Pulmonary Tuberculosis in Humanized Mice Infected with HIV-1

Co-infection with HIV increases the morbidity and mortality associated with tuberculosis due to multiple factors including a poorly understood microbial synergy. We developed a novel small animal model of co-infection in the humanized mouse to investigate how HIV infection disrupts pulmonary containment of Mtb. Following dual infection, HIV-infected cells were localized to sites of Mtb-driven inflammation and mycobacterial replication in the lung. Consistent with disease in human subjects, we observed increased mycobacterial burden, loss of granuloma structure, and increased progression of TB disease, due to HIV co-infection. Importantly, we observed an HIV-dependent pro-inflammatory cytokine signature (IL-1β, IL-6, TNFα, and IL-8), neutrophil accumulation, and greater lung pathology in the Mtb-co-infected lung. These results suggest that in the early stages of acute co-infection in the humanized mouse, infection with HIV exacerbates the pro-inflammatory response to pulmonary Mtb, leading to poorly formed granulomas, more severe lung pathology, and increased mycobacterial burden and dissemination.

reviews 21,22 . In the current studies we applied our HuMouse model of experimental TB to investigate pulmonary Mtb/HIV-1 co-infection. We chose to model one of several important clinical scenarios, that of new pulmonary Mtb infection in subjects with an existing HIV infection. Recent reports indicate that in HIV + subjects, Mtb infections are frequently a new event rather than reactivation of latent infections [23][24][25] . We observed that HIV-infected cells were localized to sites of Mtb-driven inflammation and mycobacterial replication in both human and HuMouse co-infected lung. We also observed increased mycobacterial burden and loss of granuloma structure due to HIV co-infection, consistent with disease in human subjects. Further, we show neutrophil infiltration and tissue necrosis due to HIV in the co-infected lung and increased production of pro-inflammatory cytokines and chemokines that recruit neutrophils. These results suggest that in the early stages of acute pulmonary co-infection, HIV infection compromises antibacterial functions in the lung and activates non-protective inflammatory responses and neutrophil influx that can promote tissue damage.

HIV replication in the Mtb-infected BLT humanized mouse. To emulate human HIV infection in BLT
HuMice, animals were given 2,500 TCID 50 of the JR-CSF strain of HIV-1 i.v. as described 16 . At 3 wk p.i. with HIV, some groups of animals were infected i.n. with 250 CFU of the H37Rv strain of Mtb as described 20 . Replication of HIV, based on detection of the p24 capsid protein by an ELISA assay employed in clinical settings to determine HIV status, was observed in the plasma and spleen at 8 wk post-HIV infection (Fig. 1A). Expression of p24, though variable, was detectable in HIV-and HIV/Mtb-infected mice from two independent studies. Co-infection with Mtb did not significantly alter viral load at this stage of infection (5 wk post-Mtb, and 8 wk post-HIV, infection). As previously observed in HuMice infected with the JR-CSF strain of HIV-1, CD4 + T-cell numbers decline slowly 26 . The moderate decrease in CD4 + T-cells observed in HIV-infected animal groups did not reach significance compared to non-infected and Mtb-infected animals (Fig. 1B).

HIV p24 + cells localize to Mtb lesions in the lung. The role of HIV-infected cells in local immune
suppression at the site of Mtb containment in protective granulomas is an important unknown in co-infection biology. We used hematoxylin and eosin (H&E) and immunohistochemistry (IHC) to assess tissue pathology associated with the presence and distribution of HIV + cells in Mtb-co-infected human and HuMouse lung tissue (Fig. 2). In the absence of HIV, the lesions observed in the lung of Mtb-infected animals were characteristic of TB where solid lesions arise from inflammation in the interstitium in response to Mtb-infected macrophages as shown with representative lung in Fig. 2A. In contrast, a much greater inflammatory cell influx with involvement of alveolar spaces was observed in the lung of Mtb/HIV co-infected animals ( Fig. 2A). Histological examination of the lung tissue from HuMice infected with only HIV revealed no remarkable pathology compared to non-infected animals ( Fig. 2A).
We found that HIV + cells trafficked to the lung in HuMice infected with HIV or co-infected with HIV and Mtb. In the absence of Mtb infection, HIV p24 + cells were occasionally found and were randomly distributed in the lung parenchyma (Fig. 2B) as described 27 . In a subset of co-infected HuMice, HIV + cells were observed at sites of Mtb infection-driven inflammatory foci (Fig. S1A) at 3 wk p.i. with Mtb. Following 5 wk of co-infection, we observed that HIV + cells were localized almost exclusively to the sites of TB lesions including within developing pulmonary granulomas (Fig. 2B). The distribution of HIV + cells in co-infected HuMouse lung was similar to what we observed in co-infected human lung reference tissues. As shown with representative tissues from Mtb/ HIV co-infected human lung from autopsy cases (Fig. 2B), we found that HIV + cells are localized primarily to the lesion periphery in both newly forming solid lesions as well as in more progressed and necrotic TB granuloma.

Pulmonary growth of Mtb is increased by HIV co-infection prior to peripheral CD4 + T cell depletion.
To determine the impact of HIV co-infection on the growth and dissemination of Mtb in HuMice, we measured colony-forming units (CFU) of mycobacteria in disrupted lung and liver tissue following necropsy. Analysis of randomly selected animals at 3 wk p.i. with Mtb revealed moderate mycobacterial growth in the lung that was similar among animals with mock or HIV co-infection ( Fig. S1B). At 5 wk p.i., however, Mtb load was significantly increased (> 1 log) in the lung of animals co-infected with HIV ( Fig. 3A) in two independent groups of animals reconstituted with tissues from different human donors. Increased bacterial load due to HIV infection was also observed in the liver, though the change did not reach significance (p = 0.06) in one experimental group. These results were verified by visualization of AFB numbers using light microscopy (Fig. 3B). As shown in Fig. 3B relative to a similarly sized TB lesion, few AFB are observable in the lesions of an Mtb-infected HuMouse lung while greater AFB staining is observable in the co-infected lung. To determine if there was a positive relationship between relative lung viral load and bacterial burden among the co-infected animals, a Spearman's Correlation analysis was performed. As shown in Fig. 3C, there was a positive, though weak, relationship between HIV p24 levels and mycobacterial proliferation in lung tissue. A moderately positive relationship, however, was observed among HIV p24 and the % of granulomatous tissue in the co-infected animals (Figs 3C and 4B). These positive correlations were not statistically significant with this sample size.
HIV co-infection exacerbates pulmonary TB pathology. A limited number of observational studies of human lung has suggested that HIV disrupts TB granuloma architecture 13,14 , but controlled experiments still are needed given the paucity of animal models. To determine how HIV alters TB pathology in the HuMouse, we compared lung tissue of Mtb and Mtb-HIV co-infected mice. Analysis of lung tissue from a subset of Mtb-or Mtb/ HIV-infected HuMice harvested at 3 wk p.i. with Mtb revealed mostly non-remarkable pathology with only occasional foci of inflammation observed at sites of Mtb proliferation (Fig. S1). Surprisingly, a greater cellular influx was observed at these infrequent sites of Mtb-driven inflammation in HuMice with HIV co-infection (Fig. S1A). At this stage of disease in the HuMouse model the mycobacterial burden did not differ among HIV + and HIVanimals (Fig. S1B). By 5 wk p.i., analysis of the gross lung demonstrated that co-infected mice had more numerous TB lesions compared to those infected with only Mtb (Fig. 4A). The lesion size varied among lesions within the lung, but overall the lesions in co-infected mice were larger (Fig. 4A). As a result, the percent of the lung occupied by granulomatous tissue was significantly greater in co-infected animals (Fig. 4B). These larger lesions exhibited decreased organization of the granuloma structure due to the presence of diffuse cellular aggregations at the lesion site versus the smaller and more compact lesions observed in animals infected with Mtb alone (Fig. 4C). Importantly, the lesions from co-infected animals had a less discriminate distribution of AFB relative to lung pathology (Fig. 4C). In the lung of animals infected with Mtb, AFB were fewer and were localized primarily within solid lesions ( Fig. 3B), suggesting successful containment by the host immune system. In contrast, AFB + cells were observed in co-infected lung within the granulomas, but also within macrophages in the alveolar spaces adjacent to poorly defined lesion peripheral zones (Figs 3B and 4C). Increased tissue necrosis accompanied by areas of nuclear karyorrhectic debris, was also present in co-infected animals (Fig. 4D). Consistent with the reported role of necrotic tissue to promote mycobacterial growth 28,29 , large numbers of AFB were visualized within these necrotic tissue zones (Fig. 4D).

Co-infection with HIV promotes neutrophil accumulation in pulmonary TB granulomas.
To determine the basis for exacerbated lung pathology due to co-infection, we investigated neutrophil influx in lung tissue of HuMice and of human reference tissues. Inflammatory lesions most similar in size and stage of organization were chosen through H&E visualization of pathology for assessment alongside tissue from HuMice (Fig. 5A). Neutrophils were stained using titration of an antibody specific to myeloperoxidase (MPO) (Fig. 5B). In the HuMouse, at 5 wk pi we observed solid and well organized lesions (Figs 4C and 5A) that were similar to smaller lesions found in human lung from a decedent with Mtb infection (Fig. 5A). Moderate numbers of neutrophils were distributed throughout the lung lesions from HuMice and human subjects infected with Mtb only, consistent with prior descriptions in human lung 30 and other animal models of TB [31][32][33] . In contrast, in co-infected lung from both human subjects and HuMice, we observed sites of inflammatory pathology characterized by a poorly organized inflammatory influx (Fig. 5A). Detection of MPO revealed a marked increase in the numbers of neutrophils localized to these poorly organized inflammatory areas in co-infected tissue (Fig. 5B). The specificity of staining in these necrotic tissues was confirmed using a non-specific primary antibody as a control (Fig. 5B, insert).  plus HIV. As shown in Fig. 6, the expression of most cytokines and chemokines in animals infected with only HIV were similar to the baseline presence observed in non-infected control animals (Fig. 6). The exceptions were a significant increase in MCP-1 and a trend towards increased IL-6 due to HIV infection. Consistent with the non-significant reduction of CD4 + T-cell populations that we observed at this stage of HIV infection (Fig. 1), there were no significant decreases in IFN-γ activation due to HIV co-infection with Mtb. As expected, several cytokines and chemokines were significantly increased in the lung microenvironment due to Mtb infection, including pro-inflammatory cytokines (IL-1-β , IL-6, and TNF-α ) and chemokines (MCP-1, IP-10, Eotaxin 1, and GM-CSF) and IL-12, IFN-γ , and IL-4.

HIV co-infection activates excessive pro-inflammatory cytokine responses to
An exacerbation of pro-inflammatory cytokine and chemokine production, however, was observed in the pulmonary micro-environment due to co-infection. Significantly increased production of IL-1β , IL-6, and IP-10 was observed in co-infected lung as compared to Mtb-infected lung. Significant increases in IL-8 and MIP-1α were observed in co-infected lung while these molecules were not increased following Mtb or HIV mono-infection. Activation of IL-10 occurred following Mtb infection and was further increased due to co-infection, although this effect of co-infection failed to reach significance (p = 0.07) due to variability among animals.

Pulmonary pneumonia and vascular occlusions develop as complications of HIV/Mtb co-infection.
Interestingly, TB pneumonia and vascular occlusions were noted in several animals with co-infection, while similar pathologies were not seen in in mono-infections. As shown in Fig. 7, alveolar spaces in co-infected HuMouse lung contained inflammatory cells and exudate consistent with TB pneumonia identified in infected human tissues. The development of TB pneumonia in immunocompromised persons including those with HIV/AIDS has been described 22 . Observation of this feature is an important endpoint that supports the validity of the HuMouse model to study TB/HIV. Additionally, infiltration of small vessels by mononuclear cells with endothelialitis and luminal narrowing in co-infected HuMouse lung was also observed. To confirm this finding, a pentachrome stain (Movat) was employed which identifies nuclei, elastin, cartilage, reticular fibers, mucin, fibrin, and muscle. Visualization of the black elastin fibers definitively identified vascular occlusions as shown within a representative tissue in Fig. 7A. Similar vascular occlusions were identified in lung from our co-infected human lung reference tissues as illustrated in Fig. 7B. Development of vascular occlusions is a pathology that has previously been noted in advanced stages of human TB 34-38 , but has not been associated with HIV to our knowledge. Development of bronchiole obstructions in the lung of co-infected mice (Fig. 7C)

Discussion
Evidence-based studies that can inform development of novel countermeasures to reduce and treat TB/HIV are critically needed. Our description of dual disease here is the first, to our knowledge, to demonstrate the application of HuMice in understanding the pathogenesis of TB/HIV, or of any viral and bacterial co-infection. In the absence of Mtb, infection with HIV in HuMice led to systemic viral propagation and random distribution of HIV + cells to the lung interstitium. In contrast, a definitive localization of HIV + cells to the periphery of TB granulomas was observed in HuMouse and human lung as previously described in a rhesus macaque model of SIV/TB 11 . These are important observations as a small number of HIV-infected cells can exert local immune modulatory activity through the effects of cytokines (e.g IL-10) and potentially toxic HIV proteins (e.g. gp120, tat, nef) 41,42 . These observations provide evidence to support the theory that the inflammatory response to Mtb infection may recruit HIV + cells to the granuloma, which in turn disrupts local protective innate and acquired immune function required at this critical site for containment of Mtb in the host lung.
HIV infection in HuMice markedly altered the host immune containment of pulmonary Mtb infection similar to the accelerated progression of TB disease in people with HIV 25 . Infection with HIV may be disrupting innate immune processes involved in bacterial killing or antigen presention by macrophages as demonstrated using ex vivo analysis of human cells [43][44][45] or by impairing T cell helper and/or cytotoxic activity that restricts Mtb growth in macrophages 4,46-48 . In our model of co-infection, we observed in vivo effects prior to significant loss of CD4 + T cells in peripheral blood; the latter is a clinical indicator of generalized immune suppression in human subjects with HIV 47,49 . At this early stage of infection, we did not observe increased HIV replication when assessing peripheral and splenic viral loads as has been observed in the BAL fluid of some human subjects with active TB 50 . Consistent with the limited observations of co-infected human lung 13,14 we found that disorganized Mtb lesions develop in the setting of HIV infection in the HuMouse. Our results, however, demonstrate that diffuse and poorly organized Mtb lesions can develop from the onset of acute co-infection. This is important, as the effect of HIV to compromise TB host defense is postulated to be due to HIV-mediated deterioration of the already developed protective granuloma structure, leading to bacterial escape and dissemination 9 . That scenario is most likely to apply to patients with latent TB, or a resolving TB infections, who later become infected with HIV. In those with an active HIV infection who subsequently are infected with Mtb, the initial results in the HuMouse model imply that de novo formation of the granuloma is compromised and perhaps occurs prior to lowering of CD4 + T cell counts in peripheral blood.
A surprising finding is that inappropriate or dysregulated immune responses, as opposed to immune suppression, can occur at an early stage of co-infection. It is important to note that the current studies were not extended to include late stages of disease characterized by CD4 + T cell loss and AIDS. In support of an earlier morphological description of cellular influx in a co-infected lung 14 , our analysis revealed neutrophils as the definitive inflammatory cell type accumulating in the lung of co-infected human subjects, and HuMice, at sites of Mtb proliferation. In the absence of HIV infection, neutrophils have a well-defined protective role during the early immune response to Mtb 32,51 but are associated with pulmonary pathology when the infection fails to resolve 52,53 . The tissue destruction mediated by neutrophil degranulation and the accumulating debris from dying neutrophils provides an environment that is believed to be favorable for Mtb proliferation 21 . Our findings introduce a mechanistic concept that could explain why increased blood neutrophil counts in HIV co-infected patients is associated with greater TB disease severity 54 . Our studies have also suggested candidate inflammatory mediators (e.g. IL-1β , CXCL5, IL-8) that could promote excessive neutrophil recruitment in co-infected lung. Interestingly, these effects are observed in the absence of HIV-associated defects in several cytokines (IFN-γ , IL-12, and TFN-α ) that play important roles in TB immunity 55,56 . The development of vascular occlusions, bronchial obstructions, and bacterial pneumonia that we observed in co-infected animals and human reference tissue is consistent with the effects of exacerbated inflammation. Bacterial pneumonia is known to occur in people who are immune compromised who become infected with Mtb 22 , while vascular occlusions and bronchial obstructions are features observed in advanced TB disease 35,37,38 . We have previously reported development of bronchial obstructions and pneumonia in HuMice at later stages of Mtb infection (> 7 wk p.i.) or following infection with a larger i.n. Mtb dose 20 . Our results in the HuMouse, then, suggest that HIV co-infection may accelerate the TB disease process by promoting specific aspects of the inflammatory process.
In summary, these studies describe a novel model that has been successfully applied to increase our understanding of TB/HIV co-infection pathobiology. These initial findings may provide a basis for the conflicting evidence for both immune activation and immune suppression in co-infected human subjects. Speculatively, co-infection may drive non-protective inflammation during early stages of HIV disease while immune compromise, including loss of IFN-γ , predominates once CD4 + T cell depletion ensues. In future studies it will be important to determine if these features of inflammatory pathology are promoted by the increased bacterial load in the HIV-compromised host, or if HIV-driven inflammation may play a more direct role in supporting the Mtb growth that subsequently accelerates TB disease progression. This is quite intriguing given recent studies that demonstrate inflammatory processes may augment M.tb proliferation 57,58 independent of the effect of tissue destruction to provide a favorable environment for Mtb growth. Long term, the HuMouse may serve as an important pre-clinical model to identify mechanisms of co-infection pathobiology and test therapeutic interventions prior to use in NHPs or human subjects. Animal Experiments. Humanized BLT mice were generated as we have previously described 20   as M-tropic and infects both monocyte/macrophage and CD4 + T cells. HIV was diluted in PBS and mice were infected with a 2,500 TCID 50 via tail vein infection as described 17 . Infection with the H37Rv (tdTomato) strain of Mtb 59 was performed using 250 CFU diluted in 40 ul of Dulbecco's Phosphate-Buffered Saline (PBS, Cellgro, Manassas, VA, USA) via an intranasal route (20 ul/nare) as described 20 . Animal experiments were repeated in an independent group generated from an additional human tissue donor. Humanized mice are challenging to generate and animal numbers were a limitation. Two independent studies with non-infected animals (n = 2/ study), Mtb-infected (n = 3/study) and Mtb/HIV-infected (n = 4/study) were performed. Three HIV-infected control animals were included only in one study. All animal experiments and work with Mtb were performed in a CDC-approved animal biological safety level-3 (ABSL-3) and BSL3 facilities in the Galveston National Laboratory in accordance with biosafety procedures approved by the UTMB Environmental Health and Safety Division.

Materials and Methods
Bacterial and Viral Load. Following aseptic removal at 5 wk post Mtb infection, organs (lung and liver) were placed into 1 ml of PBS in 15 ml small tissue grinders (Kendall, Mansfield, MA, USA) for tissue homogenization. Serial dilutions of homogenized organ samples were prepared in PBS to perform limiting dilution CFU enumeration as described 60 . All studies were performed in a CDC-approved biological safety level-3 (BSL-3) facility. For viral quantification, plasma and supernatants from disrupted spleen were stored frozen at − 80 °C. Plasma and splenic supernatants were then used to detect HIV p24 capsid protein using a clinical ELISA test (Zeptometrix, Buffalo, NY). Levels of p24 were quantified by generating a standard curve using the standards provided in the kit. Values were generated by linear regression to the standard curve as recommended by the manufacturer. Immunohistochemistry. Lung tissue sections from formalin-fixed, paraffin embedded HuMouse or human lung were deparaffinized and washed in 0.05% PBST. Following deparaffinization, antigen retrieval was performed by 25 min incubation in citrate buffer, pH 6 (Dako) at 95-100 °C in a rice steamer. Samples were then washed with 0.05% PBST and incubated with monoclonal antibodies to MPO (at a dilution of 1:50; Abcam; ref ab45977-100) for two hours at room temperature or HIVp24 (at a dilution of 1:50; Dako; clone Kal-1) overnight at 4 °C. Sections were stained with Permanent Red using the Dako EnVision System with Mayer's Hematoxylin as the counter stain. Images were taken using an Olympus BX53 microscope equipped with an Olympus DP71 camera.
Pathology. At specified time points, following aseptic removal of organs, the remaining half of tissues were placed in 10% Neutral Buffered Formalin (Statlab, McKinney, TX, USA) for 48 hours to inactivate infectious agent, changed after 24 hours, and finally stored in 70% ethanol. Tissues were embedded in paraffin then stained with hematoxylin and eosin (H&E) and additional sections were stained using the Ziehl-Neelson method to visualize acid-fact bacteria (AFB) to detect M.tb bacilli. A 5 color chromagen (Movats) stain was performed to visualize vascular pathology including elastin, collagen, fibrin, mucin, and muscle fibers. Processing and staining for H&E, AFB, and Movats was performed by the University of Texas Medical Branch, Research Histopathology Core Facility. Lung tissue was evaluated by a trained pathologist with expertise in tuberculosis disease progression, with confirmation done in a slide blinded manner. Granuloma number was quantified from the entire section of mouse lung. The average size of each granuloma and the total amount of granulomatous tissue in the whole lung was quantified using digital software (NIH Image J; developed at the U.S. National Institutes of Health and available on the Internet at http://rsb.info.nih.gov/nih-image/). The percent of granulomatous tissue was calculated as a percent of total lung area as described 61 . reactivity between human detection reagents and mouse cytokines could confound data (e.g. VEGF, IL-13) were excluded from the analysis. Levels of each cytokine were quantified by generating a standard curve using the standards provided in the kit. Values were generated by linear regression to the standard curve as recommended by the manufacturer and as described 62 . Values that were below the extrapolated range (< OOR) were set to 0.
Statistics. Data are shown as mean ± SEM. One-way ANOVA followed by a Dunnetts multiple comparison test was used for multiple group comparisons (GraphPad Software v5.0). Statistically significant values are designated as follows: *p < 0.05; **p < 0.01; ***p < 0.001. The strength of association between viral load and bacterial burden or lung pathology was computed using the Spearman correlation coefficient (GraphPad Software v5.0).