A multi-enzyme cascade of hemoglobin proteolysis in the intestine of blood-feeding hookworms

The aspartic proteases (APR-1), cysteine proteases (CP-2) and metalloproteases (MEP-1), the former of which is known to digest Hb. We now show Hb is degraded by a multi-enzyme, synergistic cascade of proteolysis. Recombinant APR-1 and CP-2, but not MEP-1, digested native Hb and denatured globin. MEP-1, however, did cleave globin fragments that had undergone prior digestion by APR-1 and CP-2. Proteolytic cleavage sites within Hb α and β chains were determined for the three enzymes, identifying a total of 131 cleavage sites. By scanning synthetic combinatorial peptide libraries with each enzyme, we compared the preferred residues cleaved in the libraries with the known cleavage sites within Hb. The semi-ordered pathway of Hb digestion described here is surprisingly similar to that used by Plasmodium to digest Hb and provides a potential mechanism by which these are efficacious vaccines in animal models of hookworm


INTRODUCTION
In mammals, gastric digestion of proteins derived from food involves a cascade of mechanistically distinct proteases. One of the major food sources of blood-feeding parasites, hemoglobin (Hb), is presumed to be degraded in a similar manner by a cascade of proteases that cleave the intact Hb tetramer into successively smaller fragments that are used to provide nutrients for anabolic processes such as maturation and reproduction. To date, the only hematophagous parasite where such a proteolytic cascade of Hb digestion has been identified is the malaria parasite, Plasmodium falciparum, where a semi-ordered catalytic pathway exists including members of at least three different mechanistic classes of endoproteases. Hb is degraded in the P. falciparum digestive vacuole initially by aspartic proteases termed plamepsins (1, 2), and this step can be inhibited by addition of the aspartic protease inhibitor, pepstatin A (3). P. falciparum also expresses vacuolar cysteine proteases, termed falcipains, which are able to digest native Hb and denatured globin (4)(5)(6). After initial digestion with plasmepsins and falcipains, a P. falciparum metalloprotease, falcilysin, digests globin fragments even further (7).
Hematophagous helminths include representatives from diverse phyla including nematodes and platyhelminths. Their life cycles are complex, so to ensure transmission in the face of environmental adversity, helminths release vast quantities of fertilized eggs, a physiologic process that necessitates a protein-rich diet, such as blood. This is evident in the increased consumption of blood by female schistosomes (8) and the overexpression of mRNAs encoding hemoglobinolytic proteases in female schistosomes (9). Aspartic and cysteine proteases are present in the gut lumen and surrounding gastrodermis of blood-feeding adult schistosomes, and while an ordered pathway of Hb degradation is assumed, it has yet to be characterized in any detail (10,11).
Hookworms are blood feeding nematodes that reside in the small intestine of infected mammalian hosts (12). An estimated 740 million people are infected with the human hookworms Ancylostoma duodenale and/or Necator americanus (13). Many hookworminfected people, particularly women and children, develop sub-clinical or clinical hookworm disease including anemia and impaired physical and cognitive development as a direct result by guest on March 23, 2020 http://www.jbc.org/ Downloaded from of blood loss at the site of attachment in the intestine (12). The impact of this blood loss in chronically infected populations is enormous. For example, in a single village of approximately 800 people in South China, it was estimated that hookworms caused an annual blood loss of 200 liters (14). To obtain an unimpeded blood meal, hookworms secrete potent anticoagulants, the actions of which have been well characterized (15,16). Once blood is ingested by adult hookworms, they lyse erythrocytes using a pore-forming, membrane-bound hemolysin (17), releasing the red cell contents into the intestinal lumen for proteolytic degradation. The pH of the hookworm intestine is not definitively known, however, it is presumed to be acidic in nature (18). All of the proteaes identified from this anatomic site have acidic pH optima (19)(20)(21) and Hb digestion by hookworm secretory extracts is optimal at pH 5-7 (22). Moreoever, the intestinal contents from the related blood-feeding helminth, Schistosoma mansoni, are acidic pH (11). mRNAs encoding hookworm proteases of different mechanistic classes, including aspartic (19,20,23), cysteine (24) and metalloproteases (25,26) have been cloned and the expression sites of some have been localized to the intestinal brush border of the blood-feeding adult stage (for a review, see (18)). Until now, only the aspartic proteases, APR-1 and APR-2, had been shown to digest Hb in vitro (19,20). Moreover, these hookworm proteases were more efficient at digesting Hb from permissive host species than non-permissive hosts, despite sharing identical active sites residues, highlighting the exquisite co-evolution of these enzymes with host substrates (19,20,27). We suggested that hookworms likely employ a cascade of hemoglobinolysis utilizing multiple proteases of distinct mechanistic classes (18).
A cathepsin B-like cysteine protease, Ac-CP-2 (24), was recently shown to be expressed in the intestinal brush border of adult A. caninum, and immunization of dogs with the recombinant protease resulted in an anti-fecundity effect on female hookworms (21). In addition, a neprilysin-like metalloendopeptidase was identified from the intestinal brush border of adult A. caninum (25), however the recombinant molecule was not expressed in catalytically active form, precluding determination of a functional role for this enzyme.
Here we show that the canine hookworm digests Hb in a semi-ordered cascade that consists of at least aspartic, cysteine and metalloproteases that act in synergy at acidic pH. The classes of by guest on March 23, 2020 http://www.jbc.org/ Downloaded from proteases involved and the order of their actions is strikingly similar to that used by P.
falciparum to digest Hb in the food vacuole.

Protease cDNAs and protein expression
Cloning of cDNAs encoding the proteases described here has been reported elsewhere; Ac-APR-1 (23) (GenBank U34888); Ac-CP-2 (24) (GenBank U18912); Ac-MEP-1 (25) (GenBank AF273084). Expression and purification of recombinant APR-1 in baculovirus (20) and CP-2 in the yeast, Pichia pastoris (21), have been reported previously. To express MEP-1 in baculovirus, the entire ORF minus the predicted signal peptide was amplified by PCR using an MEP-1 pBluescript plasmid as the template. The 5' PCR primer incorporated a Sac I restriction site and was designed so that the recombinant protein was fused in-frame with the honey bee melittin signal peptide encoded by pMelBac (Invitrogen). The 3' primer incorporated an Nco I restriction site preceded by nucleotides encoding a hexaHis tag and stop codon. Amplicons were ligated into pMelBac by restriction cloning and plasmids were generated by heat shock transformation of TOP 10 E. coli (Invitrogen). Plasmid preparation, sequencing and protein expression were performed as described for APR-1 (19). Secretion of the recombinant protein was confirmed using an anti-hexa-His monoclonal antibody (Invitrogen) and a polyclonal antiserum raised to insoluble MEP-1 expressed in E. coli (25).
After generation of a high titer tertiary viral stock, Trichoplusia ni High 5 cells were infected with recombinant virus and maintained in shaker flasks at 27°C with constant shaking at 100 rpm as previously described (20).

Purification of recombinant Ac-MEP-1
Supernatant from 500ml Ac-MEP-1 baculovirus culture was concentrated to 100ml using a 10 kDa mw cut-off membrane (Pall Scientific), buffer exchanged into 2L of Tris buffered saline

Assessment of enzymatic activity and pH optima
The pH optimum and substrate preferences for Ac-APR-1 have been described previously (20,27). Recombinant Ac-CP-2 cleaved the substrate, Z-Phe-Arg-7-amino-4-methylcoumarin with an optimal pH range of 5.0 -7.0 as described previously (21). Optimal assay conditions for Ac-MEP-1 were determined using pre-cast gelatin gels (Invitrogen) and the Novex ® Zymogram Gels developing system as per manufacturer's guidelines (Invitrogen). To determine the optimal pH for proteolytic activity, zymogram gels were cut into strips and incubated in developing buffer (Invitrogen, containing 10 mM CaCl 2 or ZnCl 2 ) overnight at single pH value incremements from pH 1.0 to 8.0 in the presence/absence of 10 µM 1,10phenanthroline. Gelatin gels were stained with CBB to observe clear zones of hydrolysis.

Immunohistochemistry
Adult A. caninum were prepared for immunostaining as previously described (19). Nonspecific binding sites were blocked with 5% Fetal Bovine Serum in PBS/0.05% Tween 20 (PBS-T) for 1h. Dog antisera were produced against each protease formulated with the adjuvant AS03 as described for CP-2 (21) -animals were bled after the third vaccination but before parasite challenge. Antisera (diluted 1:100, 1:500, 1:1000 in PBS) were applied to each tissue section and incubated for 2h at RT in a humidity chamber. Subsequent steps were as described elsewhere (26).

Determination of subsite preferences using positional scanning synthetic combinatorial libraries
The P1-P4 specificities of recombinant Ac-CP-2 were investigated using a combinatorial fluorogenic peptide library established by the Proteomics group at The University of California, San Francisco as previously described (28). Briefly, aliquots of 25 nM (in the appropriate activity buffer for each protease) from the 20 respective residues of each P1-P4 sub-libraries were added to wells (8000 compounds diversity per well) of a 96-well Microfluor-1 U-bottom plates (Dynex Technologies). The assay was performed in 100 mM sodium acetate, pH 5.5, 100 mM NaCl, 10 mM DTT, 1 mM EDTA, 0.01% Brij-35, and 1% DMSO (from the substrates). Assays were initiated by the addition of enzyme and monitored fluorimetrically with a SpectraMax Gemini fluorescence spectrometer (Molecular Devices) with excitation at 380 nm and emission at 460 nm. All assays were performed in triplicate.
Aspartic and metalloproteases usually exert their specificities at the prime side of the cleavage site, so to determine the P1'-P4' subsite preferences for APR-1 and MEP-1, peptide library analysis was carried out as described previously (29). The substrate peptide library acetyl-

RP-HPLC, MALDI-TOF and MS-MS analysis of Hb hydrolysates
Hb that had been incubated for 15 mins with various combinations of hookworm proteases in the presence/absence of specific protease inhibitors were subjected to RP-HPLC analysis and the respective masses of Hb-derived peptides were determined by MALDI-TOF. Briefly, aliquots of 50 µl from each hydrolysate were separated by RP-HPLC using a Vydac C18 TP54 150x4.6 mm column (Bodman) and a linear gradient of water/acetonitrile (both containing 0.1% TFA) from 15% to 65% over 50 minutes. The HPLC separation was monitored by a UV detector set at 215 nm. Fractions were collected automatically every minute and analyzed by MALDI-TOF (Axima CFR plus, Kratos) either in linear or reflectron mode, using α cyano-4-hydroxycinnamic cinnamic as MALDI matrix. Aliquots of 20 µl from each sample were also analyzed by LC-MS. The instrumental set up consisted of an HPLC (HP 1100, Agilent) connected to an electrospray-ion trap mass spectrometer (LCQ, Thermofinnigan). Chromatographic conditions for the HPLC-MS analyses were as follows: flow rate of 50 µl/min, obtained with a microflow processor (Acurate, LC Packing) set to a split ratio 1:10, Vydac 238MS 1mm x 250 mm column, mobile phases were water (A) or 100% acetonitrile (B) both containing 0.01% heptafluoro butyric acid and 1% acetic acid.
The separation was carried out at 10% B for the first five minutes followed by a linear gradient from 10% to 50% B over 40 minutes. Mass spectrometric analysis was performed either in full scan mode or in data dependent mode (MS-MS). The latter analytical approach was used to perform the selective fragmentation of the ionic species that in any given scan (from 300 to 2000 atomic mass units) yielded the highest peak intensity, allowing both MW and structural information to be obtained for chromatographically separated peptides. MS-MS data were then analyzed using TurboSequest 2.0 software (30) (ThermoFinnigan).

Aspartic, cysteine and metalloproteases localize to the intestinal microvilli of adult hookworms
We previously localized APR-1 (20), CP-2 (21) and MEP-1 (25) to transverse sections of adult A. caninum. Here we confirm that each protease bound to the intestine only of longitudinal sections of adult hookworms (Figure 1) while control serum did not bind to any structures. Both male and female worms were probed, with no significant differences in localization observed between the two sexes (not shown).

Expression and substrate affinity purification of catalytically active MEP-1
Ac-MEP-1 was expressed by baculovirus in High 5 cells and secreted into culture medium at a yield of 4 mg.L -1 . The protease was successfully purified to homogeneity (as determined by CBB staining of SDS-polyacrylamide gels) using gelatin sepharose (Figure 2a) but did not bind to nickel-NTA agarose (not shown) despite the presence of a hexa-His tag confirmed by Western blotting using an anti-hexa-His monoclonal antibody (not shown). The purified protein migrated with an apparent molecular weight of 105 kDa (Figure 2a).

Ac-MEP-1 is enzymatically-active and degrades gelatin
Ac-MEP-1 readily degraded gelatin (with as little as 1.0 ng of recombinant enzyme) and activity was completely inhibited by 1,10-phenanthroline (Figure 2b). Activity was detected at a broad range of pH values; the most intense zones of hydrolysis were seen at pH 2.0 and again at pH 5.0-7.0 (not shown). Interestingly, gelatinolytic activity was not detected at pH 3.0 and only weakly at pH 4.0. Gelatinolytic activity was detected at similar levels using both 10 mM CaCl 2 and ZnCl 2 (not shown).

Subsite specificities of hookworm proteases using synthetic combinatorial libraries
Aspartic and metalloproteases normally exert their specificities C-terminal to the scissile bond, suggesting that P1-P4-fixed positional libraries would not be suitable for subsite determinations. Papain-like cysteine proteases, however, are most selective N-terminal to the cleavage site, notably at the P2 position (31). The P1-P4 specificities of CP-2 were determined, revealing a preference for basic residues at P1, notably Lys, and a fairly unusual preference for Pro at P2 (Figure 3). Papain-like enzymes usually prefer bulky, hydrophobic amino acids in this position (31). Cysteine proteases are typically less selective at P3 and P4, and CP-2 preferred Thr and Arg at P3 and displayed less specificity at P4 but prominently cleaved acidic residues. The prime side specificities of APR-1 and MEP-1 were assessed (29) using a peptide library technique to analyze the substrate pool selected by the proteases from a completely random mixture of peptide dodecamers (Figure 3). For APR-1, hydrophobic residues were well tolerated at P1', with the major selections being for Phe, Tyr, Ile and Met, Ala and Ser. Other P1' residues were deselected. There were also subtle preferences for hydrophobic residues at P2' with Val being the most favoured amino acid. No significant selectivity was seen beyond P2'. MEP-1 was most selective at the P1' position, where it excluded charged residues as well as Pro and Trp. Though the protease tolerated a wide spectrum of amino acids at P1', it displayed a significant preference for Met, Ile, Phe and Ala at that position. MEP-1 was less selective at positions downstream of P1', but subtle preferences, for example for basic residues at the P2' and P3' positions, were apparent from the digestion profile. No significant selectivity was seen at positions downstream of the P4' position (data not shown). Furthermore, MEP-1 did not digest Hb when combined with just APR-1 or just CP-2 alone (not shown). However, MEP-1 did digest globin fragments after initial proteolysis with both APR-1 and CP-2 together ( Figure 5). By considering previously reported cleavage sites for APR-1 (20) and additional sites identified here, as well as cleavage sites for CP-2 and MEP-1, we developed a map of Hb catalysis by hookworm proteases (Figure 6). For APR-1, we detected 38 and 44 cleavage sites in the α and β chains of Hb respectively. In agreement with subsite preference data using combinatorial libraries, APR-1 was relatively promiscuous but favored hydrophobic residues at P1'. Peptides that were generated by cleavage of Hb with APR-1 were compared with Hb peptides digested with both APR-1 and CP-2; CP-2 cleaved the α chain at 15 and the β chain at 16 sites that were not cleaved by APR-1. In contrast to the subsite preference data obtained for CP-2 using synthetic combinatorial libraries, basic residues in Hb were not favored at the P1 subsite -instead, there was little selectivity, with P1-Ser, -Leu and -Asp being identified 3 times each. Only one basic P1 residue, βArg-30, was identified from CP-2-digested Hb hydrolysates. By library scanning, CP-2 was predicted to favor Pro, and to a lesser extent, the hydrophobic residues Ile, Val and norleucine at the P2 site. Hb P2 subsites favored by CP-2 included Thr, Ile and Leu; surprisingly, P2-Pro was not identified. It should be noted however, that not all cleavage sites would be detected using the methods employed here; moreover, secondary structure of Hb may mask some sites predicted to be ideal, especially if they contain proline. MEP-1 cleaved globin fragments from the α chain at 10 sites and the β chain at 8 sites. Using prime side peptide libraries, MEP-1 favored largely hydrophobic P1' residues; in agreement with this data, preferred P1' residues in degraded globin (after digestion with APR-1 and CP-2) for MEP-1 were Leu and Ala. It was more difficult to predict the preferred P2' residues because many of these sites contained cleavage sites for the other two enzymes (Figure 6), both of which are thought to act upstream of MEP-1 in the hemoglobinolysis cascade described here. Preferred P1 residues for MEP-1 in Hb were Asp and Leu. MEP-1 digested peptides that contained between one and six residues N-terminal to the scissile bond, and between one and four C-terminal residues. It cannot be discounted however, that peptides unique to the final Hb digest (cleaved with all three proteases) were generated by secondary catalysis with APR-1 or CP-2 after MEP-1 had begun to act.

DISCUSSION
We recently described the first nematode recombinant proteases to digest Hb -the cathepsin D-like enzymes, Ac-APR-1 from A. caninum and Na-APR-1 from the human hookworm, N.
americanus, (20). We now show that two additional enzymes from the intestinal microvilli, a cathepsin B-like cysteine protease, CP-2 (33) and a neprilysin-like metalloprotease, MEP-1 (25), also digest Hb in a semi-ordered pathway of catalysis. APR-1 and CP-2 digest native and denatured Hb while MEP-1 acts downstream of these two enzymes and digests globin peptides generated by prior proteolysis with APR-1 and CP-2.
The cascade described here is remarkably similar to that used by P. falciparum to digest Hb in the food vacuole of the parasite (2,3,7). Two aspartic proteases, plasmepsins I and II have been implicated in the initial steps of Hb degradation in Plasmodium. Plasmepsin I, like hookworm APR-1, cleaves the Hb tetramer at the hinge region, unraveling the molecule for subsequent enzymatic proteolysis (1). Plasmepsin II is capable of digesting native Hb, but is more active against denatured or fragmented globin, such as that produced by the action of plasmepsin I (34). In addition to Na-APR-1, we identified a second aspartic protease from the human hookworm, N. americanus, Na-APR-2; APR-2 was most similar to a family of nematode-specific aspartic proteases, termed nemepsins (19). APR-2 digested Hb but at very different sites to APR-1, suggesting that in N. americanus at least, multiple aspartic proteases are responsible for digesting Hb in the intestine of adult worms. In like fashion, P.
falciparum digestive vacuoles contain at least four distinct aspartic proteases (35). After extensive in vitro and in silico screening of A. caninum cDNA libraries and ESTs respectively, we did not identify an orthologue of Na-APR-2 from A. caninum (B. Zhan, A. Loukas and P. Hotez, unpublished).
While the reducing environment of the digestive vacuole is an unresolved issue, erythrocytes contain 1.2mM GSH and 0.006 mM cysteine (32). The reductant concentration of the hookworm intestine is unknown but given the large volume of blood that the worms ingest and subsequently lyse (17,38), we have assumed that the concentration is similar to that found in blood. Ac-CP-2, like other papain family enzymes, was dependent on a reducing agent for activity. Physiological concentrations of GSH (at least intraerythrocytic) were sufficient for CP-2 to digest Hb but were not too high to inhibit APR-1 from digesting Hb, albeit at a slower rate.
Cysteine proteases exert much of their substrate selectivity at the S2 position of the active site pocket. CP-2 displayed an unusual, but not unprecedented, preference for Pro at P2 for a clan CA protease (Merops classification). Papain-like enzymes with similar P2-Pro specificities have been reported from other parasites including trypanosomes (39) and liver flukes (40) as well as mammalian cathepsin K (41) and ginger rhizome cysteine proteases (42). Although CP-2 readily digested P2-Pro from combinatorial peptide libraries, this substrate preference was not detected in Hb hydrolysates, despite the presence of multiple prolines in canine Hb α and β chains. However, many of the Pro residues in Hb were one or two residues either side (both prime and nonprime) of the scissile bonds at APR-1 cleavage sites (Figure 6), possibly accounting for the absence of P2-Pro sites in Hb hydrolysates detected here.
Ac-MEP-1, like falcilysin from P. falciparum, did not digest Hb or heat-denatured globin, but did further digest globin fragments generated by initial cleavage with aspartic and cysteine proteases, implying a downstream role for this enzyme in the hemoglobinolysis cascade. The sheep barber's pole worm, Haemonchus contortus, expresses a family of metalloproteases in the intestine (43,44), and although these enzymes have not been expressed in active form, it is thought that they are involved in Hb digestion. MEP-1 displayed a broad pH range when assessed in zymograms. While gelatinolytic activity is relatively crude and non-quantitative, the strongest regions of hydrolysis were observed at two distinct pH values with peaks at pH 2.0, no activity at pH 3.0 and a second peak of activity at pH 5.0-7.0. Falcilysin was recently shown to be expressed in both the food vacuole and within vesicular structures elsewhere in P. falciparum; moreover, the enzyme was shown to have dual specificity, displaying globinolytic activity at acidic pH and a distinctly different substrate preference at neutral pH, prompting the authors to suggest that falcilysin functioned as two different proteases in different anatomic locations within the parasite (45). Although dual peaks in gelatinolytic activity were seen for MEP-1, the protease was only detected in the intestine of adult worms, suggesting a single function in Hb digestion. Further investigations are warranted to determine the substrate preferences of MEP-1 at different pH values before a potential dual role beyond hemoglobinolysis can be entertained for this protease.
Hydrostatic pressure of the nematode pseudocoelom ensures that blood ingested by adult hookworms has a rapid passage time through the intestine and out of the anus (38). It is therefore important for the parasite to quickly lyse ingested erythrocytes (17)  While an ordered pathway of hemoglobinolysis has not yet been determined for the trematode parasite, Schistosoma mansoni, intestinal aspartic and cysteine proteases have been cloned and expressed and shown to digest Hb in isolation (48)(49)(50)(51). Hookworms, schistosomes and plasmodia are all endoparasites, with the latter being intra-erythrocytic during its bloodfeeding periods. Interestingly, exoparasites that feed on blood, namely biting insects and acarines, rely predominantly on serine proteases to digest the blood meal (52)(53)(54).
Nevertheless, gut specific expression and secretion of cysteine proteases, similar to that seen in the free-living nematode, Caenorhabditis elegans and parasitic helminths, occurs in many insects. Moreover, feeding by these insects on plants genetically engineered to produce cysteine protease inhibitors, results in a switch to serine proteases as digestive enzymes suggesting the two protease families are redundant in insects at least (55).
Hemoglobinases are proving to be excellent targets for vaccines against blood-feeding nematodes (18,56). H-gal-GP is a protein complex from the intestine of the sheep barber's pole worm, Haemonchus contortus. Vaccination of sheep with H-gal-GP confers high levels of protection against parasite challenge; the complex is predominantly made up of aspartic, cysteine and metalloproteases (57), and recent evidence suggests that the purified enzymes from H-gal-GP are responsible for the vaccine efficacy (44,58,59). We recently described the efficacy of Ac-CP-2 as an anti-hookworm vaccine in dogs experimentally challenged with A.
caninum (21). Vaccinated animals produced antibodies that neutralized the proteolytic activity of the recombinant enzyme and resulted in decreased growth and fecundity of female worms. Moreover, antibodies from vaccinated dogs were found adhered to the intestines of worms recovered from those dogs. It is likely that the protective mechanism afforded to animals vaccinated with hemoglobinases is the generation of neutralizing antibodies that interrupt digestion of the blood-meal, affecting adult worm growth, survival and fecundity (a blood meal is required to produce eggs). Elucidation of the molecular mechanisms by which hematophagous worms digest blood will hopefully lead to the production of new generation control strategies. Perhaps hemoglobinases will form part of a cocktail vaccine against hookworms. By targeting potential antigens from hookworm infective larvae and adult blood-feeders, such as the hemoglobinases described here, recombinant vaccines against these complex multicellular pathogens are a realistic near-term goal (12).          Hb-α