1010. Cross-Species Translation of Correlates of Protection for COVID-19 Vaccine Candidates Using Quantitative Tools

Abstract Background Several COVID-19 vaccines have been authorized, and the need for rapid, further modification is anticipated. This work uses a Model-Based Meta-Analysis (MBMA) to relate, across species, immunogenicity to peak viral load (VL) after challenge and to clinical efficacy. Together with non-clinical and/or early clinical immunogenicity data (ECID), this enables prediction of a candidate vaccine’s clinical efficacy. The goal of this work was to enable the accelerated development of vaccine candidates by supporting Go/No-Go and study design decisions, and the resulting MBMA can be instrumental in decisions not to progress candidates to late stage development. Methods A literature review with pre-specified inclusion/exclusion criteria enabled creation of a database including nonclinical serum neutralizing titers (SN), peak VL after challenge with SARS-CoV-2 (VL), along with data from several clinical vaccine candidates. Rhesus Macaque (RM) and golden hamster (GH) were selected (due to availability and consistency of data) for MBMA modeling. For both RM and GH, peak post-challenge VL in lung and nasal tissues were used as surrogates for clinical disease and were related to pre-challenge SN via the MBMA. The VL predictions from the RM MBMA were scaled to incidence rates in humans, with a scaling factor between RM and human SN estimated using early Phase 3 efficacy data. This enabled clinical efficacy predictions based on ECID. To qualify the model’s predictive power, efficacies of COVID-19 vaccine candidates were compared to those predicted from the MBMA and their respective Ph1/2 SN data. More recently available clinical data enable building a clinical MBMA; comparing this to the RM MBMA further supports SN as predictive. Results The MBMA analyses identified a sigmoidal decrease in VL (increasing protection) with increase in SN in all three species, with more SN needed (in both RM and GH) for protection in nasal swabs than in BAL (see figure). The comparison between predicted and reported clinical efficacies demonstrated the model’s predictive power across vaccine platforms. RM and GH MBMA Protection Models and Translational Prediction with Observed Efficacies Sizes of circles indicate relative weight of the data in the respective quantitative model. Model and data visualizations have been harmonized (across tissue-types) separately for each of RM and GH using VACHER (Lommerse, et al., CPT:PSP, in press). Conclusion By quantifying adjustments needed between species and assays, translational MBMA can inform development decisions by using nonclinical SN and VL, and ECID to predict protection from COVID-19. Disclosures Anna Largajolli, PhD, Certara (Employee) Nele Plock, PhD, Certara (Employee, Shareholder)Merck & Co., Inc. (Independent Contractor) Bhargava Kandala, PhD, Merck & Co., Inc. (Employee, Shareholder) Akshita Chawla, PhD, Merck & Co., Inc. (Employee, Shareholder) Seth H. Robey, PhD, Merck & Co., Inc. (Employee, Shareholder) Kenny Watson, PhD, Certara (Employee, Shareholder) Raj Thatavarti, MS, Certara (Employee, Shareholder) Sheri Dubey, PhD, Merck & Co., Inc. (Employee, Shareholder) S. Y. Amy Cheung, PhD, Certara (Employee, Shareholder) Rik de Greef, MSc, Certara (Employee, Shareholder) Jeffrey R. Sachs, PhD, Merck & Co., Inc. (Employee, Shareholder)

(A) Serum from long-term non-progressors (LTNPs) compared to serum from a group of HIV infected with lower CD4 levels as a control for viral load were used to compete against biotinylated CD4 binding site (VRC01) and 76C Gp41 conformational epitope (6F11) targeting antibodies. Serum dilutions were chosen to align means near 50%. Means with 95% confidence intervals are shown. (B) Monoclonal antibody 76Canc was created using the germline sequence of the heavy chain variable region with the CDR3 and light chain of 76C member. Antibody dependent cell cytotoxicity flow cytometric based assays were performed using gp41 proteins from clade B (MN) and clade C (ZA1197).
Conclusion. Certain antibodies present early on in infection may contribute to overall clinical course. Variable gene germline sequences that support functional activity against HIV could be targeted in vaccine regimens.
Disclosures. Background. Multidrug resistant Acinetobacter baumannii (MDR-Ab) is a Gram-negative bacterium known for causing severe nosocomial infections, attributed in part to its formation of biofilm. Siderophore is a virulence factor known to support biofilm formation by regulating iron availability. In this study, we screened 44 isolates of MDR-Ab from our Gram-negative repository to determine the strains that phenotypically form biofilm and produce siderophore. The results were compared to Pseudomonas aeruginosa PAO1, which produces both biofilm and siderophore.
Methods. Isolates were grown overnight in minimal M9 medium supplemented with casamino acids and hydroxyquinones at 37°C. Bacterial cells were normalized (to OD 600=0.01) and a standard diluted 10 -3 tube was used in the study. A 96-well plate was inoculated with 100 microliters of each isolate in quadruplicates. This process was repeated in Tygon tubes with 50 microliters of each isolate in triplicates. The plate and Tygon tubes were incubated statically for 48 hours at 30°C and then stained with crystal violet. The contents were dissolved in 33% glacial acetic acid and analyzed by spectrophotometry to measure biofilm formation. Siderophore secretion was measured in supernatants with Chrome Azurol S (CAS) reagent and production was observed on CAS agar plates.
Conclusion. Many strains of MDR-Ab readily form biofilm. Overall siderophore production is lower in MDR-Ab compared to consistent production by PAO1, but this does not appear to affect MDR-Ab's ability to form biofilm. Unlike in PAO1, biofilm formation in MDR-Ab may occur independently of siderophore production. This research serves as a basis for understanding future MDR-Ab biofilm elimination in patient catheters and indwelling devices.

Disclosures. All Authors: No reported disclosures
Go/No-Go and study design decisions, and the resulting MBMA can be instrumental in decisions not to progress candidates to late stage development.

Methods.
A literature review with pre-specified inclusion/exclusion criteria enabled creation of a database including nonclinical serum neutralizing titers (SN), peak VL after challenge with SARS-CoV-2 (VL), along with data from several clinical vaccine candidates. Rhesus Macaque (RM) and golden hamster (GH) were selected (due to availability and consistency of data) for MBMA modeling. For both RM and GH, peak post-challenge VL in lung and nasal tissues were used as surrogates for clinical disease and were related to pre-challenge SN via the MBMA. The VL predictions from the RM MBMA were scaled to incidence rates in humans, with a scaling factor between RM and human SN estimated using early Phase 3 efficacy data. This enabled clinical efficacy predictions based on ECID. To qualify the model's predictive power, efficacies of COVID-19 vaccine candidates were compared to those predicted from the MBMA and their respective Ph1/2 SN data. More recently available clinical data enable building a clinical MBMA; comparing this to the RM MBMA further supports SN as predictive.
Results. The MBMA analyses identified a sigmoidal decrease in VL (increasing protection) with increase in SN in all three species, with more SN needed (in both RM and GH) for protection in nasal swabs than in BAL (see figure). The comparison between predicted and reported clinical efficacies demonstrated the model's predictive power across vaccine platforms.

RM and GH MBMA Protection Models and Translational Prediction with Observed Efficacies
Sizes of circles indicate relative weight of the data in the respective quantitative model. Model and data visualizations have been harmonized (across tissue-types) separately for each of RM and GH using VACHER (Lommerse, et al., CPT:PSP, in press).
Conclusion. By quantifying adjustments needed between species and assays, translational MBMA can inform development decisions by using nonclinical SN and VL, and ECID to predict protection from COVID-19.
Disclosures Background. The global threat of antimicrobial resistance (AMR) varies regionally. Regional differences may be related to socio-economic factors such as the Area Deprivation Index (ADI) score. Our hypothesis is that AMR spatial distribution is not random.
Methods. Patient level antibiotic susceptibility data was collected from three regionally distinct Wisconsin health systems (UW Health, Fort HealthCare, Marshfield Clinic Health System [MCHS]). Patient addresses were geocoded to coordinates and joined with US Census Block Groups. For each culture source, we included the initial E. coli isolate per patient per year with a patient address in Wisconsin. Percent susceptibility was calculated by block group. Spatial autocorrelation was determined by Global Moran's I, which quantifies the attribute being analyzed as spatially dispersed, randomly distributed, or clustered by a range of −1 to +1. Linear regression correlated ADI to susceptibility. Hot spot analysis identified blocks with statistically significant higher and lower susceptibility (Figure 1). Results. The UW Health results included more urban areas, more block groups and greater isolate geographic density (n = 44,629 E. coli, 2009-2018), compared to Fort HealthCare (n = 6,065 isolates, 2012-2018 and MCHS (50,405 isolates, 2009MCHS (50,405 isolates, -2018. A positive spatially clustered pattern was identified from the UW Health data for ciprofloxacin (Moran's I = 0.096, p = 0.005) and trimethoprim/sulfamethoxazole (TMP/SMX) susceptibility (Moran's I = 0.180, p < 0.001; Figures 2-3). Fort HealthCare and MCHS distribution was likely random for TMP/SMX and ciprofloxacin by Moran's I. Linear regression of ADI (scale 1-10, least to most disadvantaged) and susceptibility did not find significance, but susceptibility was lower in more disadvantaged block groups. At the local level, we identified hot and cold spots with 90%, 95%, and 99% confidence, with more hot spots in rural regions.

Conclusion.
Overall, Moran's I analysis is more able to identify a clustered pattern in urban versus rural areas. Yet, the local hot spot results indicate that variations in antibiotic susceptibility may be more common in rural areas. The results are limited to data from patients with access to the health systems included.