A new paradigm for Prelamin A proteolytic processing by ZMPSTE24: the upstream SY^LL cleavage occurs first and there is no CaaX processing by ZMPSTE24

Human ZMPSTE24, an integral membrane zinc metalloprotease, is required for conversion of prelamin A to mature lamin A, a component of the nuclear lamina and failure of this processing causes premature ageing disorders. ZMPSTE24 has also been implicated in both type 2 diabetes mellitus and in viral-host response mechanisms, but to date its only confirmed substrate is the precursor for lamin A. Prelamin A is thought to undergo four C-terminal post-translational modifications in the following order: farnesylation, SIM tripeptide cleavage, carboxymethylation and upstream “SY^LL” cleavage. Here we present evidence that the sequence of events does not follow the accepted dogma. We assessed cleavage of long human prelamin A sequence peptides by purified human ZMPSTE24 combined with FRET and mass spectrometry to detect products. Surprisingly, we found that the “SY^LL” cleavage occurs before and independent of the C-terminal CSIM modifications. We also found that ZMPSTE24 does not perform the predicted C^SIM tripeptide cleavage, but rather it removes an IM dipeptide. ZMPSTE24 can perform a tripeptide cleavage with a canonical CaaX box (C: cysteine; a: aliphatic; X: any residue), but the C-terminus of prelamin A is not a true CaaX sequence. Regardless of the C-terminal modifications of prelamin A, ZMPSTE24 can perform upstream SY^LL cleavage, thus removing the unwanted farnesylated C-terminus. Therefore, it is failure of SY^LL cleavage, not the C-terminal processing that is the likely cause of progeroid disorders.


Introduction
ZMPSTE24 (also called farnesylated protein-converting enzyme 1, FACE1), is an integral membrane zinc metalloprotease that is required for conversion of prelamin A to mature lamin A, a major protein component of the nuclear intermediate filaments known as lamina [1,2]. ZMPSTE24 has also been implicated in unclogging the translocon in the endoplasmic reticulum [3], removing human islet amyloid polypeptide (IAPP) from the pancreas in patients with type 2 diabetes mellitus [4], and host defenses against viruses [5,6].
Despite suggestions that ZMPSTE24 could be involved in diverse functions in cell biology, thus far the only confirmed substrate for ZMPSTE24 is prelamin A [7]. Lamins A, B1, B2, and C form the nuclear lamina, an intermediate filament meshwork that lines the nuclear side of the inner nuclear membrane [8,9]. The nuclear lamin proteins interact with genomic DNA, the nuclear membrane and nucleoplasmic proteins, and these interactions are thought to be important for an array of biological processes [1,[7][8][9].
Failure of ZMPSTE24-mediated prelamin A processing, due either to missense mutations in ZMPSTE24 or prelamin A defects that interfere with ZMPSTE24-mediated processing, result in progeria-like disease phenotypes due to an accumulation of farnesylated prelamin A [10]. Hutchinson-Gilford progeria syndrome (HGPS), mandibuloacral dysplasia type B (MAD-B), and restrictive dermopathy (RD) all involve mutations that interfere with the conversion of farnesylated prelamin A to mature lamin A [11][12][13][14][15][16][17][18][19][20][21]. ZMPSTE24-deficient mice exhibit a striking accumulation of farnesylated prelamin A and exhibit disease phenotypes that resemble those in the aforementioned progeroid disorders, including severe growth retardation, nonhealing bone fractures, alopecia, and progressive inanition [1,2]. Studies of fibroblasts from affected patients have suggested that the severity of disease phenotypes in patients is proportional to the accumulation of farnesylated prelamin A proteins in cells [22][23][24][25]. There is some evidence to suggest that defective prelamin A processing could be relevant to physiological aging, at least in certain tissues [26,27]. Given the relevance of prelamin A processing for health and disease, we believe that it is important to have a clear understanding of the biochemistry of prelamin A processing. 4 ZMPSTE24 is required for posttranslational processing at the C-terminus of prelamin A, a 661residue protein that includes a C-terminal CSIM sequence. The textbook view of prelamin A processing is that prelamin A undergoes a series of four sequential posttranslational modifications, including (i) farnesylation of the cysteine by protein farnesyltransferase, (ii) removal of the C-terminal -SIM tripeptide by either RCE1 or ZMPSTE24, (iii) carboxymethylation of the newly exposed farnesylcysteine by ICMT, and (iv) a final endoproteolytic processing step at an SY^LL site located 15 residues upstream from the C-terminal farnesylcysteine methyl ester (S1 Fig). The first three proposed steps are identical to those that occur in a large number of proteins with a C-terminal CaaX motif (where C is cysteine, a is an aliphatic residue and X is any residue) [28], including the Ras GTPases [29]. All such proteins undergo protein farnesylation, release of the C-terminal tripeptide by RCE1, and carboxymethylation of the farnesylcysteine by ICMT, enzymatic reactions collectively known as the CaaX processing reactions. However, no CaaX protein, apart from prelamin A, undergoes an additional C-terminal cleavage reaction, so they all retain their C-terminal farnesylcysteine methyl ester.
Unlike most proteins harboring a classic CaaX motif, prelamin A undergoes a second endoproteolytic cleavage step, catalyzed by ZMPSTE24, at a SY^LL motif located 18 residues from the C-terminus. This cleavage step removes the farnesylcysteine (fCys) from the protein and releases mature lamin A (S1 Fig) [7,30]. It is generally accepted that prelamin A must undergo all of the CaaX processing steps before the SY^LL cleavage can occur, although two reports raised the possibility that the SY^LL cleavage could occur in the absence of CaaX processing [31,32].
Many of the studies on the function of the human ZMPSTE24 have been based on comparisons with the yeast orthologue of ZMPSTE24, Ste24, the only other protein thought to perform both C^aaX and upstream proteolysis steps [33]. ZMPSTE24 and Ste24 share an unusual protein fold [34,35], with seven transmembrane α-helices forming an α-helical barrel enclosing a large, water-filled chamber inside the membrane. The zinc metalloprotease domain lies on the nucleoplasmic or cytoplasmic side of membrane, with the active site facing into the chamber.
Not only are the structures of ZMPSTE24 and Ste24 similar, they also appeared to have very similar proteolytic activities. In yeast Ste24 is required for the biogenesis of the mating pheromone a-factor (38). A-factor undergoes the classic CaaX processing steps (farnesylation of the C-terminal cysteine, 5 C^aaX tripeptide cleavage, carboxymethylation of the farnesylcysteine) followed by two upstream cleavage reactions. Ste24 and Rce1 can both catalyze the removal of the tripeptide from yeast a factor's classic CaaX motif (CVIA). Then Ste24 catalyses the first of two upstream cleavage reactions, the second being catalysed by Axl1 [33]. Human ZMPSTE24 complements the defect in a-factor production in yeast lacking both Rce1 and Ste24, indicating that ZMPSTE24 is capable of cleaving both the C^aaX site in a-factor and the upstream cleavage reaction [7,33], Although some evidence suggests CaaX cleavage may not be a prerequisite for second Ste24-mediatd cleavage of the a-factor precursor [36], it is generally thought that the a-factor C^aaX cleavage occurs before the upstream cleavage and that the C^aaX cleavage is necessary for the upstream cleavage to occur.
Given these observations in the yeast system, it seemed reasonable to assume that ZMPSTE24 would perform both endoproteolytic processing steps on its natural substrate, prelamin A. Thus far, however, there has been no direct experimental evidence to confirm that ZMPSTE24 does cleave the CSIM sequence of prelamin A at the predicted C^aaX site. In fact, when we examined ZMPSTE24-mediated processing of a short farnesylated prelamin A peptide (QSPQNC(f)SIM) with a mass spectrometrybased assay, we found release of the dipeptide IM, raising the possibility of a CS^IM cleavage rather than a C^SIM cleavage. Moreover, the complex that we observed between ZMPSTE24 and a CSIM peptide in the crystal structure suggested that the peptide might undergo a CS^IM cleavage. More studies were needed to explore these preliminary observations based on short peptides.
The biochemistry of the SY^LL cleavage in prelamin A is also relatively unexplored. There is little data available on the kinetics of proteolysis of prelamin A based substrates by ZMPSTE24 or comparison of the SY^LL cleavage, which is unique to ZMPSTE24, and the CSIM cleavage, which can also be performed by RCE1. Available studies followed the appearance of the final lamin A product and/or whether it was possible to carboxymethylate the C-terminus [28,35,[37][38][39][40][41][42][43]. Studies involving purified ZMPSTE24 characterized the C^aaX or upstream reaction using a short peptide based on the C-terminal sequence of human K-RAS protein [44] or the yeast a-factor peptide [43] rather than the actual prelamin A sequence. Another study with purified ZMPSTE24 used in native mass spectrometry to investigate complexes of ZMPSTE24 with peptide substrates and products with the SY^LL site but not the CSIM site [45]. 6 In this study, our goal was to provide a biochemical characterization of prelamin A processing by ZMPSTE24, including kinetics for each cleavage reaction (CSIM and SYLL) and identification of the products formed by each processing step. We used peptides containing only one of the two sites, as well as a longer prelamin A peptides containing both cleavage sites. Our results confirmed the dipeptide cleavage from the CSIM site. More importantly, we investigated the relationship between the SYLL and CSIM cleavage reactions and discovered that the processing of prelamin A by ZMPSTE24 does not conform to textbook dogma, which holds that the upstream cleavage reaction at the SYLL site can only occur after the processing at the CSIM motif is complete. Here, we show that ZMPSTE24 performs the upstream SY^LL cleavage reaction in prelamin A, regardless of any post-prenylation processing at the CSIM motif.

ZMPSTE24 reaction kinetics suggests faster cleavage at the SY^LL than CSIM site
ZMPSTE24 cleaves prelamin A at two sites (SYLL and CSIM). To measure the kinetics of ZMPSTE24-mediated cleavage reactions, we designed a fluorescence resonance energy transfer (FRET) assay using two farnesylated prelamin A peptides, a 20-mer peptide covering the SYLL site (Peptide 1, Table 1, Fig 1A) and a 9-mer peptide covering the C-terminal CSIM site (Peptide 2, Table   1, Fig 1A). The peptides were modified with a dinitrophenyl (Dnp) and n-amino benzoic acid (Abz) groups on either side of the cleavage site. In the absence of peptide cleavage, the Dnp group quenches the fluorescent signal from the Abz group ( Fig 1B). Cleavage of the peptide separates the fluorophore and quencher, resulting in an increase in the fluorescence signal.   We produced human ZMPSTE24 in both Sf9 insect cells and in human Expi293F™ cells and compared the activity of protein produced in these two systems. We found that the activities of ZMPSTE24 obtained from different expression systems was very similar (S2E and S2F Fig), confirming that the results we obtained were not an artefact of using Sf9 cells for expression. The conversion factor for the relative fluorescence intensity (RFI) and moles of the peptide cleavage product was determined with a standard curve (Fig 1C). The fitted line plot (R 2 = 0.992) revealed that the linear regression followed the experimental data almost exactly, indicating that intermolecular quenching of the fluorescent signal was negligible.
Next, we compared kinetic parameters for the two cleavage reactions. Initial velocities for peptide 1 cleavage at the SYLL site and the peptide 2 cleavage at the CSIM site were plotted against substrate concentrations and fitted to the Michaelis-Menten equation (Fig 1D and 1E). The kinetic parameters (Km, Vmax and kcat) are shown in Fig 1F. The substrate affinity was in the low micromolar range for both reactions but was slightly higher affinity with the CSIM reaction. On the other hand, the turnover of the SYLL peptide was nearly five-fold higher than the CSIM peptide, indicating that the SYLL cleavage is faster.

SYLL peptide cleavage occurs at the expected location, but the CSIM peptide cleavage gives a dipeptide product
We used electrospray ionization mass spectrometry to characterize the peptide cleavage products of the reactions described above (Fig 2 and 3). For peptide 1, which spans the SYLL site, two products were 9 detected: an N-terminal 5-mer peptide (Peptide 1-1) and a C-terminal 15-mer peptide (Peptide 1-2), which eluted at 1.8 and 6.3 min, respectively ( Fig 2C). The identity of each peptide was confirmed by mass spectrometry (Fig 2D and 2E). As predicted, the cleavage occurred between Y and L. After a 30min incubation with ZMPSTE24 at 37°C, the substrate peptide (Peptide 1) was almost completely consumed, with only a small peak remaining at 1356.65 remaining (6.6 min peak in Fig 2C and 2F).
We also observed a dimer of the detergent DDM, which co-eluted with peptides 1 and 1-2. This DDM species has been observed previously in mass spectrometry-based analyses of membrane proteins [46].  and absence (blue) of ZMPSTE24. (D-F) Mass spectrometric spectrum for peptide 1 and the reaction products at the indicated time points in the presence (red) or absence (blue) of ZMPSTE24. The retention time of the peptide substrate was 7.6 min; the retention times for the two reaction products were 3.3 and 6.5 min (C-F). The substrate and reaction products are marked with coloured dots; peptides that were not detected are marked with white dots. DDM was detected in the product sample at 6.5 min.
This experiment was repeated four times, with similar results each time.
To investigate the CSIM cleavage, we analyzed the products from treatment of the farnsylated Cterminal 8-mer peptide (peptide 2) with ZMPSTE24 (Fig 3). We detected two products: peptides 2-1 and 2-2 ( Fig 3C-F). Interestingly, neither of those peptides corresponded to the reactions products expected from a classic C^aaX cleavage. Human ZMPSTE24 is thought to function as a prelamin A CaaX protease because the yeast orthologue, Ste24, serves as a CaaX protease, removing a tripeptide from the CaaX motif (CVIA) of yeast a-factor [33,47]. Therefore, we expected that human ZMPSTE24 would cleave peptide 2 between the farnesylcysteine and adjacent serine within the C^SIM motif (a reaction that would release peptides 2-3 and 2-4). However, the products we identified by mass spectrometry were peptides 2-1 and 2-2 ( Fig 3C-F), indicating that the cleavage actually occurs between the serine and the isoleucine of the CSIM motif (CS^IM). These results are consistent with our previous observation that with a nonderivatized farnesylated peptide (QSPQNC(f)SIM), ZMPSTE24 removed an IM dipeptide [34]. This finding is also consistent with our structure of a CSIM tetrapeptide in complex with ZMPSTE24, where the tetrapeptide appeared to be positioned (with respect to the Zn 2+ ion) for a CS^IM cleavage reaction [34]. In the case of peptide 2, most of the substrate was uncleaved after a 30-min incubation with ZMPSTE24 (Fig 3F), indicating that the CS^IM cleavage reaction was far slower than the SY^LL reaction.
To determine if there were any reaction products consistent with a C^SIM cleavage, we extracted in the other two, peaks for peptides 2-3 and 2-4 were undetectable. Therefore, our results suggest that although ZMPSTE24-mediated cleavage of Peptide 2 releases a tripeptide but only in very small amounts; the primary reaction is a CS^IM cleavage, which yields an IM dipeptide product.
It has been reported that after cleavage of a farnesylated CSIM containing peptide by ZMPSTE24, one of the products could be carboxymethylated by ICMT [37]. Since carboxymethylation requires an exposed farnesylcysteine, the product of a tripeptide C^aaX cleavage, this finding suggested that some C^aaX peptide processing had occurred. However, the level of carboxymethylation in those studies was low [37], consistent with the very low levels of the C^aaX cleavage in our mass spectrometry studies. This method would not detect the products of a CS^IM cleavage, since an exposed C(f)S would not be a substrate for ICMT, so it does not reveal how much dipeptide release occurred.
Our studies suggest that a CS^IM dipeptide cleavage is the preferred ZMPSTE24 reaction [34], but those experiments were performed with a short (9-mer) peptide. To exclude the possibility that the use of a short peptide lead to abnormal cleavage, we repeated the FRET based assay with a derivatised Although there was a small peak for the 15-mer product (Peptide 3-2), it was only 1% of the size of the

ZMPSTE24 performs the SY^LL cleavage before the CS^IM cleavage
Our studies with FRET-labelled peptides revealed that the SY^LL cleavage was fivefold faster than the CSIM cleavage and the CSIM cleavage involved the release of a dipeptide rather than a tripeptide.
However, these studies involved peptide substrates modified by FRET labels. To exclude the possibility that the labels affected the cleavage process, we investigated ZMPSTE24 activity against a native 29mer farnesylated prelamin A peptide containing both cleavage sites (Peptide 4). We used mass spectrometry to assess peptide product formation at multiple time points (Fig 4). If the CS^IM cleavage were to occur before the SY^LL cleavage (as generally assumed), we would expect Peptide 4-5 (or Peptide 4-6) to appear in the reaction mixture first, followed later by the appearance of Peptides 4-1 and 4-4. However, that was not the case. Contrary to expectations, the SY^LL cleavage occurred first, releasing Peptides 4-1 and 4-2 ( Fig 4A, 4B and 4D). Peptide 4-1 was formed continuously and accumulated progressively (Fig 4B), whereas the amount of peptide 4-2 (the C-terminal product of the SY^LL cleavage) increased briefly but then remained constant (Fig 4D), implying the existence of an equilibrium between production and consumption of the peptide. If there were an initial SY^LL cleavage followed by a CS^IM dipeptide 13 cleavage, peptide 4-3 would be formed. Indeed, peptide 4-3 did appear (Fig 4C) and the amount of that peptide increased from 30 to 60 min. Furthermore, mass spectra of the 1-min sample revealed a greater abundance of the product of an initial SY^LL cleavage reaction (peptide 4-2) than the product resulting from both cleavage reactions (peptide 4-3) (S6A and S6B Fig). We did not detect peptides 4-4 or 4-5 (products that would have formed from a C^aaX tripeptide cleavage), nor did we detect peptide 4-6 (which would have indicated that the CS^IM cleavage occurred before the SY^LL cleavage). Thus, our data indicates that the first reaction is the SY^LL cleavage (yielding Peptide 4-2), which subsequently undergoes a CS^IM cleavage (yielding Peptide 4-3).
We considered the formal possibility that the initial SY^LL cleavage was a peculiarity of terminal product of the SY^LL cleavage (Peptide 5-1) (Fig 5C). Both doubly-and triply-charged states were detected (with a sodium adduct in the latter) (Fig 5D). If ZMPSTE24 were to cleave within the CVLS site, we would expect to find either peptides 5-2 and/or 5-3, depending on whether a tripeptide or a dipeptide was released. Although the m/z value for the doubly-charged Peptide 5-2 were close to that of the triply-charged Peptide 5, it was possible to resolve the peptides on the extracted ion chromatogram (Fig 5E-5G). Peptide 5-2 was identified, indicating that the tripeptide C^aaX cleavage was favoured (Fig 5F). The peak for peptide 5-2 appeared only after 5 min, whereas the peak for Peptide 5-1 could be detected at 1 min (Fig 5E), indicating that the SY^LL cleavage preceded the C^VLS cleavage. Moreover, the peak for peptide 5-2 was small, reflecting a lower efficiency for the C^VLS reaction than the SY^LL reaction. Thus, a prelamin A peptide terminating with a CVLS motif first undergoes the SY^LL cleavage and subsequently a C^aaX cleavage. These findings suggest that ZMPSTE24 is capable of acting as a bona fide C^aaX protease but not with a substrate terminating in CSIM. Substrate and product peptides are marked with coloured dots; peptides that were not detected are marked with white dots. Product peptide 5-1 (peach dot) has a retention time of 6.1 min and was detected in doubly-and triply-charged forms with a Na + adduct (C-D); product peptide 5-2 (yellow dot) has a retention time of 5.6 min and was detected in its doubly-charged form with a Na + adduct (E and 15 F); the peak at 5.8 min was substrate peptide 5 (orange dot) in its doubly-and triply-charged forms (E,G). The mass spectrometry measurements were performed in three biological replicates.

The farnesyl lipid modification influences the order in which the cleavage reactions occur
To investigate the importance of the farnesyl lipid on ZMPSTE24-mediated prelamin A cleavage, we performed time-course mass spectrometry studies with a nonfarnesylated 29-mer prelamin A peptide (Fig 6). The first processing event was a SY^LL cleavage (Peptides 6-1 and 6-2), but we also observed peptide products resulting from a CS^IM cleavage (Peptides 6-3 and 6-6). We found no evidence for a C^SIM cleavage (Peptides 6-4 and 6-5). The peak corresponding to Peptide 6-6 appeared at approximately the same rate as the SY^LL cleavage products, suggesting that CS^IM cleavage and SY^LL reactions occur simultaneously in the absence of the farnesyl lipid modification, unlike the faster reaction we observed at SY^LL when farnesyl is present (Fig 3). 16

Assessing the SY^LL cleavage in vivo with a yeast model
To examine ZMPSTE24-dependent cleavage of prelamin A in vivo, we expressed human ZMPSTE24 and prelamin A in yeast strains lacking yeast Ste24 and we observed SY^LL cleavage by ZMPSTE24 (i.e., endoproteolytic release of mature lamin A) (Fig 7A), as we previously reported [48].
Notably, we show here that the SY^LL cleavage occurred regardless of whether the prelamin A terminated in CSIM or CVLS, or other CaaX motifs, and regardless of whether Rce1 was present or absent. We used a yeast halo assay to determine if Ste24 is capable of producing mature a-factor from a-factor constructs terminating in CSIM and CVLS [31,47]. We observed production of mature bioactive a-factor from both CSIM and CVLA constructs in WT and ste24∆ yeast (Fig 7B). The production of some a-factor in ste24∆ yeast, albeit a reduced amount, is likely due to the ability of Axl1 to carry out the final N-terminal processing step in the absence of the Ste24-dependent N-terminal processing step [36]. Notably, in rce1Δ yeast, only a-factor constructs ending with a bona fide CaaX motif (CVIA, CVLS) yielded mature a-factor; constructs terminating with CTLM and CSIM (sequence motifs in which the cysteine is followed by a polar residue) did not yield mature a-factor (Fig 7B). This observation is likely explained by the fact that a-factor secretion and activity require carboxymethylation of a C-terminal farnesylcysteine (which is only possible after a tripeptide cleavage) [36]. The CSIM sequence likely undergoes dipeptide cleavage, thereby preventing carboxymethylation and the production of mature a-factor. Our findings suggest that both ZMPSTE24 and Ste24 carry out a tripeptide cleavage when proteins terminate with a bona fide CaaX motif but not when proteins terminate with a modified C-terminal sequence motif where the cysteine is followed by a polar residue (e.g., Thr, Ser).

Discussion
Although the requirement for ZMPSTE24 activity in prelamin A processing has been known for some time [1,2], this is the first study to use human prelamin A peptides to rigorously examine prelamin A processing at a biochemical level. Our results suggest that the often-repeated and wellaccepted paradigm for prelamin A processing needs revision (Fig 8A, compared to S1 Fig). First, we show that the rate of cleavage by ZMPSTE24 at the upstream SY^LL site is five times greater than at the CSIM site. Second, we showed that SY^LL cleavage is not dependent on prior CSIM processing.
Instead, when both sequences are present in one peptide, ZMPSTE24 first cleaves at the SY^LL site then at the CSIM site ( Fig 8A). Hence, the SY^LL cleavage occurs before the C-terminal processing.
Third, we show that processing of the C-terminal CSIM involves a CS^IM dipeptide cleavage reaction, not a C^SIM tripeptide cleavage. The dipeptide cleavage was observed regardless of peptide length, whether or not the peptide was farnesylated, whether the ZMPSTE24 had been purified from insect or mammalian cells, or whether ZMPSTE24 was in a membrane or a detergent micelle. The discovery of a dipeptide cleavage is consistent with our atomic structure of ZMPSTE24 as a complex with a CSIM tetrapeptide, which revealed that the CSIM peptide was perfectly aligned for a dipeptide cleavage [34] ( Fig 9). Additional potential processing routes when RCE1 and ICMT are present in addition to ZMPSTE24.
In this case farnesylated prelamin A could be processed by RCE1-mediated release of the SIM tripeptide from the C-terminal CSIM motif. This product could then undergo SY^LL cleavage by ZMPSTE24 directly or be further modified by ICMT before the SY^LL cleavage. We took advantage of yeast expressing both human ZMPSTE24 and full-length human prelamin A constructs to examine ZMPSTE24-mediated prelamin A processing in living cells.
Consistent with our biochemical studies with prelamin A peptides, ZMPSTE24 faithfully carried out the upstream SY^LL cleavage within prelamin A, releasing mature lamin A, regardless of whether the prelamin A terminated in CSIM or was modified to terminate with CVIA, CVLS, or CTLM. Yeast afactor biogenesis studies were also consistent with our in vitro biochemical findings. The biogenesis, and secretion of mature active a-factor, which can be monitored with a halo assay, requires farnesylation, a tripeptide Caa^X cleavage, and methylation of the newly exposed farnesylcysteine, followed by upstream cleavages within the a-factor precursor. In our current studies, we showed that Ste24, the yeast orthologue of ZMPSTE24, is capable of promoting a-factor biogenesis in rce1Δ yeast from a-factor precursors that terminate with CVIA or CVLS but not from precursors that terminate with CTLM or CSIM. It is likely ZMPSTE24 carries out a C-terminal dipeptide rather than tripeptide cleavage, in the CTLM and CSIM proteins, preventing farnesylcysteine methylation and thereby preventing secretion of mature active a-factor which requires farnesylation and methylation for secretion and activity [36]. In contrast, Ste24 likely carries out a tripeptide cleavage in the CVIA and CVLS a-factor precursors, allowing carboxyl methylation of the farnesylcysteine and production of mature a-factor. Consistent with this interpretation, mature bioactive a-factor was produced from the CTLM and CSIM precursors in yeast that expressed Rce1, which is a bona fide CaaX tripeptide cleavage enzyme.
The significance of the farnesylation on the C-terminal CSIM cysteine is an intriguing question.
In our biochemical studies, we observed cleavage of nonfarnesylated prelamin A peptide at the SYLL site, suggesting that protein farnesylation is not absolutely critical for processing. However, in vivo evidence with genetically modified mice suggests that the SY^LL cleavage occurs only after protein farnesylation. Davies et al., [49] created prelamin A knock-in mice in which the C-terminal CSIM motif was changed to SSIM, thus abolishing prelamin A farnesylation. In these mice, there was no detectable processing of the nonfarnesylated prelamin A to mature lamin A, despite normal ZMPSTE24 expression. In addition, a prelamin A with a C-terminal CSIM sequence synthesized in yeast that are expressing either yeast Ste24 or human ZMPSTE24, the prelamin A is correctly processed. However, prelamin A with a C-terminal SSIM sequence is not cleaved [48]. These findings imply that in vivo the SY^LL cleavage depends on prelamin A farnesylation. Farnesylation of prelamin A may facilitate localisation to the membrane surface, bringing it in close proximity to the intramembrane ZMPSTE24.
In cells, the C-terminus of prelamin A will also be exposed to RCE1 and is thus a substrate for RCE1-mediated C^SIM tripeptide cleavage (Fig 8B). However, ZMPSTE24-mediated processing of prelamin A, with release of mature lamin A, does occur in RCE1-or ICMT-deficient fibroblasts, although the efficiency is slightly reduced (evident from a small amount of uncleaved prelamin A in cells) [42]. In any case, it seems likely that there will be a variety of modifications to the C-terminus of 20 prelamin A and that the proportion of each modification will depend on expression and localisation of the substrates, intermediates and enzymes. However, it would appear that ZMPSTE24 is capable of processing farnesylated prelamin A, and any of the C-terminal modified prelamin A variants, removing 15-18 residues, including the farnesyl membrane anchor, thus producing the required mature lamin A.
The physiologic purpose for prelamin A processing, and why prelamin A processing has been conserved through mammalian evolution, has remained elusive. Coffinier et al., [50] generated a knockin mouse in which a stop codon was introduced into the prelamin A gene after the codon for the last amino acid in mature lamin A. This mouse produced mature lamin A directly (bypassing ZMPSTE24mediated processing). Surprisingly, no overt histopathology was observed in those mice, but fibroblasts from the mice had greater numbers of nuclear blebs and appeared to have relatively lower amounts of lamin A at the nuclear rim. In light of these findings, we suspect ZMPSTE24-mediated processing of prelamin A optimizes targeting of mature lamin A to the nuclear rim. While direct production of mature lamin A did not result in overt histopathology in laboratory mice, we nevertheless suspect that optimized delivery of mature lamin A is probably important for optimal function of the nuclear lamina and for that reason has been favored during evolution. Conversely, it is clear that that lack of prelamin A processing by ZMPSTE24, with accumulation of permanently farnesylated prelamin A, is detrimental to human health.
Given the importance of prelamin A post-translational processing in a range of conditions such as progeria, lipodystrophies, and potentially even in normal ageing, an understanding of how ZMPSTE24 functions in normal and pathological conditions is essential. This work provides a new paradigm for the role of ZMPSTE24 in prelamin A processing, suggesting alternative views of its activity in cells. Now that we are clear which reaction is central to ZMPSTE24 function, we are in a better position to develop therapies for ZMPSTE24 related diseases.

Peptides and reagents
The peptide substrates were designed based on sequence of the C terminus of human prelamin A and purchased from commercial sources: Peptides 1 to 4 were purchased from Peptide Protein Research Limited (Fareham, United Kingdom) and Peptide 5 was purchased from Severn Biotech (Kidderminster, United Kingdom). The sequences are shown in Table 1. Nε-DNP-L-lysine hydrochloride, and 2-Aminobenzoic acid were purchased from Sigma-Aldrich Company Ltd. The peptides, Abz, and K(Dnp) were dissolved in DMSO at a final concentration of 10 mM. The stocks were aliquoted and kept at -20°C for future use.

Protein expression and purification
ZMPSTE24 was expressed and purified as previously published with modifications [34]. In brief, the full-length protein was expressed using Bac-to-Bac® Baculovirus Expression System  prior to the injection. The peak fractions were pooled together and TEV protease was added at a weight ratio of 1:10 for overnight treatment. The His-tagged TEV protease was removed by a second TALON resin binding for 1 h, and the flow through was collected and analyzed by SDS-PAGE and MSD-ToF electrospray ionization orthogonal time-of-flight mass spectrometer (Agilent Technologies Inc.). The protein was stored at 4°C for up to 2 days or flash-frozen and kept at -80°C for longer periods.

In vitro proteolysis assay
In brief, the 10 mM peptide stock was firstly diluted to 1 mM in 100 mM HEPES-NaOH, pH 7.5 and further diluted to twice of the desired concentrations in GF buffer.

Activity and kinetic analysis
The raw data were graphed using Microsoft Excel. The relative fluorescent intensity (RFI) was converted to concentration units (nM) by referencing the standard curve, in which the RFI of a 1:1 mixture of fluorophore and quencher pair at varying concentrations (0 to 90 µM) was measured in the presence of the same range of substrate concentrations (0 to 90 µM).
The initial velocities (nM s -1 ) were plotted against the substrate concentrations and analyzed using PRISM software (GraphPad Software Inc.). To obtain the apparent kinetic parameters (i.e. Km and Vmax), the graph was fitted to a non-linear regression one-phase decay equation as the catalytic reaction obeyed Michaelis-Menten kinetics. The S.E. values shown represent error with respect to curve fitting.

Mass Spectrometry
All the mass spectrometry measurements in this study were performed using an Agilent 1290 Infinity LC System in-line with an Agilent 6530 Accurate-Mass Q-TOF LC/MS (Agilent Technologies Inc.). The reaction mixture of the proteolysis assay at 100 µM peptide concentration was used as the sample for mass spectrometry analysis. 5 µl of the reaction mixture was diluted to 60 µl with 30% methanol in 0.1% formic acid. 60 µl of sample was injected onto a ZORBAX StableBond 300 C3 column (Agilent Technologies Inc.) by an auto sampler. The solvent system consisted of 0.1% formic acid in ultra-high purity water (Millipore) (solvent A) and 0.1% formic acid in 100% methanol (solvent B). Initially, 30% solvent B was applied at a flowrate of 0.5 ml/min. The sample was eluted by a linear gradient from 30% to 95% of solvent B over 7 min. Elution was then isocratic at 95 % B for 2 min, followed by a further 2 min equilibration at 30% B. The mass spectrometer was operated in positive ion, 2 GHz detector mode. Source parameters were drying gas 350 °C, flow 12 l/min, nebulizer 60 psi, capillary 4000 V. Fragmentor was 250 V, collision energy 0 V and data acquired from 100-3200 m/z.
Protein or peptide (fragments of peptide) intact mass was acquired using an MSD-ToF electrospray ionisation orthogonal time-of-flight mass spectrometer (Agilent Technologies Inc.). Data analysis was performed using MassHunter Qualitative Analysis Version B.07.00 (Agilent) software.

Time course mass spectrometry
The time course mass spectrometry was performed to characterize the cleavage of Peptide 4, 5 and 6. In brief, the peptide stock was prepared as previously specified. 20 µl of WT ZMPSTE24 at the concentration of 0.08 mg ml -1 was mixed with 20 µl of 200 µM Peptide 4, 5 or 6, and incubated at 37°C in a heating block. 5 µl of the reaction mixture was taken out and added into 55 µl of 30% methanol to terminate the reaction at time point 1 min, 2 min, 5 min, 10 min, 30 min and 60 min. The samples were kept on ice until loaded onto the Q-TOF LC/MS. The mass spectrometry analysis was performed as previously described. When looking for specific species from the reaction mixture, the extracted ion chromatogram (EIC) function was used.

Proteoliposome reconstitution
A stock of phosphatidyl-choline (POPC) and phosphatidyl-ethanolamine (POPE) (Avanti) at 3:1 (w/w) was dried under argon and re suspended in detergent free GF buffer to a final concentration of 20 mg ml -1 . The liposomes were disrupted by the addition of 22 mM sodium cholate (Anatrace).
Purified ZMPSTE24 was added into the mixture at lipid protein ratio (LPR) at 20:1 (w/w). After 30 min incubation on ice, Bio-Beads SM-2 (Bio-rad) were supplemented into the mixture to remove the detergent. The Biobeads were changed to fresh batch after 2 h for an overnight incubation. A further 1 h incubation with fresh Biobeads was performed the next morning. To make sure the complete removal of detergent, the proteoliposome was centrifuged at 100, 000 g for 1 h and the pellet was re-suspended in detergent free GF buffer at lipid concentration 5 mg/ml. The resuspension was then applied onto a LiposoFast extruder (avestin) and extruded against a membrane with a pore size of 400 nm.