A Cleavage-potentiated Fragment of Tear Lacritin Is Bactericidal*

Background: The wet visual surface of the eye is essentially a sterile environment. Results: Proteolytic processing of the prosecretory mitogen lacritin in tears releases a fragment that is required for much of the bactericidal activity of tears. Conclusion: The protease-released C terminus of lacritin is bactericidal under physiological conditions. Significance: All known lacritin activities are bundled within the same C-terminal region, although at different dose optimum. Antimicrobial peptides are important as the first line of innate defense, through their tendency to disrupt bacterial membranes or intracellular pathways and potentially as the next generation of antibiotics. How they protect wet epithelia is not entirely clear, with most individually inactive under physiological conditions and many preferentially targeting Gram-positive bacteria. Tears covering the surface of the eye are bactericidal for Gram-positive and -negative bacteria. Here we narrow much of the bactericidal activity to a latent C-terminal fragment in the prosecretory mitogen lacritin and report that the mechanism combines membrane permeabilization with rapid metabolic changes, including reduced levels of dephosphocoenzyme A, spermidine, putrescine, and phosphatidylethanolamines and elevated alanine, leucine, phenylalanine, tryptophan, proline, glycine, lysine, serine, glutamate, cadaverine, and pyrophosphate. Thus, death by metabolic stress parallels cellular attempts to survive. Cleavage-dependent appearance of the C-terminal cationic amphipathic α-helix is inducible within hours by Staphylococcus epidermidis and slowly by another mechanism, in a chymotrypsin- or leupeptin protease-inhibitable manner. Although bactericidal at low micromolar levels, within a biphasic 1–10 nm dose optimum, the same domain is mitogenic and cytoprotective for epithelia via a syndecan-1 targeting mechanism dependent on heparanase. Thus, the C terminus of lacritin is multifunctional by dose and proteolytic processing and appears to play a key role in the innate protection of the eye, with wider potential benefit elsewhere as lacritin flows from exocrine secretory cells.

2). Although Ͼ1900 antibacterial peptides have been identified to date, most (3)(4)(5) are individually inactive under normal physiological conditions (6,7), with electrostatic binding of the anionic bacterial outer membrane a common characteristic. Yet others are primarily hydrophobic. Although membrane disruption is typical with ensuing lysis or pore formation (8), some pass intracellularly to disrupt function (9). Antimicrobial peptides may represent the future of antibiotics, with sensitivity to proteolysis in the gut being a primary weakness.
Different epithelia have evolved distinct protective mechanisms. The surface epithelium of the eye lacks the enhanced cornified barrier of skin, yet is rarely subject to bacterial penetration (7,10). This property is largely attributable to the bactericidal tear film that covers the surface of the eye with lipids (11), metabolites (12), salts, and at least 1500 different proteins, some only recently characterized (13). Originally, it was thought that tears only immobilized pathogens by salt agglutination for subsequent removal or that salt levels were responsible for tear bactericidal activity because heat was ineffective (14) or, instead, that a moderate heat-resistant activity could be designated as a lysozyme (15). Since then, a variety of antimicrobial factors have been identified in human tear film, including lactoferrin, immunoglobulin A antibodies (IgA), secretory phospholipase A 2 , mucins, ␣and ␤-defensins, histatins, and cathelicidin LL-37 (7). Gene knock-out studies support antimicrobial roles for lipocalin 2 (16 -18), cathelicidin antimicrobial peptide (LL37 (19)), and defensin ␤1 (20). However, all except lipocalin 2 are individually inactive under physiological conditions (6).
Lacritin is a pleiotropic tear protein (21) that promotes the survival of stressed human corneal epithelial cells (22), basal tearing (23), and corneal epithelial cell proliferation (24). Lacritin is 21% identical (25) with dermcidin, whose cathepsin D-releasable C-terminal domain is processed (26) to the potently bactericidal SSL-25 peptide, the main bactericidal activity in sweat (27). These observations were the rationale for challenging cultures of Escherichia coli, Pseudomonas aeruginosa, and Staphylococcus epidermidis with lacritin and lacritin truncation mutants, peptides, and fragments. We report that cleavage of lacritin releases a potent bactericidal fragment that is distinct from SSL-25 and is active on both Gram-negative and -positive bacteria when applied at low micromolar doses. Activity is retained in 280 mosmol/ liter buffer and only slightly diminished at 380 mosmol/liter. Thus, a growth factor with a biphasic dose optimum of 1-10 nM is a potent bactericide at low micromolar levels after proteolytic processing.

EXPERIMENTAL PROCEDURES
Tears and Tear Immunodepletion-Normal human basal tears were collected as described previously (28). The institutional review board at Walter Reed Army Medical Center Department of Clinical Investigation granted approval prior to the initiation of the study. Each participant gave informed consent, and all research adhered to the tenets of the Declaration of Helsinki. Briefly, tears from 0.5% proparacaineanesthetized eyes were collected onto preweighed wicks and flash-frozen for Ϫ70°C storage. Tears were eluted by immersion of each strip in 30 l of PBS for 20 min, followed by centrifugation. For immunodepletion, 10-fold diluted tears were incubated overnight (4°C) or for 1 h at room temperature with protein A beads (0.2 ml, NAb Spin Kit, Peirce/Thermo Scientific) saturated with "anti-N-65 Lac C-term" or preimmune Ig. N-65 is a lacritin truncation mutant lacking 65 N-terminal amino acids (22). The tear flow-through after centrifugation (5000 ϫ g for 1 min) was then assayed for antibacterial activity.
The nature of the lacritin ϳ9-kDa fragment was pursued by Western blotting. Briefly, lacritin before and after DEAE separation was separated by SDS-PAGE and then transferred and blotted with anti-Pep Lac N-terminal and anti-N-65 Lac C-terminal antibodies, respectively diluted 1:200 or 1:400 in PBS containing 0.3% Tween 20. Detection was with ECL. For fragment purification, chitin-enriched lacritin was dialyzed against phosphate buffer containing 14 mM NaCl (pH 7.2). Following incubation with DEAE equilibrated in the same buffer, the ϳ9-kDa fragment was collected in the flow-through, whereas intact (18 kDa) lacritin was eluted with 140 mM NaCl in phosphate buffer, pH 7.2. After determination of protein concentration (BCA assay), both were aliquoted, lyophilized, and stored at Ϫ70°C. Analysis was by SDS-PAGE on 4 -20% gradient gels. The identity of the ϳ9-kDa fragment was determined by mass spectrometry.
Bacterial Growth, SYTOX Green Assays, and on Column Cleavage-E. coli (ATCC (Manassas VA) catalog no. 10536), S. epidermidis (ATCC catalog no. 12228), and P. aeruginosa (ATCC catalog no. 9027) were grown to mid-log phase in 50 ml of Luria-Bertani (LB) medium and washed three times in phosphate buffer containing 10 mM NaCl (pH 7.2; PB-NaCl) with centrifugation. Pellets were resuspended in 1 ml of PB-NaCl.
For lacritin inhibition assays, 50 l of bacterial pellets each diluted 1:100 in PB-NaCl were incubated for 1.5 h (37°C) with 100 l of lacritin, lacritin truncations, or synthetic peptides at a final concentration of 0.1-6 M. Mixtures were diluted 1:10 in PB-NaCl before plating 100 l in quadruplicate on LB agar plates for overnight growth at 37°C. Colonies were manually counted. In other experiments, mid-log E. coli was treated at 37°C for 0, 1, 2, or 3 h with 2 M lacritin or lacritin truncations or with ampicillin (5 M) or tetracycline (2 M). After each treatment, 100 l was centrifuged, resuspended in 1 ml of PB-NaCl, and plated (100 l) onto LB agar for overnight growth (37°C) and colony counting.
For salt sensitivity studies, pelleted and washed mid-log phase E. coli, S. epidermidis, or P. aeruginosa were resuspended in 1 ml of PB-NaCl and then treated as above with PB-NaCl or with 3 M N-65 in 130, 280, or 380 mosmol/liter PB-NaCl for 1.5 h (37°C). Mixtures were diluted 1:10 in PB-NaCl before plating 100 l of each in quadruplicate on LB agar plates for overnight growth at 37°C. Colonies were manually counted.
For bacterial permeability assays, pelleted and washed midlog phase E. coli were resuspended in 1 ml of PB-NaCl and then treated as above with 3 M lacritin, N-65, or C-25 or with 10% Triton X-100. Similarly, washed mid-log phase S. epidermidis were resuspended in 1 ml of PB-NaCl and then treated with lacritin or C-25 or a ϳ9-kDa purified lacritin fragment. Later, 1 l of 0.5 mM SYTOX Green was added to each well of 96-well fluorescent microtiter plates. Readings were taken at 5-min intervals at respective excitation and emission wavelengths of 485 and 538 nm using a Fluoroskan Ascent FL fluorometer (Thermo Fisher Scientific). In parallel, SYTOX Green internalization was visualized by confocal microscopy after 1 h of 10% Triton X-100, PB-NaCl, or 3 M N-65 treatment of washed mid-log phase E. coli.
For cell-free synthesis without glycosylation, full-length lacritin cDNA in pLacSL was PCR-amplified and subcloned into pTXB1 supplied by the manufacturer (New England Biolabs, Ipswich, MA). Cell-free synthesis and subsequent removal of ribosomes, followed by metal affinity resin adsorption of His-tagged factors, was performed as per the manufacturer's instructions (New England Biolabs; PURExpress). Immediately following expression, an aliquot was stored at Ϫ60°C. Other aliquots were incubated at 37°C for 24 and 48 h. Each was separated by SDS-PAGE, transferred, and blotted with anti-N-65 Lac C-terminal antibodies.
For lacritin cleavage assays, supernatants from saturated 50-ml overnight cultures of S. epidermidis were collected by centrifugation (10 min; 11,000 rpm). Each supernatant was then incubated for 4, 16, and 20 h (37°C) in PB-NaCl with chitin beads containing lacritin-intein immobilized via N termini. C-terminal cleavage products were collected by PB- Statistical Analyses-With the exception of the single metabolomic analysis, all experiments were performed at least FIGURE 1. Tear bactericidal activity is largely attributable to lacritin, as suggested by immunodepletion using an anti-lacritin C-terminal antibody. A, line diagram of lacritin, drawn together with dermcidin, with which it is 20% identical. In brackets is the N-65 truncation mutant. Bracketed in dermcidin is the anti-bacterial SSL-25 fragment. Rectangles, PSIPRED (version 3.0)-predicted (or validated) (24, 29) ␣-helices. B, schematic diagram of the eye with tear film. C, washed mid-log E. coli were incubated with phosphate (phos) buffer containing 10 mM NaCl (PB-NaCl) without or with half-diluted human basal tears for 1.5 h (37°C). The mixture was then diluted and transferred to agar plates for overnight growth (37°C) and cfu counting. Tears completely inhibit growth. D, 10-fold diluted tears were incubated with immobilized "anti-N-65 Lac C-term" or preimmune Ig. Material not bound (lanes 2 (ab C-term) and 4 (pre-immune)) and starting tears (lanes 1 and 3) were separated by SDS-PAGE, transferred, and blotted for lacritin. E, washed mid-log E. coli were incubated for 1.5 h (37°C) with phosphate (phos) buffer containing increasing volumes of tear flow-through from anti-N-65 Lac C-terminal or preimmune Ig columns. The mixture was then diluted and transferred to agar plates for overnight growth (37°C) and cfu counting. F, same as E but with washed mid-log P. aeruginosa. three times. Statistical analysis of metabolite data was as described previously (22), where raw data values were first log transformed to be closely distributed as a normal distribution and then assessed by a non-parametric Wilcoxon test and two-sample t test. For both tests with p Ͻ 0.05, metabolites were considered significantly different and further analyzed by hierarchical clustering for their association patterns. Data are reported as the mean Ϯ S.E.

RESULTS
Lacritin Bactericidal Activity in Tears-Tears protect the surface of the eye against environmental pathogens and are enriched in the prosecretory mitogen lacritin (Fig. 1A), which flows onto the eye during basal and reflex tearing (21,31). Lacritin is 21% identical to dermcidin, whose proteolytically processed C terminus contributes to the bactericidal activity of human sweat (26,27,32). We first sought to validate whether basal human tears (33)(34)(35) (Fig. 1B), like reflex tears (36,37), are bactericidal and, if so, whether lacritin or a lacritin fragment(s) is in part responsible. Indeed, half-diluted basal tears completely blocked E. coli growth (Fig. 1C). E. coli is a significant contributor to bacterial conjunctivitis in the developing world, as is P. aeruginosa (38,39). We next tested tears that had been passed over immobilized anti-N-65 Lac C-terminal antibodies (ab C-term) to immunodeplete both lacritin and C-terminal lacritin fragments (Fig. 1D, lane 2), or over preimmune Ig (mock-depleted; Fig. 1D, lane 4). Both were diluted 10-fold for dose-dependent challenge of E. coli and P. aeruginosa. Mock-depleted tears suppressed E. coli and P. aeruginosa colonies in a tear volume-dependent manner (Fig. 1, E and  F). This contrasted with C-terminal antibody-immunodepleted tears, which were as ineffective as the phosphate buffer negative control (Fig. 1, E and F).
Lacritin's C Terminus Contains a Bactericidal Domain-Lacritin's C terminus contains three predicted ␣-helices (Fig.  1A), each validated by circular dichroism (24,29) (Fig. 2G). The most C-terminal ␣-helix is amphipathic and targets syndecan-1 as an initiator of corneal epithelial cell proliferation (24) and survival (22), largely via hydrophobic face residues (29). Association of amphipathic ␣-helices with bacterial membranes can be destabilizing (40). To explore whether these or other lacritin domains are bactericidal, we generated recombinant lacritin and lacritin truncations (24) (Fig. 2A). Each was generated as an intein fusion protein, purified on chitin to also remove the intein tag and then on DEAE to exclude bacterial contaminants. Lacritin and truncations were then assayed in equimolar (2 M) amounts in the presence of mid-log E. coli, P. aeruginosa, or S. epidermidis. P. aeruginosa is an eye pathogen often responsible for keratitis in contact lens wear (41). S. epidermidis is a common cause of conjunctivitis and keratitis and is abundant in blepharitis (42,43), an eyelid inflammation associated with slightly altered tear composition, including selectively less lacritin (44). Lacritin without truncation had no effect on the appearance of colonies, with numbers equivalent to the phosphate buffer negative control (Fig. 2, B-D). However, few colonies were apparent with lacritin lacking 65 (N-65) or 80 (N-80) amino acids from the N terminus, an effect completely or partly negated by removing six additional amino acids (N-86) in E. coli (Fig. 2B) or P. aeruginosa (Fig. 2C) but not S. epidermidis (Fig.  2D). Amino acids 81-86 comprise the sequence LAKAGKG, which ClustalW2 (Fig. 2E) and FASTA align with LDGAKKA from potent dermcidin fragment SSL-25 with an amino acid identity of 44%.
Antimicrobial Mechanism-To ask whether N-65 was destabilizing the outer bacterial membrane such that small extracellular molecules were gaining entry, we challenged mid-log E. coli with it in the presence of the membrane-impermeable dye SYTOX Green (45). After entry, SYTOX Green binds nucleic acids. We also monitored the release of hemoglobin from sheep red blood cells to control for lysis of mammalian cells under identical incubation conditions. N-65 (Fig. 3, A and B), but not C-25 or lacritin (Fig. 3A), opened E. coli to SYTOX Green with kinetics similar to the 10% Triton X-100 positive control (Fig. 3, A and B). None, including N-104, lysed sheep red blood cells (Fig. 3, C and D). We wondered whether the interaction might involve a cell surface protein(s) because the same region binds a hydrophobic patch within the eukaryotic cell surface proteoglycan syndecan-1 (46). E. coli were biotinylated, lysed, and incubated with immobilized lacritin or C-25. After washing and then exposure to 1 M NaCl, no streptavidin-peroxidase-detectable bands were eluted (not shown), suggesting a lack of high affinity protein binding. N-65 may bind LPS or peptidoglycan. However, binding via LPS would be incompatible with targeting Gram-positive bacteria. Conversely, peptidogly-can is the main component of the cell wall of Gram-positive bacteria, but it is primarily periplasmic in Gram-negative bacteria.
An alternative possibility is that the interaction might be largely electrostatic between the cationic C terminus and anionic phospholipids that predominate in bacterial membranes. However, electrostatic binding renders cationic antimicrobial proteins, such as cathelicidins and most defensins, inactive in physiological solutions (3)(4)(5). Tear osmolarity is ϳ302 (47) to 318 mosmol/liter (48), rising to ϳ324 mosmol/liter or more in tears from dry eye patients (49). We therefore challenged all three bacteria with 2 M N-65 in 130, 280, or 380 mosmol/liter (pH 7.2) buffer (Fig. 3, E-G). N-65 was effective against P. aeruginosa (Fig. 3F) and S. epidermidis (Fig. 3G) in all three buffers and against E. coli at 130 and 280 mosmol/liter ( Fig. 3E; 380 mosmol/liter buffer alone was toxic for E. coli). Thus, lacritin C-terminal bactericidal activity is largely insensitive to osmolarity. Electrostatic capture of bacteria by N-65 may thus be substantially strengthened by mutual hydrophobic interaction, should hydrophobic residues penetrate the membrane possibly to form pores, as implied by SYTOX Green studies.
Although compromised membrane integrity may be sufficient to promote death, additional mechanisms could be in play either prior to or as a consequence of putative pore formation. We monitored 80 E. coli metabolites 15 min after treatment without or with 6 M N-65 and observed significant changes in 34 (Fig. 4). Notably, N-65-treated cells dis- played less dephosphocoenzyme A, spermidine, putrescine, and phosphatidylethanolamines. Dephosphocoenzyme A is the immediate precursor of widely employed cofactor coenzyme A (50). Spermidine and putrescine counter damage from reactive oxygen species (51), and phosphatidylethanolamine is the primary bacterial membrane phospholipid (52) that proportionally decreases in osmotic stress and is necessary for transport protein function (53). Thus, N-65 appears to widely compromise cellular metabolic capacity, protection against reactive oxygen species that are elevated by antibiotics, and perhaps integral transmembrane transport processes. We also observed an accumulation of alanine, leucine, phenylalanine, tryptophan, proline, glycine, lysine, serine, and glutamate (Fig. 4) and a decrease of valine, as also occurs in cold-or heat-stressed E. coli (54). Other changes included more cadaverine and pyrophosphate. Generation of cadaverine from lysine decarboxylation is a mechanism by which E. coli counters acid stress (55), and pyrophosphate is beneficial to bacterial growth (56). Thus, stress from N-65 disruption of the bacterial membrane and the availability of dephosphocoenzyme A, spermidine, putrescine, and phosphatidylethanolamines appears to go hand in hand with mechanisms attempting to counteract it.
Proteolysis of Lacritin-If this bactericidal mechanism is available in basal tears, as suggested by lacritin immunodepletion experiments (Fig. 1, E and F), C-terminal fragments must be available at a sufficiently high micromolar level. Levels may be enhanced by bacteria-dependent cleavage, for which there is plenty of lacritin available. Basal human tears appear to contain ϳ27 M lacritin antigen, as implied by an ELISA estimate of 4.2 ng of lacritin per 100 ng of total tear protein (28) and a suggested basal tear protein concentration of ϳ8 mg/ml (57). We therefore searched for evidence of lacritin C-terminal processing in normal tears and in manipulations of recombinant lacritin without or with bacterial supernatant. Monomer is ϳ25 kDa, with dimer and trimer at 50 and 75 kDa, respectively. Blotting of normal basal tears with anti-C-terminal lacritin antibodies detected all three, as well as immunoreactive bands of ϳ12, 10, and 9 kDa, suggesting that C-terminal fragments are natural constituents of basal tears (Fig. 5A). Quantities of these fragments appear to be relatively low. To ask whether lacritin is subject to bacteria-dependent cleavage, we incubated recombinant lacritin (ϳ18 kDa) with S. epidermidis supernatant. This promoted the appearance of ϳ17, 14, 9, 8.5, and 6 kDa bands within 4 h at 37°C that resolved to a single ϳ6 kDa (Fig. 5B) or ϳ8.7 kDa (Fig. 5C) anti-C-terminal detectable band by 18 -20 h. Supernatants from other species were tested. S. aureus supernatant also gave rise to a ϳ8.7 kDa anti-C-terminal detectable band (Fig. 5C). P. aeruginosa and E. coli supernatants were less effective over 18 h, with monomer largely left intact but with some ϳ8.7-kDa fragment detectable (not shown). Purification of recombinant lacritin has often yielded a second ϳ9 kDa band after DEAE purification (Fig. 5D) that promotes entry of SYTOX Green (Fig. 5E); is detectable with anti-C-terminal, but not anti-N-terminal, specific antibodies (Fig. 5F); and was validated as a lacritin C-terminal fragment by mass spectrometry (Fig. 5G). Its migration in SDS-PAGE is similar to that of N-80 (Fig. 5H). The same band also slowly developed with time of lacritin incubation alone at 37°C (Fig. 6A), suggesting that an E. coli proteolytic enzyme co-purifies with lacritin, as is not uncommon with recombinant proteins. It is unlikely but possible that lacritin has self-cleavage activity (58). To consider this possibility, we generated lacritin using a cell-free translation system. Lacritin was detectable initially as a doublet (Fig. 6B), which decreased substantially by 24 h at 37°C and was barely apparent at 48 h. No C-terminal fragment was detected (Fig. 6B). To address the nature of the C-terminal fragment generating proteolytic activity, we subjected E. coli recombinant lacritin to a panel of proteolytic inhibitors and discovered that processing was inhibitable with chymostatin, with leupeptin, or by boiling but not with 4-(2-aminoethyl) benzenesulfonyl fluoride hydrochloride, antipain, bestatin, E64, pepstatin, or phosphoramidon (Fig.  6C). Thus, a serine protease-like activity may be responsible.

DISCUSSION
Antimicrobial peptides protect all classes of life as the first line of innate defense. The surface epithelium of the eye lacks the enhanced cornified barrier of skin and yet is rarely subject to bacterial penetration (7, 10), a property largely attributed to the bactericidal tear film. Tears are enriched in the prosecretory mitogen lacritin (21) that flows onto the eye during basal and reflex tearing. Here we discover that lacritin is subject to C-terminal proteolytic processing and that the amphipathic ␣-helixcontaining fragment appears to account for much of the bactericidal activity of normal basal tears by creating pores without hemolysis and a rapid form of bacterial death that may be regulated.
Our rationale for exploring whether lacritin might be bactericidal was its 21% identity with dermcidin (Figs. 1A and 2E), whose proteolytically processed C terminus contributes to the bactericidal activity of human sweat (26,27,31) and is in tears (13). Tear bactericidal activity has been the subject of much curiosity for over a century (7), including the original discovery of lysozyme (15) and a variety of other tear antimicrobial factors, particularly lactoferrin, lipocalin 2, immunoglobulin A antibodies (IgA), secretory phospholipase A 2 , mucins, ␣and ␤-defensins (20), histatins, and cathelicidin LL-37 (7,19,37), and recently cytokeratin fragments (59). However, few are individually insensitive to the ionic strength (6) of normal or dry eye tears. This might be overcome in combination (7), or there could be other contributors. Seeking clarity, we took advantage of a tear immunodepletion strategy (22) with anti-C lacritin terminal specific antibodies, thereby providing confirmation of a C-terminal bactericidal lacritin fragment resident in normal tears that is salt-resistant.
Surprisingly, dermcidin primary sequence homology was not the source of lacritin activity. Only 40.7% identity is shared between dermcidin's bactericidal SSL-25 peptide and the homologous lacritin region that as a synthetic peptide was inactive. Instead, lacritin N-104 fragment with 7% dermcidin identity embodies the core activity, a hybrid domain consisting of an N-terminal amphipathic ␣-helix and hydrophobic C-terminal coiled coil tail, together appropriate for bacterial membrane contact and insertion, as was apparent by rapid entry of membrane-impermeable SYTOX Green in N-65-treated cells. Yet, dermcidin's SSL-25 peptide also forms an amphipathic ␣-helix (60) and, together with an adjoining C-terminal ␣-helix (60), binds artificial 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-(1-rac-glycerol)-rich membranes to form Zn 2ϩ -dependent toroidal pores (61). If N-104 or -65 create pores, as per SYTOX Green entry and metabolomic changes incompatible with lysis, how is pore formation linked to death? The answer is not clear, although levels of some cellular elements fell, such as phosphatidylethanolamine, the primary bacterial membrane phospholipid (52) necessary for transport protein function (53). Also lower were spermidine and putrescine to potentially diminish cellular repair from reactive oxygen species damage. Not yet addressed is the question of whether lacritin or N-104 affects redox potential. Dephosphocoenzyme A, the immediate precursor of essential cofactor coenzyme A necessary for key metabolic processes (50), was also lower. Spermidine and putrescine are notably essential for virulence in Salmonella gallinarum (62) and in P. aeruginosa, where spermidine is necessary for expression of the type III secretion system that delivers virulence proteins into eukaryotic cells (63). This differs from S. aureus, which cannot produce spermine or spermidine or the precursors agmatine and putrescine (64). Guanosine was lower. Synthesis of guanosine penta-or tetraphosphate is necessary for the bacterial "stringent response" to nutritional stress (65). Other metabolites rose, some apparently in a failed attempt to restore homeostasis. Examples include the increase of glycine, cadaverine, and pyrophosphate. Cellular importation of dimethylglycine protects Bacillus subtilis from osmotic stress (66). Cadaverine is the proton consuming product of lysine decarboxylase (55). The exogenous addition of pyrophosphate increases catabolic and anabolic processes for bacterial growth (56). Thus, cleavage of lacritin, like dermcidin, releases an amphipathic ␣-helical bactericidal fragment with poorly conserved primary sequence that nonetheless promotes metabolic stress leading to rapid death.
That this cleavage may be regulated by a serine protease, as per proteolytic inhibitor sensitivity, is in keeping with in silico analysis by the Protease Specificity Prediction Server (67), which predicts serine protease sensitivity after phenylalanine 96, leaving intact the complete C-terminal amphipathic ␣-helix. Dermcidin and cathelicidin are primarily processed by cathepsin D (26) and proteinase 3 (68), respectively. Whether lacritin is partially cleaved within tears as a consequence of or as an innate defense mechanism against infection or, alternatively, is in part processed intracellularly is not known. Most lacritin in normal human tears is uncleaved in monomeric or polymeric forms, the latter from tissue transglutaminase cross-linking (69), but some C-terminal fragment is resident. Curiously, manufacture in E. coli persistently yields a contaminating C-terminal fragment. Fragment abundance increases with time of affinity tag-purified lacritin at 37°C, suggesting that an E. coli serine protease co-purifies with lacritin or that lacritin has selfcleavage activity, as per the SEA module of MUC1 (58), or generation of the active N-terminal domain from the hedgehog protein (70), among other examples. Loss of detectable cell-free translated lacritin with incubation leaves open the possibility that lacritin may be self-cleaving, yet instability when dilute could be an alternative explanation. That the same C-terminal region also supports lacritin's mitogenic (24) and prosurvival (22) activities, although at a much lower dose optimum, is an interesting example of pleiotropic conservation.