A, a CAMQ a-factor peptide was mixed with Zmpste24Ϫ/Ϫ or Zmpste24ϩ/ϩ liver membranes (75 g/ml) and incubated for 2 h at 30°C. Incubations were performed in the presence and absence of 1,10-phenanthroline. Serial dilutions of the samples were plated on a lawn of MAT␣ sst2 yeast cells (SM1086); the production of mature a-factor was judged by inhibition of yeast growth. There was some residual activity in Zmpste24Ϫ/Ϫ membranes, which could be largely blocked with 1,10-phenanthroline. The residual activity was probably not due to Rce1 because Rce1 activity is not affected significantly by 1,10-phenanthroline (33) and because Rce1 cannot process a CAMQ a-factor peptide. B, independent experiment testing the capacity of membranes from different Zmpste24Ϫ/Ϫ or Zmpste24ϩ/ϩ tissues to generate mature a-factor from the same CAMQ a-factor. The Zmpste24Ϫ/Ϫ cells were generated by growing Zmpste24ϩ/Ϫ in high concentrations of G418, as described by Mortensen et al. (26). The genomic DNA was digested with EcoRI, and the blot was hybridized with the 5Ј-flanking probe. STE14, had already been shown to be involved in the processing of yeast Ras2p (4,5,31,32). The fact that knockouts of the two mouse genes yielded lethal developmental phenotypes was also not too surprising given that both are involved in the processing of dozens of isoprenylated proteins. In the current studies, we focused on mouse Zmpste24, the "other" CAAX endoprotease and arguably the most mysterious and intriguing of the three post-isoprenylation protein-processing enzymes.
In yeast, the only known substrate for Ste24p is the mating pheromone a-factor (2, 5, 6, 8 -10, 29). Although no a-factor orthologue has yet been identified in mammals, zinc metalloproteases similar to Ste24p can be readily identified in many species throughout the plant and animal kingdoms (2,9). Interestingly, the level of amino acid sequence identity is higher for yeast Ste24p and mouse Zmpste24 than for yeast Rce1p/ mouse Rce1 or for yeast Ste14p and mouse ICMT (2). The high level of sequence similarity initially led us to suspect that Zmpste24 might also encode a key housekeeping protein required for mouse survival. This suspicion was not borne out. Zmpste24-deficient mice developed normally and were fertile. The absence of a lethal phenotype cannot be attributed to a leaky phenotype or a poorly designed knockout experiment. Our gene-targeting vector eliminated the critical HEXXH zincbinding domain and introduced a frameshift, and we had proven with yeast expression studies, even before embarking on the mouse experiments, that the mutation inactivated the enzyme.
In assessing the biochemical phenotype associated with mouse Zmpste24 deficiency, we used yeast a-factor as a surrogate substrate, since no natural mammalian substrates for Zmpste24 have been identified. Nonetheless, our biochemical studies uncovered new and intriguing findings. First, we found that an enzymatic activity capable of cleaving the amino terminus of a-factor is expressed at readily detectable levels in wild-type mouse ES cells and in multiple tissues of wild-type mice. This enzymatic activity was absent in the setting of Zmpste24 deficiency. Thus, Zmpste24 was the only mammalian enzyme capable of processing the amino terminus of a-factor in vitro; no redundant "Ste24p/Zmpste24-like" amino-terminal cleavage activities were detected. We also demonstrated that Zmpste24-deficient cells are deficient in a specific CAAX endoprotease activity, the ability to cleave the -AAX from a CAMQ a-factor peptide. The use of a CAMQ a-factor peptide was essential for these experiments because it cannot be cleaved by Rce1 (5, 6).
One possible conclusion from the phenotype of the Zmpste24Ϫ/Ϫ mice is that there are simply no essential Zmpste24 substrates. Alternatively, developmentally essential Zmpste24 substrates may exist but are processed, at least to a degree, by other proteases. In this regard, it is worthwhile pointing out that STE24-deficient yeast are not completely deficient in their ability to produce mature a-factor (2,5,6,9). Some mature a-factor is produced because the carboxyl terminus of a-factor precursor can be cleaved by Rce1p and because the Axl1pmediated amino-terminal cleavage step proceeds, albeit inefficiently, in the absence of Ste24p. Thus, it is quite conceivable that the viability of Zmpste24Ϫ/Ϫ mice results from processing of Zmpste24 substrates by other enzymes.
The only known Ste24p substrate in yeast, a-factor, is meth- 6. Reduced ability of membranes from Zmpste24؊/؊ cells and tissues to carry out the amino-terminal processing of a-factor. An 35 S-labeled P1 a-factor intermediate was prepared from a yeast strain that lacked STE24 and expressed MFA1 from a high-copy plasmid. Proteolysis reactions were performed by incubating the radiolabeled P1 with mouse membranes for 2 h at 30°C in the absence or presence of 1,10-phenanthroline. The a-factor intermediates were then immunoprecipitated with an a-factor-specific antiserum and size-fractionated by SDS-PAGE. Dried gels were analyzed by autoradiography and with a PhosphorImager. The amino-terminal processing reaction cleaves seven amino acids from the amino terminus of P1, yielding the shorter intermediate P2 (9,29). A, ability of membranes from ES cells and fibroblasts to cleave the P1 a-factor intermediate. B, ability of membranes from Zmpste24-deficient mouse tissues to cleave the P1 a-factor intermediate.

FIG. 7.
Absence of a detectable accumulation of "methylatable" substrates in Zmpste24؊/؊ cells. A, methylation of whole cell extracts from primary fibroblasts derived from Zmpste24ϩ/ϩ, Zmpste24Ϫ/Ϫ, ICMTϩ/ϩ, ICMTϪ/Ϫ, Rce1ϩ/ϩ, and Rce1Ϫ/Ϫ embryos. The cell extracts were incubated with S-adenosyl-L-[methyl-14 C]methionine, membranes containing high levels of Ste14p, yeast membranes containing high levels of mouse Rce1, and yeast membranes containing high levels of mouse Zmpste24. The relative level of methylatable substrates in wild-type and knockout cells was assessed with a base hydrolysis/methanol diffusion assay. For the Zmpste24ϩ/ϩ and Zmp-ste24Ϫ/Ϫ cells, we observed identical results when membranes expressing Rce1 were left out of the reaction mixture. B, methylation of whole cell extracts from Rce1ϩ/ϩ and Rce1Ϫ/Ϫ fibroblasts in the presence of yeast membranes containing high levels of mouse Rce1 or membranes containing high levels of mouse Zmpste24. In these experiments, we demonstrated that the yeast membranes contained active Zmpste24 by showing that they promoted the processing and methylation of the 15-mer CAMQ a-factor substrate, as shown in Fig. 3D. ylated after the proteolytic release of the carboxyl-terminal tripeptide (4). If unique Zmpste24 CAAX protein substrates were to exist in mammalian cells (i.e. substrates not cleaved by Rce1), we predicted that we would be able to document an accumulation of methylatable substrates in Zmpste24-deficient cells. This prediction was not upheld. No accumulation of methylatable substrates was detected either in Zmpste24Ϫ/Ϫ ES cells or in a variety of Zmpste24Ϫ/Ϫ tissues. In contrast, an accumulation of methylatable substrates was easily detectable in both Rce1-deficient cells and ICMT-deficient cells. Recombinant Zmpste24 could not cleave the accumulated substrates in Rce1 knockout cells. How should these studies be interpreted? One could properly conclude that these studies show that a greater number of methylatable protein substrates accumulate in the setting of Rce1 or ICMT deficiency than in Zmpste24 deficiency. One could also conclude that Zmpste24 cannot process the uncleaved substrates in Rce1-deficient cells. However, we would be reluctant to conclude that mammalian cells have no unique Zmpste24 CAAX protein substrates or that all Zmpste24 substrates are cleaved by other enzymes (e.g. Rce1). Unique Zmpste24 substrates may accumulate in Zmpste24 deficiency but at concentrations too low to yield levels of carboxyl methylation above background.
The conservation in the enzymatic specificities for yeast Ste24p and mouse Zmpste24 is remarkable. Mouse Zmpste24 as well as the human enzyme faithfully cleave the amino terminus of yeast a-factor. Moreover, both mouse Zmpste24 and yeast Ste24p share the property of being able to cleave the carboxyl terminus of a wild-type a-factor and a CAMQ a-factor mutant, but not a CTLM a-factor mutant. The striking conservation of Ste24p/Zmpste24 specificities makes it tempting to speculate that an a-factor orthologue is lurking in the mammalian genome. We have not yet identified such a protein, but recognizing an open reading frame as short as 36 amino acids (the length of the a-factor precursor) could be quite challenging. This challenge is compounded when one considers the fact that a-factor-like CAAX proteins from diverse organisms (e.g. Saccharomyces cerevisiae, Schizosaccharomyces pombe, Ustilago maydis, Ustilago hordei, Filobasidiella neoformans) exhibit only limited sequence similarities.