Processing of predicted substrates of fungal Kex2 proteinases from Candida albicans, C. glabrata, Saccharomyces cerevisiae and Pichia pastoris

Heterologous expression and purification of Kex2 proteinases

In its native form Kex2 is a type I membrane protein with the catalytic domain located inside the trans Golgi network lumen (Figure 1A). As neither the transmembrane domain nor the cytosolic domain are necessary for the catalytic activity, it is possible to produce a soluble and secreted version of this enzyme by truncation of the gene just before the sequences encoding the transmembrane domain [21].

For the expression of the soluble forms of S. cerevisiae, C. glabrata and P. pastoris Kex2 enzymes the P. pastoris expression system (Invitrogen) was used. The strain expressing S. cerevisiae Kex2 was a kind gift of Guy Boileau [22]. For the expression of C. glabrata and P. pastoris Kex2 enzymes the 5' part of the gene coding for the luminal domain of the enzyme, including the native signal- and pro-peptide, plus a C-terminal 6 × His-tag were cloned into the pic3.5 vector (Figure 1B and 1C) and transformed into P. pastoris strain GS115. The transformants displaying the strongest extracellular proteolytic activity (ppCgKex2#12 and ppPpKex2#5) in test expressions were used for large-scale production of the enzymes.

Attempts to purify the C. glabrata and P. pastoris Kex2 enzymes via 6 × His-affinity chromatography were not successful, possibly due to burial of the epitope inside the protein. Thus, all three enzymes, including the one from S. cerevisiae, were purified to near homogeneity by a combination of anion exchange and size exclusion chromatography (Additional file 1).

Because several attempts to produce the intact, soluble form of Kex2 of C. albicans in the Pichia system failed, ultimately the native host C. albicans was used for production of this enzyme: the 5' part of the C. albicans KEX2 gene coding for the luminal domain of the enzyme, again including the native signal- and pro-peptide as well as a C-terminal 6 × His tag was put under the control of the constitutive and strong promoter of the ACT1 gene, as described under Methods (Figure 1D). The linearized plasmid was transformed into C. albicans strain CAI4 and the transformant giving the strongest Kex2-like activity in the supernatant (CaAct1-Kex2#7) was used for further large scale production of the enzyme, as above.

While we were able to produce the high Kex2 activity in supernatants, the efficiency of its purification remained low. Highest yields of enzyme were achieved using complex media including yeast extract and peptone, but this resulted in only impure enzyme preparations. However, the parental strain did not produce this activity (Additional file 1, Figure 1B). To avoid low-weight impurities in the enzyme preparations, which would have disguised product bands in further analytical experiments, the medium was passed over a 10 kDa size-exclusion column prior to the expression. In combination with the purification methods as outlined above this resulted in an enzyme preparation that contained only few other proteins and was devoid of low molecular weight contaminants.

Activity testing of the enzyme preparations

Prior to use, enzyme preparations were adjusted to a common activity of one nmol/min per μl proteinase added in a standard reaction setup with the chromogenic substrate Z-Tyr-Lys-Arg-pNA. Neither preparations from a P. pastoris negative control strain nor from the C. albicans parental CAI4 strain displayed this activity (data not shown). In addition we performed controls with the C. albicans enzyme preparation to ascertain that the proteolytic activity was Kex2-dependent: The activity was indeed inhibited by PMSF, EDTA and ZnCl₂, but not by pepstatin A (Additional file 1, Figure 1C).

To test whether the enzymes had similar properties we first tested the enzymes for optimal pH and temperature with the chromogenic substrate Z-Tyr-Lys-Arg-pNA. The optimal pH for all enzymes was between 7.2 and 7.4 (data not shown), as described earlier for the S. cerevisiae enzyme [23] and this pH was therefore used throughout all further experiments. In contrast, the result for the optimal temperature was surprising: all enzymes showed an elevated activity at unphysiological temperatures from 40°C to 50°C (data not shown), at which none of the source organisms display optimal growth, if any. Nevertheless, all following experiments were carried out at 37°C, reflecting human body temperature, as our main focus lay on the enzymes of the human pathogenic fungi C. albicans and C. glabrata.

Since the KEX2 gene of C. albicans can complement the kex2 deletion in S. cerevisiae [8] and the KEX2 gene from S. cerevisiae can complement the loss of the KEX2-ortholog KRP1 in Schizosaccharomyces pombe [12], it is feasible to assume that these enzymes have similar to identical biological functions and biochemical properties. To show that this is also the case for the Kex2 proteinases from C. glabrata and P. pastoris, we tested whether all four proteinases cleaved the S. cerevisiae α-mating pheromone, a natural and proven substrate of Kex2 from S. cerevisiae, in a similar manner (Figure 2): The α-mating pheromone precursor protein was purified using the pET100-D E. coli expression system as described below for the other substrate proteins. Indeed, the four proteinases showed the same digestion pattern of the pheromone precursor into the expected fragments of the N-terminal 11 kDa peptide and peptides of 2–3 kDa size (Figure 2).

Prediction of potential Kex2 substrates

Next, we developed a prediction method for potential Kex2 cleavage sites in substrate proteins to identify proteins from C. albicans, C. glabrata or S. cerevisiae for testing with the proteinases. Earlier studies [9, 10] used very stringent search parameters and only looked in the N-terminal region of protein sequences. However, there is biochemical and biological evidence for processing of sites containing other amino acids in the P2 position [4, 24] as well as activity on C-terminal motifs in other organisms such as the chloroperoxidase CPO of A. niger [25] and on membrane proteins such as Kex2 itself. Therefore we included Golgi-luminal portions of transmembrane proteins as well as full-length sequences of soluble proteins into our search. ER-retained proteins were excluded, as they should not come into contact with Kex2.

Expression of substrate proteins

From the 297 predicted potential Kex2 substrate proteins we selected a total of 43 proteins (three of S. cerevisiae, 26 of C. albicans and 14 of C. glabrata) for heterologous expression in E. coli (Additional file 3). These were chosen to cover a wide range of different cleavage sites and protein types and expressed using the TOPO-pET D100 system. The DNA fragments cloned were devoid of domains encoding signal peptides and putative GPI-anchor sequences. Out of these selected proteins, we were able to express and purify thirteen from C. albicans, ten from C. glabrata and one from S. cerevisiae (the α-pheromone mentioned above). Since the majority of the chosen proteins accumulated as inclusion bodies, we converted these proteins into a soluble form by on-column refolding. To test for overall correct folding of the refolded proteins, we performed an activity test for the substrate CA5147, an acid phosphatase, which was the only protein with a known activity in this set. Indeed, we were able to confirm the activity of this protein and observe a maximum activity at pH 4.2–4.3 (Figure 3) using para-nitrophenol phosphate as a substrate. This shows that at least some refolded protein assumes its native structure and can thus be used for specific proteolysis assays.

In vitro proteolytic processing of substrate proteins by Kex2 from C. albicans, C. glabrata, S. cerevisiae and P. pastoris

Very few studies provided experimental evidence that predicted Kex2 cleavage sites in potential substrate proteins are in fact processed by Kex2 proteinases. In order to determine susceptibility of the purified proteins to proteolytic processing by Kex2, all potential substrate proteins purified above were digested with each of the four proteinases. A selection of digestions is depicted in Figure 4. Based on the scores given by the algorithm, we expected most proteins to be cleaved. Indeed, we observed rapid cleavage at the predicted cleavage sites for 2/3 of the proteins. This also included cleavage at sites with lower scores in polypeptide precursors (e.g. CaEce1, position 92, Additional file 2). In contrast, some proteins were not cleaved even though they contained sites with high scores, such as CaCcw14 (see Additional file 2). Furthermore, one protein (CA0365) was cleaved very differently by the proteinases: while it was not cleaved at all by ScKex2, CaKex2 rapidly processed the precursor into peptide sized fragments, without any noteworthy appearance of intermediates under the standard reaction conditions (Figure 4). A similar activity was observed with CgKex2 and PpKex2, while at a considerably slower rate of hydrolysis.

Proteins that were cleaved into fragments of the expected sizes were CA0365, CaEce1 (CA1402), CA1873, CA2974, CaPga17(CA4679), CaTos1 (CA2303), CaSun41 (CA0883), CgScw4 (CAGL0M13805g), CgSUN4 (CAGL0L05434g), CgPir1 (CAGL0M08492g), CgPry1 (CAGL0F05137g) and CgPry2 (CAGL0G07667g). Proteins that remained fully uncleaved were CaCcw14 (CA2942), CaPho11 (CA5147), CaRbt4 (CA0104), CaCrh1 (CA0375), the Plb-homolog CAGL0J11770g and the three proteins of unknown function CA1394, CAGL0H08910g and CAGL0A02277g.

The pattern of cleavage vs. non-cleavage observed was not sufficiently explained by the score calculated from the prediction algorithm among the proteins tested. Therefore, we inspected the amino acid distribution surrounding the investigated and other known cleavage sites for other patterns: indeed, a high overrepresentation of negatively charged (aspartic/glutamic acid) and small (alanine, valine, leucine) residues in the P1', P2' positions and a similar moderate overrepresentation in the P4' position was found, while positively charged residues were underrepresented at those substrate sites which were digested. In the case for the sites not cleaved, no over- or under-representation was observed (Figure 5).

Reflection of substrate recognition in proteinase structure

Next, we asked whether the apparent preference for negatively charged residues in the P1'–P4' region of substrates digested by Kex2 proteins is reflected by the structure of the proteinases in the substrate binding cleft.

Recently, a 3D model of the bacterial subtilisin kumamolisin of Bacillus novospec was published [28]. The enzyme studied there was incapable of autoproteolytic activation thus retaining the pro-peptide. By superimposition with the coordinate sets of S. cerevisiae Kex2 [29] and Mus musculus furin [30] we were able to investigate the potential substrate binding pockets in the P1' – P4' region (Figure 6A) as outlined by the intact pro-peptide cleavage site still bound into the substrate binding cleft. Indeed, the P4-P1 positions of the Kumamolisin pro-domain aligned with the known S4-S1 pockets of the enzymes (not shown), as well as the P1'–P3' positions with the S1'–S3' pockets predicted in the literature [31, 32] (Figure 6A).

The neighbouring S1' and S3' pockets are characterized by positive charges in ScKex2 (H213, H381) as well as in furin (R193, H194, H364), and both pockets may well accommodate aspartate or glutamate residues in the substrate. In furin, the excess charge possibly results in a stronger selection for negatively charged residues in the P1' position, but as the S2 pocket is directly adjacent to the S1' pocket, the lack of a positively charged P2 residue in furin substrates may compensate this effect. The S2' pocket, located on the opposite side of the cleft, as well contains a terminal positive charge (R318 in ScKex2, R298 in furin) which would favour negatively charged residues in the P2' position.

A potential P4' pocket was also identified (Figure 6B and 6C). The P4' residue aligns between S363 and Q350 and extends towards E362 in the furin model (Figure 6C). The alignment with ScKex2 is of lesser quality in this region, but nevertheless a similarly built potential binding pocket is seen in the ScKex2 enzyme bordered by S380 and Y367 (Figure 6B). However, the equivalent to the negative terminal charge of E362 in furin would be the positive charge of H369 in ScKex2.

In summary, the structure of the enzymes explains the increased preference for negatively charged P1'–P4' residues in the substrates.

Conservation of residues involved in substrate recognition

It is known from previous studies, that C. albicans KEX2 can complement KEX2 in S. cerevisiae [8] and this gene in turn can complement the KEX2 ortholog KRP1 in S. pombe [12]. Therefore, it must be concluded that the corresponding proteinases have similar substrate specificities and activities. Nevertheless, we have been able to show that at least in the case of one substrate (CA0365) the proteinases of S. cerevisiae and C. albicans behave differently. To investigate whether this difference as well as the question whether or not the substrate specificity in general is the same in different fungi, we generated a sequence alignment of Kex2-orthologous proteins from fungi and furin-orthologous proteins from mammals (Figure 7) and investigated the residues involved in substrate recognition (Table 2) for their degree of conservation between the different species.

#	ScKex2	MmFurin	Pocket	Reference
a	D175	D153	S2	[29]
b	D176	D154	S2	[29]
c	D210	D191	S2	[29]
d	D211	N192	S2	[29]
e	Y212	R193	S3' S1'	[31, 32] This work
f	H213	H194	S1'	[32]
g	R216	R197	S3'	[32]
h	I245	M226	S4	[29]
i	L246	L227	S4	[29]
J	I250	V231	S4	[29]
k	T252	D233	S4, S6	[67]
l	D254	V235	S4	[29]
m	E255	E236	S4	[29, 67]
n	W273	W254	S4	[29]
o	A276	E257	S3, S5	[32, 29]
p	D277	D258	S1	[29]
q	D278	D259	S5	[32]
r	H281	T262	S5	[32]
s	Q283	D264	S4, S6	[67]
t	A311	A292	S1	[29]
u	N314	N295	S1	[68]
v	T317	R298	S2'	[31]
w	R318	E299	S2'	[31]
x	D320	D301	S1	[29]
y	D325	D306	S1	[29]
z	Y327	Y308	S4	[31]
A	P347	W328	S2'	[31, 32]
B	Y348	Y329	S2'	[31]
C	E350	E331	S1	[29]
F	Y367	Q350	S4'	This work
G	H369	-	S4'	This work
H	N379	E362	S4'	This work
D	S380	S363	S3' S4'	[31] This work
E	H381	H364	S1'	[32]

Residues relevant for substrate recognition given in the literature or identified in this study were compiled. Column a: Residues were numbered "a" trough "E". The identical numbering was also used in Figure 6 and Figure 7; columns 2 and 3: The respective analogous residues in ScKex2 (Genebank accession no. CAA96143) and MmFurin (Genebank accession no. CAA37988) are given along with the reference.

The S1 pocket (composed of positions p, t, u, x, y and C) is fully conserved and among fungi this is also true for the four negative charges of the S2 pocket (positions a, b, c and d). Interestingly, we observed for the S4 and the S1' position that the enzymes from Ascomycetales combine the charge-selective properties of the S. cerevisiae Kex2 enzyme with those from the furin enzymes, and thus probably display the most discrete substrate recognition. Among the Saccharomycetales the residues are conserved for the major subsites S4, S2, S1 and S1' with minor exceptions only. Differences are visible in subsites where there is no strong selection to or discrimination against substrate residues, such as the S5 pocket (positions q and r). The S2' pocket is generally positively charged, however, this charge is mediated by one histidine in either the v or the w position.

In summary, it is seen, that the substrate selectivity among Saccharomycetales Kex2 enzymes is very conserved, and that there are no substitutions that would explain the differential processing of substrate CA0365 between the four proteinases. Therefore, the enzymes must discriminate their substrates either through further subsites or through processes independent of the primary sequence surrounding the cleavage site.

Relevance of substrate structural features for cleavage

During the in vitro cleavage experiments, we observed that proteins purified from the soluble fraction of E. coli lysates were generally processed more efficiently than those purified and refolded from inclusion bodies. Therefore, we predicted that the three dimensional structure of the substrate and the exposure of the putative processing site on the protein surface is crucial for processing to occur. To investigate this further, we tested if sites that were readily cleaved in the native protein were still cleaved in a denatured form of the protein: two substrate proteins that were readily cleaved by ScKex2 (CaEce1 and CA1873) were heat denatured prior to addition of the ScKex2 proteinase (Figure 8A). As expected, both were cleaved less in the denatured form. This effect is more pronounced for CA1873 than for CaEce1, as CaEce1 contains seven equal cleavage sites and is thus generally more prone to processing than CA1873. The reduced cleavage of partially denatured/refolded proteins can be explained by either inaccessibility of the site due to burial in the denatured structure or by the failure to form a specific secondary structure needed for processing.

Also, we did not observe cleavage for all proteins with potentially good sites. Therefore, we tested if this was due to an uncleavable primary sequence or if there were structural constraints preventing cleavage: site 3 of CaCcw14 and site 1 of CA0365, were each fused between a GST and a GFP domain and so exposed to the solvent. The GST-CA0365 [1]-GFP fusion protein was not cleaved (Figure 8B, lane 2), indicating that this sequence is not a substrate of ScKex2 and the non-cleavage of the full length protein is not due to structural constraints, as was expected due to the cleavage by the other three Kex2 enzymes. In contrast, the GST-CaCcw14 [3]-GFP fusion protein was readily cleaved by ScKex2 (Figure 8B, lane 5), demonstrating that this primary sequence reflects a good substrate and the non-cleavage in the full length protein must be due to structural constraints. This gives further evidence that accessibility and/or secondary structure of the cleavage site are essential for processing.

Oligonucleotides

Oligonucleotides (TIBMolBiol, Germany) used for cloning of expression vectors in this study are given in Additional file 4.

Heterologous proteinase expression in Pichia pastoris

Candida glabrata and Pichia pastoris Kex2 enzymes were expressed using the Pichia expression system (Invitrogen) according to manufacturer's instructions. Briefly, the DNA coding for the Golgi-luminal part of the protein was PCR-amplified from genomic DNA with oligonucleotides containing terminal restriction sites (BamHI/NotI and SnaBI/NotI, respectively) and a sequence for a C-terminal 6 × His-tag, cloned into the pic3.5 vector and transformed into Pichia strain GS115 using an optimized electroporation protocol [49]. Transformants were screened by testing of enzymatic activity against the chromogenic substrate Z-Tyr-Lys-Arg-pNA (see below) in the supernatant of pilot expressions and the clone exhibiting maximum activity used for scale-up. For large-scale production, cells were grown in 500 ml buffered minimal glycerol medium at 30°C over night, harvested by centrifugation, washed and resuspended in 50 ml buffered minimal methanol medium. Maximum activity was detected after 16 h of growth, after which the culture supernatants were harvested and the recombinant enzymes purified as described below.

Heterologous proteinase expression in Candida albicans

The soluble form of C. albicans Kex2 was expressed in the native host, as several attempts of heterologous expression in P. pastoris failed. The KEX2 gene was PCR-amplified from genomic DNA of C. albicans with primers containing restriction sites (HinDIII/NheI) and the sequence for a 6 × His-tag, cloned into pCIp10 [50] and thus put under control of the mainly constitutive C. albicans ACT1 promoter. The plasmid was linearized with NcoI and transformed into C. albicans strain CAI4 using the same protocol as above for Pichia transformation. Transformants were selected on minimal medium and screened using the supernatant of 5 ml YPD (1% yeast extract, 2% peptone, 1% glucose) over night cultures for testing of enzymatic activity as above. For preparative expression, a 500 ml YPD culture was grown over night, the cells harvested, washed twice with 50 ml YPD, resuspended in 50 ml YPD and further cultivated at 30°C. Maximum Kex2 activity in the supernatant was observed after 12 h of growth at 30°C, after which the supernatant was collected and the recombinant protein purified as described below. YPD medium used for expression was previously freed from low molecular weight impurities by passing over a 10 kDa size exclusion Centricon-20 column (Millipore),

Purification of secreted soluble Kex2 proteins

To purify the recombinant enzymes, 50 ml sterile filtered expression culture supernatant were concentrated on a 30 kDa size-exclusion Centricon-20 column (Millipore) to a volume of approximately 1–2 ml, desalted using a PD-10 column (Amersham Biosciences) and eluent diluted to a volume of 20 ml into IAEX buffer (50 mM BisTris pH 4.5, 10 mM NaCl). This was loaded onto an HiTrap ANX FF anion exchange column (Amersham Biosciences), washed, and eluted with IAEX buffer containing 100 mM NaCl. The eluent was then again concentrated, the buffer changed into storage buffer (50 mM BisTris, pH 7.2 50% w/v glycerol) and the enzymes kept at -20°C.

Proteinase activity quantification

Proteolytic activity of the purified enzymes was assayed using the chromogenic substrate Z-Tyr-Lys-Arg-pNA (Bachem, Switzerland) as described previously [23]. Assays were done in buffer containing 50 mM BisTris (pH 7.2), 1 mM CaCl₂, 0.5 mM substrate in a total volume of 100 μl at 37°C. For the measurement of time kinetic data, the reaction was started by mixing 50 μl of solution containing the proteinase with 50 μl containing the substrate. The temperature gradient for optimal reaction temperature measurement was generated in a thermocycler (Biometra) and the reaction terminated by the addition of EDTA to a final concentration of 10 mM. Liberation of p-nitroannilide (pNA) was measured at 405 nm in a spectrophotometer (Tecan). All measurements were calibrated against negative controls without proteinase and repeated at least three times.

In silicoidentification and analysis of substrate sequences

The protein sequence sets analysed here were downloaded from the Genolevures Website [51, 52], from CandidaDB [53, 54] and from the Stanford Genomic Resources FTP server [55]. Sequence analysis on genome-scale was done using custom perl scripts within the bioperl framework, incorporated into a local database, as described in the results section. Entry into the secretory pathway and membrane topology were predicted with the Phobius algorithm [56]. Sequence logos were created using the Weblogo website [57]. All programs were run under SUSE Linux 10.1.

Heterologous expression, renaturation and purification of substrate proteins

Heterologous expression of substrate proteins was done using the pET100-D TOPO vector system (Invitrogen) in E. coli, strain Rosetta (Novagen). Either, exponentially growing cells were induced with IPTG and grown for 3 h in a volume of 50 ml or autoinduced by growth in 50 ml LB containing 0.05% glucose and 0.2% lactose [58]. Harvested cells were lysed in 5 ml BugBuster (Novagen) with Benzonase nuclease (Novagen) and lysosyme (Sigma) added according to the manufacturers' description. Proteins expressed in a soluble form were purified using 6 × His chelating chromatography and analyzed by SDS-PAGE. Proteins expressed in form of inclusion bodies were refolded using β-cyclodextrin [59]. Inclusion body pellets were dissolved in solubilization buffer (100 mM Na₂HPO₄, 100 mM NaCl and 8 M urea) at 60°C. The denatured protein was bound to Ni-Agarose and washed with 20 bed volumes of solubilization buffer. The urea was thoroughly removed with buffer A (100 mM Na₂HPO₄) containing 0.1% Triton X-100 (Sigma). Excess Triton X-100 was removed by washing with 10 bed volumes of buffer A and the bound protein refolded over night in buffer A containing 5 mM β-cyclodextrin (Sigma). The refolded, bound protein was then again washed with 20 bed volumes buffer A and eluted with 200 mM imidazole and 0.001% Triton X-100. The eluent was passed through a 20 μm sterile filter to remove aggregates and then analyzed via SDS-PAGE.

To confirm proper folding by this method we tested one of the proteins, a putative acid phosphatase, for activity. One μl each of the refolded protein solution was assayed in 100 mM NaAcetate buffered from pH 3 to 6 at room temperature towards its activity against 0.5 mM para-nitrophenol phosphate (pNPP). Liberation of p-nitrophenol was measured in a spectrophotometer (Tecan) at 405 nm and the experiment repeated three times.

Proteolytic digestion of substrate proteins and detection of digestion products

Proteolytic digests of Kex2 substrates were performed in Kex2 buffer (20 mM Tris-HCl, 1 mM CaCl₂, pH 7.2, modified from [60]) at 37°C. To observe intermediate products in the case of proteins with several potential cleavage sites, time series of the reactions were conducted by stopping the reaction of aliquots by fast heating to 95°C. Depending on the expected product sizes, the proteolytic digests were resolved either on Tris-glycine or Tris-tricine [61] polyacrylamide gels and visualized by silver staining and/or western blotting using an antibody against the N-terminal Xpress epitope (Invitrogen).

Alignment of proteinase structures and sequences

Sequences were aligned locally with the ClustalW algorithm [62] and edited with the BioEdit software [63]. Coordinate sets for proteinase structures were downloaded from the Protein Data Bank [64], superimposed using the MASS algorithm [65] and visualized with RasMol [66].