Treatment of Porphyromonas gulae infection and downstream pathology in the aged dog by lysine‐gingipain inhibitor COR388

Abstract COR388, a small‐molecule lysine‐gingipain inhibitor, is currently being investigated in a Phase 2/3 clinical trial for Alzheimer's disease (AD) with exploratory endpoints in periodontal disease. Gingipains are produced by two species of bacteria, Porphyromonas gingivalis and Porphyromonas gulae, typically associated with periodontal disease and systemic infections in humans and dogs, respectively. P. gulae infection in dogs is associated with periodontal disease, which provides a physiologically relevant model to investigate the pharmacology of COR388. In the current study, aged dogs with a natural oral infection of P. gulae and periodontal disease were treated with COR388 by oral administration for up to 90 days to assess lysine‐gingipain target engagement and reduction of bacterial load and downstream pathology. In a 28‐day dose‐response study, COR388 inhibited the lysine‐gingipain target and reduced P. gulae load in saliva, buccal cells, and gingival crevicular fluid. The lowest effective dose was continued for 90 days and was efficacious in continuous reduction of bacterial load and downstream periodontal disease pathology. In a separate histology study, dog brain tissue showed evidence of P. gulae DNA and neuronal lysine‐gingipain, demonstrating that P. gulae infection is systemic and spreads beyond its oral reservoir, similar to recent observations of P. gingivalis in humans. Together, the pharmacokinetics and pharmacodynamics of COR388 lysine‐gingipain inhibition, along with reduction of bacterial load and periodontal disease in naturally occurring P. gulae infection in the dog, support the use of COR388 in targeting lysine‐gingipain and eliminating P. gingivalis infection in humans.


| INTRODUC TI ON
COR388 is an irreversible active-site inhibitor developed to target lysine-gingipain (Kgp) in the brain of Alzheimer's disease (AD) patients. 1 Kgp is a cysteine protease virulence factor secreted by Porphyromonas gingivalis, a keystone bacterium in the development of periodontal disease. 2 The secretion of gingipain proteases is part of the asaccharolytic metabolism of P. gingivalis, and the gingipains are known to contribute to dysbiosis, immune pathway induction and dysregulation, chronic inflammation, and cellular toxicity. [3][4][5][6][7] While P. gingivalis resides in oral biofluids and tissues, it has also been shown to translocate to other tissues where it is associated with disease pathology including atherosclerosis, 8 cancer, 9 arthritis, 10 and AD. 1,11 P. gingivalis is best known for its pathogenic role in periodontal disease, and periodontal disease is a risk factor for the development of AD, 12,13 indicating that a high bacterial load in the oral cavity may be one of many risk factors, along with age, genetics, bacterial strain virulence, and other variables that may contribute to a chronic brain infection with P. gingivalis. Additionally, ongoing moderate to severe periodontal disease has been shown to result in more rapid worsening of cognitive decline in clinical AD patients. 14 We recently identified gingipains and P. gingivalis in the brains of subjects with AD, and brain gingipain levels positively correlated to severity of AD diagnosis and pathology. 1 Mechanistic studies in wild-type mice demonstrated that oral infection with P. gingivalis results in brain colonization and pathology characteristic of AD, 1,11 and these effects were blocked or reversed by COR388. 1 As noted above, we and others have found that oral infection of mice with P. gingivalis results in translocation to the brain, resulting in neurodegeneration and AD pathology, while other oral bacteria did not translocate. 15 Because dogs are naturally infected with the closely related species, P. gulae, and commonly develop both periodontal disease 16 and canine cognitive dysfunction associated with cerebral neuropathology, including amyloidosis, tau hyperphosphorylation and neuronal loss that resembles AD in humans, 17,18 we aimed to examine the presence of P. gulae and associated gingipain proteases in dogs and investigate the pharmacology of the selective Kgp inhibitor COR388 in these animals.
The prevalence of periodontal disease in dogs increases with age, with 40.8% of dogs 1-4 years old, and 53.6% of dogs 5-8 years old, having evidence of the disease, with prevalence increasing to >85% in dogs older than 8 years of age. 19 It was originally thought that P. gulae was the animal biotype of P. gingivalis, but recent research indicates that P. gulae is also found in humans with periodontal disease, 20 and is able to adhere to and invade human gingival epithelial cells. 21 Importantly, P. gingivalis and P. gulae are the only bacterial species known to produce gingipains, which include the arginine-gingipains RgpA/B in addition to Kgp. 22 Oral biofluid sampling in dogs enabled us to analyze COR388 Kgp target engagement and pharmacologic effects on P. gulae bacterial load at multiple time points and doses. These data were used to understand the pharmacology of COR388 in naturally infected tissues and biofluids of the oral cavity. In oral tissues, systemic COR388 treatment reduced Kgp activity and P. gulae levels over a 28-day period in a dose-dependent manner. Based on these data, the lowest effective dose of COR388 was chosen for administration for 90 days, and this dose demonstrated efficacy in reducing periodontal disease pathology.
In addition, we demonstrate that P. gulae DNA and Kgp antigens are present in the aged dog brain, with similar histopathology to what we identified for P. gingivalis and gingipains in the human AD brain. 1 Thus, the preclinical work with COR388 reported here allowed us to pharmacologically evaluate the drug's ability to inhibit Kgp, reduce P. gulae bacterial burden and disease pathology in a naturally occurring infection in dogs, providing proof-of-concept for targeting Kgp and P. gingivalis in human studies. in the evenings with water provided ad libitum but not within 30 minutes of the start of anesthetic procedures.

| Dog pharmacology model
Gingival crevicular fluid (GCF), subgingival plaque (SGP), buccal cells, saliva, and plasma samples were collected and BANA enzyme testing in the oral cavity was performed throughout the studies.
Procedures for oral biomarker sample collection are described below and were performed under general anesthesia induced with 8 mg kg −1 of propofol (intravenous) to effect. Following intubation, anesthesia was maintained with an isoflurane-oxygen mixture.
Procedures on study Day 1 were performed under sedation with medetomidine (reversed with atipamezole), as full anesthesia was not required for the collections performed on this day.
For evaluation of pharmacokinetics parameters, approximately 2 mL of blood was collected at the indicated time points following test article administration into K 2 EDTA tubes and plasma was isolated for bioanalytical determination of plasma

| BANA enzyme test
The test provided a qualitative assessment of arginine-gingipain protease activity derived from oral bacteria by assessing proteolytic cleavage of N-benzoyl-dL-arginine-2-napthylamide substrate.
Subgingival plaque was collected using a dental scaler from the base of two teeth above the gum line. Plaque samples were applied to a BANA test strip and analyzed with the manufacturer's instructions (BANAMet LLC).
Enzyme activity was then immediately measured in a Synergy 2 microplate reader (Biotek) at 380/460 nm excitation/emission wavelength, read every 1.5 minutes for 30 minutes with incubation and shaking at 37°C. Enzyme activity data were analyzed using Gen5 software (Biotek).

| Kgp enzyme inhibition assay in bacterial
cultures using an activity-based probe 5 × 10 6 CFU in PBS were incubated with COR388 at the concentrations ranging from 0.3 to 1000 nmol L −1 37°C for 30 minutes, followed by the addition of 1 μmol L −1 COR553 Cy5-labeled activity probe to bind remaining active sites and incubated at 37°C for another 30 minutes. The COR553 activity probe was prepared by the copper-catalyzed azide-alkyne cycloaddition reaction between an azide derivative of the irreversible Kgp inhibitor COR553 and an alkyne amide derivative of the Cy5 fluorophore. The probe forms an irreversible covalent bond with a catalytic cysteine residue in the active site of Kgp by displacement of a phenol leaving group, and its use has been previously described in detail. 1 Sample loading dye was added, samples heated at 95°C for 5 minutes, and analyzed on XT Criterion gels (Biorad). Cy5 probe labeled protein was detected using a Chemidoc imaging system (Biorad) and quantified by Imagelab software (Biorad). A rabbit polyclonal anti-Kgp antibody, CAB102, described previously, 1 was used to detect Kgp by Western Blot analysis to ensure equal loading of bacteria and presence of the enzyme. This sample loading was confirmed in each experiment with one example included.

| Oral biofluid sample collection
Saliva: Saliva was collected using a SalivaBio collection swab (Salimetrics LLC) on the mouth interior. The swab was inserted into the saliva collection vial, centrifuged at 1500 g for 15 minutes at room temperature. The swab was then removed from the vial and the liquid sample was stored as aliquots of 50-100 μL at −80°C prior to analysis. Gingival Crevicular Fluid: GCF was collected from the gingival pockets at times following dosing ±30 minutes, as indicated. Eight sampling sites were identified per dog and four sites were used per timepoint. Sampling sites were isolated with a cotton roll and airdried. Gingival fluid collection strips (Periopaper™, Oraflow) were gently held with tweezers and placed into the pocket until resistance was felt. Strips were held in place for 30 seconds and then processed as described above for SGP isolation.

| qPCR detection of P. gulae in saliva, buccal cells, and GCF
Biofluids were evaluated for P. gulae load using a qPCR assay that was developed to specifically detect P. gulae 16S gene with the following primers and probe: Forward primer CGAGGGGCAGCATGAACTTA, Reverse primer TTGCCCGATCATGCAACCAA, and probe GCGTAACGCGTATGCAACTTGCCTTAC. Primers were designed from regions of P. gulae 16S that have sequence mismatches with P. gingivalis sequence to ensure a specific detection of P. gulae. Saliva was evaluated with either 2 μL neat saliva or purified DNA from 2 μL saliva. Buccal cells were evaluated with either 2 μL neat buccal cell slurry or purified DNA from 2 μL buccal cell slurry. SGP was evaluated with 2 μL from the elution in B-Per. GCF was evaluated with either purified DNA from 50 μL of B-Per eluted GCF or neat 2 μL elution.

| Kgp enzyme activity analysis in GCF and SGP using an activity-based probe
Twenty microliters of neat saliva, SGP and GCF eluates and 50 μL lysed buccal cell slurry were incubated with the COR553 activity probe as described above, at a final concentration of 1 μmol L −1 .
Probe binding analysis was performed as described above. BCA protein detection assay results were used to equally load the gels. Active Kgp was extrapolated from the purified Kgp (purified Kgp was a kind gift of Barbara Potempa, University of Louisville) standard curve that was run on each gel. To normalize samples run on separate gels, the extrapolated Kgp value was divided by the ng of total protein loaded per lane since the value varied between gels.

| Dental assessments
Assessments were performed 2 hours following the evening dose time 7 days prior to the start of dosing (predose) and also included on days 27, 28, 29, 91, and 92. These included an examination of pocket depth (listed in mm). Dental assessments were performed after all biomarker samples were collected so as not to disturb biofluids with this assessment.

| Immunohistochemistry analysis of gingipains in dog brain tissue
Immunohistochemistry (IHC) staining was performed on postmortem brain samples from four dogs not utilized in the described pharmacology studies. Kgp was detected using polyclonal antibody generated to a peptide from lysine-gingipain CAB102 as previously described. 1 Antibody specificity was confirmed through IHC staining of bacteriapositive gingival tissues and pre-absorption of the antibody to Kgp eliminated the brain tissue staining in dog brain tissues. 1 Sections were deparaffinized in xylenes and alcohol series and submitted to heat mediated antigen retrieval in citric acid at pH 6.0 (Vector Laboratories, Burlingame CA). Sections were blocked in 5% normal horse serum with 0.3% triton-x-100 in PBS and incubated overnight at 4°C in primary antibody 1:2000 in 2.5% NHS and 0.15% triton-x-100. After washing three times in PBS, Impress HRP polymer detection kit and Impact DAB EqV substrate were used to visualize the primary antibody binding. Images were taken on an Olympus motorized microscope BX61 equipped with a color CCD camera (Olympus, DP27) and processed for brightness and contrast correction, cropping and addition of scale bars with CellSens 1.14 Dimension software.

| PCR detection and sequence analysis of bacteria in aged dog brain tissue
Frozen brain tissue samples from the same postmortem dog samples used for IHC analysis of Kgp protein were also analyzed for the presence of P. gulae bacteria by qPCR followed by sequence analysis of the PCR product. qPCR copy number detected is shown normalized per brain cell number listed knowing the quantity of DNA utilized in the PCR reaction and average DNA per cell.
Hippocampus was dissected from the frozen brain, the weight measured and then homogenized in 1 mL of ice cold RIPA buffer (VWR Life Science RIPA Lysis Buffer) with the Qiagen TissueLyser II at 30 Hz for 2 minutes. Protein lysate (200 μL) was immediately transferred into 1 mL cold Trizol solution (TRI reagent, Millipore).
After removal of the aqueous phase (standard TRI Reagent RNA extraction protocol, Millipore), the DNA in the phenol phase was precipitated with 100% ethanol, centrifuged at 13 500 g, washed again with 75% ethanol, and air dried for 10 minutes. To ensure protein removal, the pellet was solubilized in ATL buffer (Qiagen) and digested over night with Proteinase K. DNA purification was performed according to the instructions of the Qiagen DNeasy Blood &amp; Tissue Kit. DNA was eluted from the Qiagen column in 60 mL of TE buffer. DNA (40 ng) was used for qPCR by the method described above for biomarker qPCR evaluations.

| Data and statistical analysis
IC 50 values were extrapolated using a non-linear fit for the data plotted with log(conc) vs normalized response and the 95% confidence interval (CI) is shown for each data point (average of N = 3 replicates) using GraphPad Prism 7. The normalized response for the IC 50 determination was performed with respect to no inhibitor control that was set to 100% Kgp activity. The curve fit to the data was evaluated by the R 2 value, which was ≥.9 for all the IC 50 curves. All PK graphs (statistical significance in GraphPad Prism 7) and parameters (statistical significance in WinNonlin) are shown as mean with standard deviation (SD). All target engagement data in biomarker samples is presented as the mean with standard error of the mean (SEM) determined using the Unpaired t-test in GraphPad Prism 7. All values presented as % of baseline were calculated by setting the baseline to 100% response and statistical significance is shown compared to baseline. Statistical significance where specific groups on a graph were compared is shown by bracketed horizontal lines. The asterisk convention to denote P value was followed: *P < .05; **P < .01; ***P < .001; ****P < .0001.

| Lysine-gingipains from P. gingivalis and P. gulae have similar proteolytic activity
Kgp from P. gulae and P. gingivalis have similar amino acid sequences and specific activity, suggesting similar functionality on substrates. 22 Using in vitro substrate cleavage assays, we characterized the specific activity of Kgp from two species of P. gingivalis as well as P. gulae. Proteolytic activity using the Z-His-Glu-Lys-MCA peptide substrate was assayed from bacterial cultures to assess activity of bacterial surface-associated Kgp as well as bacterial lysates in a buffer which preserves protease activity. Values for maximal velocity (V max ), Kgp concentration, and specific activity (rate of proteolysis) were similar between two strains of P. gingivalis and P. gulae (Table 1). These data suggest that these Kgp enzymes have similar structure and specific activity and indicate COR388 may have potency across these bacterial species and strains. Note: Protease activity in intact bacterial culture for the three strains listed and separately from lysates prepared from these bacteria was measured as described in Methods. The specific activity was determined by dividing the protease activity V max with the concentration of Kgp identified in the assays.

| COR388 potently inhibits lysine-gingipain from both P. gingivalis and P. gulae
The inhibition of Kgp protease activity from P. gingivalis by COR388 and the kinetics of binding were previously reported. 1 In order to confirm the potency of COR388 on Kgp expressed by P. gulae, three in vitro assays were conducted. Potency was assayed on intact bacterial surface-associated Kgp ( Figure 1A) and on a bacterial lysate ( Figure 1B). Potency of inhibition was single digit nmol L −1 or lower in both assays and similar between strains.
Inhibition of Kgp was also assayed with an activity-based probe binding assay. COR553 is an irreversible active-site Kgp probe labeled with Cy5 for detection, previously described.

| COR388 is orally bioavailable in dogs with moderate metabolic stability
Plasma concentrations of COR388 were measured after oral and intravenous administration using LC-MS/MS methods. The pharmacokinetic profile of oral doses of COR388 administered in gelatin capsules is depicted in Figure 2A and after a single intravenous dose of 1 mg kg −1 COR388 in Figure 2B.

| Oral administration of COR388 for 28 days results in target engagement and sustained reduction in P. gulae bacterial load
Periodontal disease mediated by pathogenic bacteria is naturally occurring in dogs and has a similar disease course and microbiology as that found in humans. 23 SGP was also collected and assayed for P. gulae DNA by qPCR and Kgp activity using the COR553 probe. Kgp activity is significantly reduced with COR388 administration similar to the data in GCF ( Figure 3C). In addition, RgpB protease activity was assayed using a substrate cleavage assay in SGP as described in Methods.
Active RgpB was detected at baseline on day 1 with a reduction following COR388 administration ( Figure 3E). This is consistent with the reduction in P. gulae levels also observed, as levels of all bacterial-produced proteins would likewise be expected to fall concurrent with a reduction in bacteria.
Together these data on multiple readouts of gingipain activity and P. gulae load, from independent oral biofluids harboring the bacteria, are consistent with COR388 target engagement on Kgp and subsequent reduction in bacterial burden.

| Oral administration of COR388 results in dosedependent Kgp target engagement
Kgp active site inhibition as well as a robust anti-bacterial response was observed in the first study. A second study in an independent cohort of aged dogs was performed to assess the dose-response relationship of COR388 including a vehicle control. In this second study, after the first dose was included to assess biomarker activity at the plasma pharmacokinetic trough concentration (C min ).
Kgp activity in GCF was again assessed using the COR553 probe.
Early time points allow assessment of target engagement in which  (Figure 4; Figure S1) indicating active site target engagement by COR388. Kgp inhibition increased between 2 and 12 hours after the first dose and was maximal at the highest dose administered ( Figure 4A-C). Furthermore, the F I G U R E 4 COR388 inhibits Kgp in a dose and time dependent manner. (A-C) Target engagement at 0 (predose), 2 and 12 hours after the first dose on day 1 was measured using COR553 probe binding and quantitated by densitometry to measure inhibition in all dosing groups. Fluorescent probe images from Figure  S1 were quantitated and ng of Kgp was determined by the purified Kgp standard curve. Percent inhibition from baseline (normalized to 0% inhibition) is shown, plotting mean ± SEM.  Figure   S1A and E). These data support dose-dependent inhibition of Kgp in oral biofluids following administration of COR388. with P. gulae, postmortem brain tissue sections from the hippocampus of 4 dogs that were not part of the current COR388 PK/PD study reported here were stained for the presence of Kgp protein by IHC using a Kgp specific antibody. Kgp staining was detected in the hippocampus in the three older dogs (aged 8-14 years), most prominently in pyramidal neurons in the cornu ammonis (CA), while little to no staining was seen in the 5-year-old dog ( Figure 6A).

| Extended dosing of COR388 for 90 days is efficacious in reducing pocket depth in dogs
Next, frozen hippocampal tissue samples from these same dogs were assessed by qPCR, with the results confirming the presence of P. gulae DNA ( Figure 6B). Sequence analysis of the PCR product confirmed the identity of the qPCR product as P. gulae ( Figure 6C).
These results are consistent with a systemic infection of P. gulae in aged dogs and are concordant with the ability of this bacterium to translocate from the oral cavity and colonize other tissues including the brain.

| D ISCUSS I ON AND CON CLUS I ON S
The results reported here support identification of a dose of

COR388 necessary for inhibition of the gingipain virulence factor
Kgp and the reduction of a naturally occurring oral P. gulae infection in dogs. These findings are consistent with our prior study demonstrating that targeting Kgp with COR388 decreased the bacterial load of P. gingivalis and AD related disease pathology in the brain of wild-type mice orally infected with P. gingivalis. 1 Importantly, we demonstrate here that Kgp proteases derived from P. gulae and P. gingivalis have similar proteolytic activity and that COR388 is a potent irreversible inhibitor of both proteases. Therefore, COR388 is predicted to have efficacy in diseases in which one or both of these bacterial pathogens are causative agents for disease pathol- The focus of the current studies was to understand the pharmacokinetics and pharmacodynamic target engagement of COR388 in aged dogs, as such we did not power the study for cognitive ef-

| E THI C S TATEMENT
Aged Beagle dogs of both sexes ranging from 8 to 15 years of age, in generally good health and located in the Vivocore Inc dog colony, were utilized in this study. All procedures were reviewed and approved by Vivocore's Institutional Animal Care and Use Committee (IACUC) and were performed in accordance with the principles of the Animal for Research Act of Ontario and the guidelines of Canadian Council on Animal Care (CCAC).

ACK N OWLED G EM ENTS
Cortexyme, Inc funded this study.

D I SCLOS U R E S
Cortexyme funded this study. SSD and CL are co-founders of

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.