Publications

IL-1R1 deficiency in mice causes severe susceptibility to Mycobacterium tuberculosis Mice and macrophage cultures lacking IL-1R1 display increased bacterial growth, suggesting that phagocytes may require IL-1R1-dependent antimicrobial signals to limit intracellular M. tuberculosis replication directly. However, the myeloid-cell-intrinsic versus -extrinsic requirements for IL-1R1 to control M. tuberculosis infection in mice have not been directly addressed. Using single-cell analysis of infected cells, competitive mixed bone marrow chimeras, and IL-1R1 conditional mutant mice, we show in this article that IL-1R1 expression by pulmonary phagocytes is uncoupled from their ability to control intracellular M. tuberculosis growth. Importantly, IL-1R1-dependent control was provided to infected cells in trans by both nonhematopoietic and hematopoietic cells. Thus, IL-1R1-mediated host resistance to M. tuberculosis infection does not involve mechanisms of cell-autonomous antimicrobicidal effector functions in phagocytes but requires the cooperation between infected cells and other cells of hematopoietic or nonhematopoietic origin to promote bacterial containment and control of infection.

Recombination in HIV-1 is well documented, but its importance in the low-diversity setting of within-host diversification is less understood. Here we develop a novel computational tool (RAPR (Recombination Analysis PRogram)) to enable a detailed view of in vivo viral recombination during early infection, and we apply it to near-full-length HIV-1 genome sequences from longitudinal samples. Recombinant genomes rapidly replace transmitted/founder (T/F) lineages, with a median half-time of 27 days, increasing the genetic complexity of the viral population. We identify recombination hot and cold spots that differ from those observed in inter-subtype recombinants. Furthermore, RAPR analysis of longitudinal samples from an individual with well-characterized neutralizing antibody responses shows that recombination helps carry forward resistance-conferring mutations in the diversifying quasispecies. These findings provide insight into molecular mechanisms by which viral recombination contributes to HIV-1 persistence and immunopathogenesis and have implications for studies of HIV transmission and evolution in vivo.

Cytotoxic T lymphocytes (CTLs) have been suggested to play an important role in controlling human immunodeficiency virus (HIV-1 or simply HIV) infection. HIV, due to its high mutation rate, can evade recognition of T cell responses by generating escape variants that cannot be recognized by HIV-specific CTLs. Although HIV escape from CTL responses has been well documented, factors contributing to the timing and the rate of viral escape from T cells have not been fully elucidated. Fitness costs associated with escape and magnitude of the epitope-specific T cell response are generally considered to be the key in determining timing of HIV escape. Several previous analyses generally ignored the kinetics of T cell responses in predicting viral escape by either considering constant or maximal T cell response; several studies also considered escape from different T cell responses to be independent. Here, we focus our analysis on data from two patients from a recent study with relatively frequent measurements of both virus sequences and HIV-specific T cell response to determine impact of CTL kinetics on viral escape. In contrast with our expectation, we found that including temporal dynamics of epitope-specific T cell response did not improve the quality of fit of different models to escape data. We also found that for well-sampled escape data, the estimates of the model parameters including T cell killing efficacy did not strongly depend on the underlying model for escapes: models assuming independent, sequential, or concurrent escapes from multiple CTL responses gave similar estimates for CTL killing efficacy. Interestingly, the model assuming sequential escapes (i.e., escapes occurring along a defined pathway) was unable to accurately describe data on escapes occurring rapidly within a short-time window, suggesting that some of model assumptions must be violated for such escapes. Our results thus suggest that the current sparse measurements of temporal CTL dynamics in blood bear little quantitative information to improve predictions of HIV escape kinetics. More frequent measurements using more sensitive techniques and sampling in secondary lymphoid tissues may allow to better understand whether and how CTL kinetics impacts viral escape.

The ability of human immunodeficiency virus (HIV) to avoid recognition by humoral and cellular immunity (viral escape) is well-documented, but the strength of the immune response needed to cause such a viral escape remains poorly quantified. Several previous studies observed a more rapid escape of HIV from CD8 T cell responses in the acute phase of infection compared to chronic infection. The rate of HIV escape was estimated with the help of simple mathematical models, and results were interpreted to suggest that CD8 T cell responses causing escape in acute HIV infection may be more efficient at killing virus-infected cells than responses that cause escape in chronic infection, or alternatively, that early escapes occur in epitopes mutations in which there is minimal fitness cost to the virus. However, these conclusions were challenged on several grounds, including linkage and interference of multiple escape mutations due to a low population size and because of potential issues associated with modifying the data to estimate escape rates. Here we use a sampling method which does not require data modification to show that previous results on the decline of the viral escape rate with time since infection remain unchanged. However, using this method we also show that estimates of the escape rate are highly sensitive to the time interval between measurements, with longer intervals biasing estimates of the escape rate downwards. Our results thus suggest that data modifications for early and late escapes were not the primary reason for the observed decline in the escape rate with time since infection. However, longer sampling periods for escapes in chronic infection strongly influence estimates of the escape rate. More frequent sampling of viral sequences in chronic infection may improve our understanding of factors influencing the rate of HIV escape from CD8 T cell responses.

Liver-resident CD8+ T cells are highly motile cells that patrol the vasculature and provide protection against liver pathogens. A key question is: how can these liver CD8+ T cells be simultaneously present in the circulation and tissue-resident? Because liver-resident T cells do not express CD103 - a key integrin for T cell residence in epithelial tissues - we investigated other candidate adhesion molecules. Using intra-vital imaging we found that CD8+ T cell patrolling in the hepatic sinusoids is dependent upon LFA-1-ICAM-1 interactions. Interestingly, liver-resident CD8+ T cells up-regulate LFA-1 compared to effector-memory cells, presumably to facilitate this behavior. Finally, we found that LFA-1 deficient CD8+ T cells failed to form substantial liver-resident memory populations following Plasmodium or LCMV immunization. Collectively, our results demonstrate that it is adhesion through LFA-1 that allows liver-resident memory CD8+ T cells to patrol and remain in the hepatic sinusoids.

The suppression of viral loads and identification of selection signatures in non-human primates after challenge are indicators for effective human immunodeficiency virus (HIV)/simian immunodeficiency virus (SIV) vaccines. To mimic the protective immunity elicited by attenuated SIV vaccines, we developed an integration-defective SIV (idSIV) vaccine by inactivating integrase, mutating sequence motifs critical for integration, and inserting the cytomegalovirus (CMV) promoter for more efficient expression in the SIVmac239 genome. Chinese rhesus macaques were immunized with idSIV DNA and idSIV particles, and the cellular and humoral immune responses were measured. After the intravenous SIVmac239 challenge, viral loads were monitored and selection signatures in viral genomes from vaccinated monkeys were identified by single genome sequencing. T cell responses, heterologous neutralization against tier-1 viruses, and antibody-dependent cellular cytotoxicity (ADCC) were detected in idSIV-vaccinated macaques post immunization. After challenge, the median peak viral load in the vaccine group was significantly lower than that in the control group. However, this initial viral control did not last as viral set-points were similar between vaccinated and control animals. Selection signatures were identified in Nef, Gag, and Env proteins in vaccinated and control macaques, but these signatures were different, suggesting selection pressure on viruses from vaccine-induced immunity in the vaccinated animals. Our results showed that the idSIV vaccine exerted some pressure on the virus population early during the infection but future modifications are needed in order to induce more potent immune responses.

It is generally thought that Mycobacterium tuberculosis (Mtb)-specific CD4+ Th1 cells producing IFN-γ are essential for protection against tuberculosis (TB). In some studies, protection has recently been associated with polyfunctional subpopulation of Mtb-specific Th1 cells, i.e., with cells able to simultaneously secrete several type 1 cytokines. However, the role for Mtb-specific Th1 cells and their polyfunctional subpopulations during established TB disease is not fully defined. Pulmonary TB is characterized by a great variability of disease manifestations. To address the role for Mtb-specific Th1 responses during TB, we investigated how Th1 and other immune cells correlated with particular TB manifestations, such as the degree of pulmonary destruction, TB extent, the level of bacteria excretion, clinical disease severity, clinical TB forms, and "Timika X-ray score," an integrative parameter of pulmonary TB pathology. In comparison with healthy Mtb-exposed controls, TB patients (TBP) did not exhibit deficiency in Mtb-specific cytokine-producing CD4+ cells circulating in the blood and differed by a polyfunctional profile of these cells, which was biased toward the accumulation of bifunctional TNF-α+IFN-γ+IL-2- lymphocytes. Importantly, however, severity of different TB manifestations was not associated with Mtb-specific cytokine-producing cells or their polyfunctional profile. In contrast, several TB manifestations were strongly correlated with leukocyte numbers, the percent or the absolute number of lymphocytes, segmented or band neutrophils. In multiple alternative statistical analyses, band neutrophils appeared as the strongest positive correlate of pulmonary destruction, bacteria excretion, and "Timika X-ray score." In contrast, clinical TB severity was primarily and inversely correlated with the number of lymphocytes in the blood. The results suggest that: (i) different TB manifestations may be driven by distinct mechanisms; (ii) quantitative parameters and polyfunctional profile of circulating Mtb-specific CD4+ cells play a minor role in determining TB severity; and (iii) general shifts in production/removal of granulocytic and lymphocytic lineages represent an important factor of TB pathogenesis. Mechanisms leading to these shifts and their specific role during TB are yet to be determined but are likely to involve changes in human hematopoietic system.

When an individual is exposed to Mycobacterium tuberculosis (Mtb) three outcomes are possible: bacterial clearance, active disease, or latent infection. It is generally believed that most individuals exposed to Mtb become latently infected and carry the mycobacteria for life. How Mtb is maintained during this latent infection remains largely unknown. During an Mtb infection in mice, there is a phase of rapid increase in bacterial numbers in the murine lungs within the first 3 weeks, and then bacterial numbers either stabilize or increase slowly over the period of many months. It has been debated whether the relatively constant numbers of bacteria in the chronic infection result from latent (dormant, quiescent), non-replicating bacteria, or whether the observed Mtb cell numbers are due to balance between rapid replication and death. A recent study of mice, infected with a Mtb strain carrying an unstable plasmid, showed that during the chronic phase, Mtb was replicating at significant rates. Using experimental data from this study and mathematical modeling we investigated the limits of the rates of bacterial replication, death, and quiescence during Mtb infection of mice. First, we found that to explain the data the rates of bacterial replication and death could not be constant and had to decrease with time since infection unless there were large changes in plasmid segregation probability over time. While a decrease in the rate of Mtb replication with time since infection was expected due to depletion of host's resources, a decrease in the Mtb death rate was counterintuitive since Mtb-specific immune response, appearing in the lungs 3-4 weeks after infection, should increase removal of bacteria. Interestingly, we found no significant correlation between estimated rates of Mtb replication and death suggesting the decline in these rates was driven by independent mechanisms. Second, we found that the data could not be explained by assuming that bacteria do not die, suggesting that some removal of bacteria from lungs of these mice had to occur even though the total bacterial counts in these mice always increased over time. Third and finally, we showed that to explain the data the majority of bacterial cells (at least  60%) must be replicating in the chronic phase of infection further challenging widespread belief of nonreplicating Mtb in latency. Our predictions were robust to some changes in the structure of the model, for example, when the loss of plasmid-bearing cells was mainly due to high fitness cost of the plasmid. Further studies should determine if more mechanistic models for Mtb dynamics are also able to accurately explain these data.

While there are many opinions on what mathematical modeling in biology is, in essence, modeling is a mathematical tool, like a microscope, which allows consequences to logically follow from a set of assumptions. Only when this tool is applied appropriately, as microscope is used to look at small items, it may allow to understand importance of specific mechanisms/assumptions in biological processes. Mathematical modeling can be less useful or even misleading if used inappropriately, for example, when a microscope is used to study stars. According to some philosophers (Oreskes et al., 1994), the best use of mathematical models is not when a model is used to confirm a hypothesis but rather when a model shows inconsistency of the model (defined by a specific set of assumptions) and data. Following the principle of strong inference for experimental sciences proposed by Platt (1964), I suggest "strong inference in mathematical modeling" as an effective and robust way of using mathematical modeling to understand mechanisms driving dynamics of biological systems. The major steps of strong inference in mathematical modeling are (1) to develop multiple alternative models for the phenomenon in question; (2) to compare the models with available experimental data and to determine which of the models are not consistent with the data; (3) to determine reasons why rejected models failed to explain the data, and (4) to suggest experiments which would allow to discriminate between remaining alternative models. The use of strong inference is likely to provide better robustness of predictions of mathematical models and it should be strongly encouraged in mathematical modeling-based publications in the Twenty-First century.