Abstract
The two recent cases of HIV cure/stable remission following allogeneic stem cell transplantation are difficult to reproduce because of the toxicity of the procedure and rarity of donors homozygous for the CCR5Δ32 deletion. One approach to overcome these barriers is the use of autologous, CCR5 gene-edited hematopoietic stem and progenitor cell (HSPC) products. Unlike allogeneic transplantation, in which the frequency of CCR5Δ32 donor cells approaches 100%, the CCR5 gene can currently only be edited ex vivo in a fraction of autologous HSPCs. Therefore, we sought to determine the minimum fraction required for viral control using mathematical modeling. We analyzed data from eight juvenile pigtail macaques infected intravenously with SHIV-1157ipd3N4, treated with combination antiretroviral therapy (cART), and infused with autologous HSPCs without CCR5 gene editing. We developed a mathematical model that simultaneously describes reconstitution of CD4+CCR5+, CD4+CCR5−, and CD8+ T cell counts, as well as SHIV plasma viral loads in control and transplanted macaques. The model predicts that transplantation decreases the immunologic response to SHIV to varying degrees in macaques. By modifying the model to hypothetically describe transplantation with a given fraction of protected CCR5-edited cells we found that loss of immunologic response correlated with a more profound depletion of CCR5+CD4+ T cells and a higher fractions of gene-edited cells required for cART-free viral remission. Our results provide a framework to predict the likelihood of post-rebound control in vivo using the percentage of CCR5-edited cells in peripheral blood and the loss of HIV-specific immunity following autologous HSPC.
Key Points
Data-validated modeling suggest that loss of immune response after transplantation produces more depletion of CD4+CCR5+ T cells post-ATI.
The minimum fraction of transplanted gene-edited cells for viral control linearly relates with the CD4+CCR5+ T cell nadir 10 weeks post-ATI.
INTRODUCTION
The major obstacle to HIV-1 eradication is a latent reservoir of long-lived infected cells1–3. Cure strategies aim to eliminate all infected cells or permanently prevent viral reactivation from latency. The only known case of HIV cure4, 5 and an additional, recently-reported case of prolonged remission6, resulted from allogeneic hematopoietic stem cell transplant with homozygous CCR5Δ32 donor cells4–6. The success of this procedure is likely multifactorial—in part attributable to HIV resistance of the transplant product, the conditioning regimen that facilitates engraftment and also eliminates infected cells, graft-versus-host effect, and immunosuppressive therapies for graft-versus-host disease7–11.
A current research focus is to recapitulate this method of cure with minimal toxicity. One method is to perform autologous transplantation following ex vivo inactivation of the CCR5 gene with gene editing nucleases, eliminating the need for allogeneic CCR5-negative donors12, 13. While this procedure is safe and feasible in pigtail macaques infected with simian-human immunodeficiency virus (SHIV)13–16, only a fraction of HSPCs can be genetically modified ex vivo to be HIV-resistant.
To address this challenge, we developed a mathematical model to predict the minimum threshold of persisting, gene-modified cells necessary for functional cure. First, we modeled the kinetics of CD4+CCR5+, CD4+ CCR5−, and CD8+ T cell reconstitution after autologous transplantation. We then modeled SHIV rebound kinetics following analytical treatment interruption (ATI) and identified the degree of loss of anti-HIV cytolytic immunity following transplantation. Finally, we projected the proportion of gene-modified cells and the levels of SHIV-specific immunity required to eliminate viral replication following ATI.
METHODS
Experimental Data
Eight juvenile pigtail macaques were intravenously challenged with 9500 TCID50 SHIV-1157ipd3N4 (SHIV-C)14, 17. After 6 months, the macaques received combination antiretroviral therapy (cART: tenofovir [PMPA], emtricitabine [FTC], and raltegravir [RAL]). After ∼30 weeks on cART, four animals received total body irradiation (TBI) followed by transplantation of autologous HSPCs. After an additional 25 weeks following transplant, when viral load was fully suppressed, animals underwent analytical treatment interruption (ATI) of cART14. A control group of four animals did not receive TBI or HSPC transplantation and underwent ATI after ∼50 weeks of treatment (Fig. 1A). Plasma viral loads and absolute quantified CD4+CCR5−, CD4+CCR5+ and CD8+ total and subsets (naïve, central memory [TCM], and effector memory [TEM]) T cell counts from peripheral blood were measured as described previously14, 17. We analyzed peripheral T cell counts and plasma viral load from transplant until 43 weeks post-transplant (∼25 weeks pre-ATI and ∼20 weeks post-ATI).
Mathematical modeling
We employed several series of ordinary differential equation models of cellular and viral dynamics after transplantation (Fig. 1B). First, we modeled T cell dynamics and reconstitution following transplant and before ATI, assuming that low viral loads on ART do not affect cell dynamics (Fig. 1C). After curation of that model, we introduced viral dynamics and fit those to the T cell and viral rebound dynamics from the animals pre- and post-ATI (Fig. 1D). Lastly, we applied our complete model to a transplant scenario with gene editing of CCR5 to predict the minimal threshold of editing for functional HIV cure (Fig. 1E).
T cell reconstitution after transplantation
We modeled the kinetics of CD4+ and CD8+ T cell subsets in blood, transplanted cells that home to the BM, and progenitor cells in the BM/thymus as shown in Fig. 1C. We included CD8+ T cells in the model because CD8+ and CD4+ T cells may arise from new naïve cells from the thymus and compete for resources that impact clonal expansion and cell survival18–20. At the moment of HSPC infusion, transplanted animals are lymphopenic due to TBI. The control group did not have a transplanted-cell compartment, and all other compartments remained in steady state. We assumed that CD4+ and CD8+ T cell expansion may have two possible drivers: (1) lymphopenia-induced proliferation of mature cells that persist through myeloablative TBI18, 21–25, and (2) differentiation from naïve cells from progenitors in the thymus (from transplanted CD34+ HSPCs26, 27 or CD34+ HSPCs that persist following TBI) and further differentiation to an activated effector state24, 25, 28–32. We assumed that in a lymphopenic environment, factors that drive T cell proliferation are more accessible (i.e., self-MHC molecules on antigen-presenting cells28, 29, 33, 34 and γ-chain cytokines such as IL-7 and IL-1521–23, 35–37). However, as they grow, cells compete for access to these resources, limiting clonal expansion18 such that logistic growth models are appropriate19. We assume that new peripheral CD4+ and CD8+ T naïve cells come from a progenitor compartment in the BM/Thymus38, 39. For CD4+ T cells, we assume that naïve cells do not express CCR540–42, and subsequently up- and/or down-regulate expression of the CCR5 receptor30. For CD8+ T cells, we included a single CD8+ memory precursor compartment of TN and TCM cells that differentiate linearly into TEM during lymphopenia43–45. The details of the model are presented in the Supp. Material and in Fig. 1C with the notation described in Table 1. A parsimonious, curated version of this model was selected from a series of models with varying mechanistic and statistical complexity as presented in the Supp. Materials.
T cell and viral dynamics
We next adapted the curated T cell reconstitution model by combining several adaptations of the canonical model of viral dynamics43–53 as shown in Fig. 1D. The model assumes that SHIV infects only CD4+CCR5+ T cells17 and that a small fraction (∼ 5%) of those infected cells are able to produce infectious virus51, 54, 55. We modeled cART by reducing the infection rate to zero and modeled ATI by assuming infection is greater than zero after some time Δt after interruption. This model assumes that productively infected cells arise also from activation of a steady set of latently infected cells. The presence of both unproductively and productively infected cells leads to the expansion of CD8+ Tnaïve and TCM cells, from which the majority of dividing cells differentiate into SHIV-specific effector cells30, 46, 47, 52, 53. The details of the model are presented in the Supp. Material and in Fig. 1D with the notation described in Table 1. A parsimonious version of this model was selected from a series of models with varying mechanistic and statistical complexity as presented in the Supp. Materials.
Viral and T cell dynamics in the setting of ΔCCR5 HSPC transplantation
We next adapted our full model to simulate scenarios in which autologous transplantation includes cells that are CCR5-edited (Fig 1E). We added variables representing CCR5-edited HSPCs in different compartments: (1) infused HSPCs in blood, (2) T cell progenitors in BM/thymus, and (3) CD4+CCR5−T cells in blood. These compartments have the same structure as CCR5-non-edited cells but with two differences. First, the value of gene-edited HSPCs at transplantation is a fraction fp of the total number of infused cells. Second, mature, CCR5-edited CD4+CCR5− cells do not upregulate CCR5 (see full model in Supp. Materials).
Fitting procedure and model selection
To fit our models (Fig. 1C-D) to the transplant data, we used a nonlinear mixed-effects modeling approach56 described in detail in the Supp. materials. Briefly, we estimated the population mean and variance for each model parameter using the Stochastic Approximation Expectation Maximization (SAEM) algorithm embedded in the Monolix software (www.lixoft.eu). For a subset of parameters, random effects were specified, and those variances were estimated. Measurement error variance was also estimated assuming an additive error model for the logged outcome variables.
We first fit instances of models with varying statistical and parameter complexity in Fig. 1C to blood T cell counts during transplant and before ATI assuming that one or multiple mechanisms are absent, or that certain mechanisms have equal kinetics (Table S1 includes all 19 competing models). Then, we fit several instances of the model Fig. 1D to blood T cell counts and plasma viral load during the period after transplant including ATI using the best competing model (solid lines) for the model in Fig. 1C (Table S2 includes all 15 competing models including viral dynamics). To determine the best and most parsimonious model among the instances, we computed the log-likelihood (log L) and the Akaike Information Criteria (AIC=-2log L+2 m, where m is the number of parameters estimated)57. We assumed a model has similar support from the data if the difference between its AIC and the best model (lowest) AIC is less than two57 (see Supp. materials for details).
Simulations for each animal were computed using individual-level parameter estimates generated from the predicted random effects of the fitted population model.
Data sharing statement. Original data will be shared upon request.
RESULTS
CD4+CCR5+ and CD8+ T cells recover more rapidly than CD4+CCR5− T cells after HSPC transplantation
We analyzed the kinetics of peripheral blood CD4+CCR5+ and CD4+CCR5− T-cells, and total, Tnaïve, TCM, and TEM CD8+ T-cells in macaques after HSPC transplantation. In controls, levels of CD4+ and CD8+ T cells oscillated around a persistent set point (Fig. S1). CD4+ CCR5+ T cell levels were ∼100 cells/μl and were uniformly lower than the CD4+CCR5− T cell counts (∼1200 cells/μl) (p=0.01, paired t-test of the averaged post-transplant measures, Fig. 2A). Total CD8+ T cell levels in the control group were ∼1400 cells/μl with a greater contribution from TEM (73%) than TN+TCM (27%) (based on median values, Fig. 2B).
In the transplant group, CD4+ and CD8+ T cells expanded at different rates from the remaining levels post-TBI (Fig. 2C-D). The levels of CD4+CCR5+ T cells started at 1-10 cells/μl and reconstituted to levels similar to the control group over 5-10 weeks (Fig. 2D). After TBI, CD4+CCR5− T cells remained at higher levels (∼100 cells/μl) than CD4+CCR5+ T cells but expanded more slowly and did not reach the values of the control group after 25 weeks (Figs. 2C-D). The CD4+CCR5+ T cell compartment expanded 8-fold more rapidly than the CD4+CCR5− compartment (p=0.01, paired t-test, Figs. 2C-D). CD8+ T cells decreased to levels between 10 and 100 cells/μl after TBI but recovered to levels below the control group in 5 weeks (Figs. 2C-D); CD8+ T cells recovered as rapidly as the CD4+CCR5+ population (Figs. 2C-D).
Overall, these results show that after transplantation there is a faster reconstitution of CD4+CCR5+ and CD8+ T cells compared to CD4+CCR5− cells, suggesting each cell subset may have different mechanisms that drive their expansion. To explore these mechanisms, we analyzed the data in the context of mechanistic mathematical models of cell dynamics.
Lymphopenia-induced proliferation drives early CD4+CCR5+ and CD8+ T cell reconstitution after HPSC transplantation
To identify the main drivers of T cell reconstitution after transplant, we developed a mathematical model that considered plausible mechanisms underlying reconstitution of distinct T cell subsets following autologous transplantation (Fig. 1C). We built 19 versions of the model by assuming that one or multiple mechanisms are absent, or by assuming certain mechanisms have equivalent or differing kinetics (Table S1). Using model selection theory based on AIC, we identified the model in Fig. 1C without the dashed lines which most parsimoniously reproduces the data (Table S1). The best fits of this model are presented in Fig. 2D and Fig. S1 with the respective parameter estimates in Tables S3-S4. The main features of this model are: (1) CD4+CCR5+ T cell reconstitution after transplant is driven by proliferation and upregulation of CCR5; (2) CD4+CCR5− T cell expansion is driven only by new naïve cells from the thymus and not proliferation; and (3), thymic export rates are equal between CD4+ and CD8+ naïve T cells.
The best fit model predicts that CD4+CCR5− T cells have a delayed reconstitution that occurs only when cells from the thymus (estimated with rate ∼0.01/day) outnumber their loss due to death, trafficking to tissues, or upregulation of CCR5. Furthermore, the estimated CD4+CCR5+ T cell proliferation rate (∼0.1/day) far exceeds the estimated CCR5 upregulation (∼0.004/day) and thymic export rates (∼0.01/day). Therefore, one month after transplantation, the total concentration of CD4+CCR5+ T new cells generated by proliferation is predicted to be 40-fold higher than the concentration generated by up-regulation of CCR5 (Fig. 3).
Our model predicts that CD8+ T cells follow a similar pattern to CD4+CCR5+ T cells (Fig. 2D), as the CD8+ TEM proliferation rate is up to 10-fold higher than the CD8+ T cell differentiation rate (Fig. S2). Overall, these results suggest that following autologous HSPC transplant: (1) slow thymic export is the main driver of CD4+CCR5− T cell growth, and (2) rapid lymphopenia-induced proliferation of remaining cells (rather than transplanted cells) after TBI is the main driver for CD4+CCR5+ and CD8+ T cell expansion.
Reduction of blood CD4+CCR5+ T cell counts correlates with plasma viral rebound after ATI in animals that underwent HSPC transplantation
We next aimed to extend previous analysis comparing plasma viral load rebound kinetics to CD4+CCR5+ and CCR5− T cell subset dynamics after ATI14, 51. Fig. 4 and Fig. S3 presents the plasma viral loads and the blood CD4+CCR5+ and CD4+CCR5− T cell kinetics before and after ATI in transplanted and control macaques respectively. Viral burden after ATI was more pronounced in the transplant group compared to the control group: median peak viral load was 10-fold higher (p=0.06, Mann-Whitney test. See Fig. 4A) and median final viral load measurements at necropsy were 2-log10 higher (p=0.06, Mann-Whitney test. See Fig.4B). CD4+CCR5+ T-cell counts decreased after ATI in the transplant group reaching a significantly lower nadir (∼8-fold) than the control animals (p=0.01, Mann-Whitney test. Fig. 4C). The two animals with the largest reduction of CD4+CCR5+ T cells had the highest viral set points. There was no difference between control and transplant groups’ CD4+CCR5− T cell nadir: both groups had an average reduction of ∼200 cells/μL (Fig.4D).
In the control group, individual plasma viral loads did not correlate with corresponding CD4+CCR5+ T-cell counts post-ATI. However, in three animals in the transplant group, viral load observations post-ATI correlated negatively with their corresponding CD4+CCR5+ T cell counts (Fig. S4).
Overall, these results show that transplanted animals had higher viral load that correlated with the reduction of CD4+CCR5+ T cells after ATI. This suggests that transplantation might affect macaques’ immunologic response to SHIV so that the presence of the virus leads to more depletion of CD4+CCR5+ T cells. We explore this possibility by simultaneously analyzing the viral and T cell subset observations using novel mechanistic mathematical models.
Higher viral set points and CD4+CCR5+ T-cell depletion following transplantation and ATI are due to a reduction in SHIV-specific immunity
To understand why transplantation may have an effect on plasma viral load and CD4+CCR5+ T cell kinetics during ATI, we modified our mathematical model to include SHIV infection as described in Fig. 1D (Methods). Using model selection theory based on AIC, we found that the most parsimonious model to explain the data was the one without the dashed lines in Fig. 1D (Table S2). The best fit model simultaneously recapitulates plasma viral rebound and the kinetics of CD4+ CCR5+ and CCR5− T cells in each animal as shown in Fig. 4E and Fig. S3 with corresponding estimated parameters in Table 1 and Table S5-S6. In the best fit model, SHIV-specific CD8+ effector cells reduce virus production in a non-cytolytic manner58–60, possibly by secretion of HIV-antiviral factors61–64—not included in the model. Additionally, the model suggests that infection leads to enhanced activation of CD4+CCR5− T cells leading to replenishment of CD4+CCR5+ T cells, explaining the concentration reduction of the CD4+CCR5− compartment after ATI65–68.
From the estimated parameters, only SHIV-based CD8+ proliferation rate, ω8, correlated positively with post-ATI CD4+CCR5+ T-cell nadir and negatively with viral load set point (Fig. 5A-B). We also found that the estimated SHIV-based CD8+ proliferation rate was significantly lower in the transplant group, and the estimated time to viral rebound (Δt) was significantly higher in the transplant group (Fig. 5C-D). The projected fraction of SHIV-specific CD8+ T cells in the transplant group approached zero (Fig. S5). Overall, these results suggest that a lower nadir of CD4+CCR5+ T cells and a higher viral load after ATI in transplanted animals is due to a loss of the immune response to SHIV-infected cells.
Greater loss of immunologic control during TBI/transplant requires higher numbers of CCR5-edited HSPCs to control viral rebound after ATI
To calculate the minimum threshold of CCR5-edited cells necessary to induce cART-independent virus suppression, we added a population of transplanted, gene-edited CCR5 HSPCs to our complete, fitted model of T cell subset and viral dynamics. We assumed that in the infused product there is a fraction fp of HSPCs that have a biallelically-modified CCR5 gene and do not express CCR5. In the model we added state variables for protected progenitors and CD4+CCR5− T cells that cannot become CD4+CCR5+ T cells (Fig. 1E, full model in Supp. Materials).
We simulated the model using parameter values obtained from the best fit in the previous section for each animal in the transplant group using 100 values of fp from zero to one (0-100% CCR5-edited HSPCs). The minimal initial fraction to achieve post-rebound viral control was dependent on the underlying viral and immune dynamics of the given animal. For example, Fig. 6A depicts projections of the model using the best estimates from the fits of the model to transplanted animal Z09144 using six values of fp: an initial fraction of protected cells smaller than or equal to 40% will not lead to post-rebound viral control after ATI, even after a year. However, when fp is 60% or greater than 80% it is possible to have a spontaneous post-rebound viral control at ∼40 weeks and 10 weeks after ATI respectively. In both cases, a period of high-level viremia occurs prior to control.
The heatmaps in Fig. 6B-E show plasma viral load projections over 2 years after the start of ATI for different values of fp. The model predicts that the minimum fp to maintain post-rebound control for 2 years after ATI is higher for animals with lower estimated SHIV-specific immune response rates. For the two animals in the transplant group with lower viral setpoints, the minimum fpfor viral control was 35% and 19% (Fig. 6D-E). In contrast, for the other two animals, the minimum fp for viral control was 56% and 97% (Fig. 6B-C). These model projections suggest that a larger loss of immunologic control during TBI/HSPC transplant results in a higher fraction of CCR5 gene-edited cells required for control of viral rebound after ATI.
The model also predicts that for some values of fpit is possible to have two viral load stages after ATI: a temporary high viral set point in the first weeks after ATI may be followed by a delayed ART-free viral remission stage (e.g. when fp is between 60% and 70% in Fig. 6B or between 5% and 20% in Fig. 6E). Therefore, in some cases the viral load set point determined during the initial weeks after ATI might not be a sufficient surrogate to predict viral control further in the future. On the other hand, when we project the CD4+CCR5+ T cell count for the same example in Fig. 6A we find that this cell subset does not undergo a significant change between weeks 2 and 10 after ATI for different scenarios of fp (Fig. 7A). Moreover, the maximum decrease of CD4+CCR5+ T cells observed during the first 10 weeks after ATI is predicted to have a linear relationship with the minimum initial fraction of protected cells required to obtain post-rebound control after 2 years (Fig. 7B). Therefore, the maximum initial change in CD4+CCR5+ T cells 10 weeks after ATI, as well as the observed experimental value for fp, might predict late viral control.
Discussion
Here we introduce a data-validated mathematical model that to our knowledge is the first to simultaneously recapitulate SHIV viral loads, and CD4+ and CD8+ T cell subset counts during HIV or SHIV infection. We systematically selected from a series of models to arrive at a set of equations that most parsimoniously explains the available data. We recapitulated (1) peripheral CD4+ and CD8+ T-cell subset reconstitution dynamics following transplant, and (2) T-cell dynamics and SHIV viral rebound following ATI. Before ATI, all animals suppressed plasma viral load below the limit of detection, allowing analysis of T cell reconstitution dynamics independent of virus-mediated pressure. At each step, we applied model selection theory to select the simplest set of mechanisms capable of explaining the observed data57.
The model predicts that post-rebound viral control might be possible during autologous gene-edited HSPC transplantation if therapy achieves (1) a sufficient fraction of gene-protected, autologous HSPCs, and (2) maintenance or enhancement of SHIV-specific immune responses following transplantation. Specifically, the model predicts that increasing amounts of conditioning regimen-dependent depletion of the SHIV-specific immune response leads to a higher threshold of CCR5-gene-edited cells in the transplanted HSPC product that is required to obtain stable, ART-free viral control. These results are consistent with the cure achieved by the Berlin patient who received transplant with 100% HIV-resistant cells after intense conditioning4, 5. In the autologous setting where 100% CCR5 editing may not be feasible, adjunctive measures that augment virus-specific immunity, such as therapeutic vaccination, infusion of HIV-specific CAR T cells or use of neutralizing antibodies, may synergize with HSPC transplantation to achieve post-treatment control11, 69.
The best model predicts that the lack of complete elimination of lymphocytes by TBI prevents CD4+CCR5− cells from predominating post-transplant: the rapid expansion of CD4+CCR5+ and CD8+ T cells during the first few weeks after HSPC transplantation is most likely due to lymphopenia-induced proliferation of remaining cells after TBI via a thymus-independent pathway; the slower expansion of CD4+CCR5− T cells is due to thymic export of both transplanted and remaining cells. An important future research question will be to identify anatomic sites and mechanisms that allow activated CD4+CCR5+ to survive conditioning.
A challenge is that more intense conditioning may decrease remaining CD4+CCR5+ cells but will also lower SHIV specific immunity. We previously demonstrated the link between disruption of the immune response during transplant and increased magnitude of viral rebound during treatment interruption14, 51. Here we predict that the magnitude of the SHIV-specific immune response is correlated not only with viral load set point, but also with the reduction of CD4+CCR5+ T cells after ATI. CD4+CCR5+ T cell depletion might also be predictive of the loss of depletion of virus-specific immunity following conditioning.
A final important observation from the model is that viral control may be delayed beyond the first ten weeks after ATI, and instead occur many months after ATI. Thus, viral load levels during the initial weeks after ATI may not completely define success (stable ART-free remission), whereas CD4+CCR5+ T-cell nadir should more strongly correlate with the degree of depletion of virus-specific immunity. In this sense, minimal CD4+CCR5+ T-cell nadir may predict post-rebound viral control, if the starting fraction of protected cells is known.
Our results are limited by a small sample size of eight animals. For that reason, several model parameters were assumed to be the same among the population (i.e., without random effects). However, the number of observations for each animal was large enough to discriminate among different plausible model candidates. Therefore, we performed projections using only the individual estimated parameters. Reassuringly, our results align with prior mechanistic studies of cellular reconstitution after stem cell transplantation18, 26, 38, 70, 71. Our analysis also suggests that the majority of reconstituting CD4+CCR5− T cells do not proliferate and have a slow expansion that concurs with estimates of thymic export from previous studies26, 70, 71.
The interplay between reconstituting HIV susceptible CD4+ T cells, HIV-resistant CD4+ T cells, infected cells, virus-specific immune cells, and replicating virus following autologous, CCR5-edited HPSC transplantation is extremely complex. Our results illustrate the capabilities of mathematical models to glean insight from this system and highlight that modeling will be required to optimize strategies for HIV cure studies, both in the macaque model, as well as in HIV+ individuals.
Authorship Contributions
E.F.C. and J.T.S. conceived the study. C.W.P. and H.P.K. contributed ideas and data sources for the project. E.R.D. and D.B.R. contributed ideas for the development of mechanistic mathematical models. B.T.M. contributed ideas and support for statistical models and analyzes. E.F.C. assembled data, wrote all code, performed all calculations and derivations, ran the models, and analyzed output data. J.T.S. and E.F.C. wrote the manuscript with contributions from all other authors.
Disclosure of Conflicts of Interest
The authors declare no competing financial interests.
Animal Welfare
The data used in this work were collected in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The study protocol was approved by the Institutional Animal Care and Use Committees (3235-03) of the Fred Hutchinson Cancer Research Center and the University of Washington.
Acknowledgements
This study was supported by grants from the National Institutes of Health, National Institute of Allergy and Infectious Diseases (UM1 AI126623). ERD is supported by the National Center for Advancing Translational Sciences of the National Institutes of Health under Award Number KL2 TR002317. DBR is supported by a Washington Research Foundation postdoctoral fellowship, and a CFAR NIA P30 AI027757. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or the Washington Research Foundation.