Host resistance to Mycobacterium tuberculosis (Mtb) infection requires the activities of multiple leukocyte subsets, yet the roles of the different innate effector cells during tuberculosis are incompletely understood. Here we uncover an unexpected association between eosinophils and Mtb infection. In humans, eosinophils are decreased in the blood but enriched in resected human tuberculosis lung lesions and autopsy granulomas. An influx of eosinophils is also evident in infected zebrafish, mice, and nonhuman primate granulomas, where they are functionally activated and degranulate. Importantly, using complementary genetic models of eosinophil deficiency, we demonstrate that in mice, eosinophils are required for optimal pulmonary bacterial control and host survival after Mtb infection. Collectively, our findings uncover an unexpected recruitment of eosinophils to the infected lung tissue and a protective role for these cells in the control of Mtb infection in mice.
Introduction
Mycobacterium tuberculosis (Mtb) is an intracellular bacterium that causes tuberculosis (TB), which through 2020 accounted for the highest yearly global mortality due to a single pathogen (World Health Organization, 2020). Mtb primarily infects innate immune cells, and host resistance depends on antimicrobial effector functions of infected macrophages alongside a robust type I and T helper 1 cell immune response (Cooper et al., 1993; Ernst, 2012; Flynn et al., 1993; Flynn et al., 2015). Eosinophils, in contrast, are major innate effector cells in type II inflammatory settings (Klion and Nutman, 2004; Simon et al., 2020). Accordingly, reports of type II immunity-associated cells and responses during Mtb infection are rare, with one report of eosinophils in lungs of Mtb-infected guinea pigs (Lasco et al., 2004) and one case series of pulmonary eosinophilia in three TB patients (Vijayan et al., 1992). As a result, eosinophils have been assumed to be largely uninvolved in the immune response to Mtb or actively repressed by the strong IFN-γ–driven responses required for control of mycobacteria (Cooper et al., 1993; Flynn and Ernst, 2000; Kirman et al., 2000; O’Garra et al., 2013).
Studies of eosinophil responses in vivo have historically focused on allergic responses and parasitic helminth infections (Klion et al., 2020; Travers and Rothenberg, 2015). Eosinophils are associated with an established type II response during lung infections with respiratory syncytial virus and Aspergillus (Lilly et al., 2014; Phipps et al., 2007), and recent data have highlighted complex roles for eosinophils in tissue homeostasis and remodeling, barrier function, and wound healing (Lee et al., 2010; Rosenberg et al., 2013; Shah et al., 2020). Indeed, eosinophils are pluripotent, armed with a variety of bioactive molecules stored preformed in granules or synthesized in lipid bodies. These include a large repertoire of cytokines, chemokines, lipid mediators, and cationic granule proteins, including eosinophil cationic protein, eosinophil-derived neurotoxin, eosinophil peroxidase (EPX), and major basic protein (Acharya and Ackerman, 2014; Klion et al., 2020; Shamri et al., 2011). However, our understanding of the in vivo function of eosinophils during bacterial infections is limited. Reports are currently restricted to extracellular bacteria where direct cell-autonomous bactericidal properties of eosinophils, including extracellular DNA traps and bactericidal granule proteins, can promote extracellular clearance of Escherichia coli, Citrobacter rodentium, Staphylococcus aureus, and Pseudomonas aeruginosa bacteria (Arnold et al., 2018; Krishack et al., 2019; Linch et al., 2009; Yousefi et al., 2008). However, the physiological role of eosinophils in host resistance to intracellular bacterial infections remains unknown.
Here we establish that eosinophils are a significant cellular component of TB granulomas and Mtb-infected lungs in mice, nonhuman primates (NHPs), and patients. Using complementary genetic approaches to create eosinophil deficiency, we demonstrate that eosinophil responses are required for optimal host resistance and survival after Mtb infection in mice. Together, our findings reveal a previously unrecognized association between eosinophils and TB in multiple hosts and uncover a protective role for these cells in mice.
Results and discussion
We started by investigating whether eosinophils participated in the granulocytic response to TB by quantifying circulating eosinophils in three independent clinical cohort studies (Du Bruyn et al., 2021; Lowe et al., 2013; Mayer-Barber et al., 2014). We asked whether the numbers of peripheral blood eosinophils of TB patients changed based on disease status and antitubercular treatment (ATT) in a South African cohort (Du Bruyn et al., 2021). Indeed, blood eosinophil numbers were significantly higher in individuals with latent TB infection (LTBI) compared with active pulmonary TB (PTB) patients (Fig. 1 A), with a corresponding decreased eosinophil-to-neutrophil (E/N) ratio in the PTB group (Fig. S1, A and B). When patients in an independent Chinese cohort (Mayer-Barber et al., 2014) were stratified according to disease severity based on sputum positivity for acid-fast staining of bacilli (AFB), individuals with active AFB+ PTB and extrapulmonary TB (EPTB) disease exhibited a significantly decreased E/N ratio compared with AFB− PTB patients (Fig. S1 C). These data suggest that a decrease in circulating E/N ratio may reflect an increase in sputum bacterial loads. Consistent with the hypothesis that blood eosinophil numbers reflect bacterial sputum status and disease burden, circulating blood eosinophil numbers increased significantly within 2 wk of commencing ATT in both cohorts (Fig. 1 B and Fig. S1 D). Moreover, in a third British cohort that included patients with both PTB and EPTB (Lowe et al., 2013), TB disease survivors exhibited a significantly higher E/N ratio on blood tests taken at the time of diagnosis compared with individuals who subsequently died (Fig. S1 E). In fact, lower baseline eosinophil numbers in blood correlated with shorter time to TB-related death (Fig. S1 F). Taken together, our data from three independent clinical cohorts suggest that peripheral blood eosinophils are decreased according to disease severity, raising the possibility that rather than being depleted, eosinophils may be recruited to distant disease sites where they could participate in the local immune response to infection.
To directly address whether eosinophils could be part of the local pulmonary granulocyte response in human TB, we measured the relative abundance of granulocytes in human TB lesions. To this end, we generated single-cell suspensions from freshly resected human lung tissue (n = 9 TB patients) directly after clinically indicated surgery and identified tissue-resident granulocytes with established markers Siglec-8+/CD66b+ for eosinophils and Siglec-8−/CD66b+ for neutrophils (Fig. 1 C and Fig. S1 G). Neutrophils were abundant CD45+ immune cells in human lung lesions divided into different macropathologies (Fig. S1 H). However, eosinophils were also frequently present in appreciable numbers relative to neutrophils in the same lung lesion sample (Fig. 1 D and Fig. S1 H). In fact, E/N ratios and eosinophil frequencies of CD45+ immune cells were sometimes similar to or even higher than neutrophils in some, but not all, samples from fibrotic areas and granulomas compared with the blood in the same individuals (Fig. 1, D and E; and Fig. S1, H and I). We then asked whether eosinophils would selectively infiltrate metabolically active inflamed lung lesions via 2-deoxy-2-fluorine-18-fluoro-D-glucose (18FDG) positron emission tomography (PET)/computed tomography (CT) since the more recent patients received 18FDG PET/CT scans to aid in defining the surgical regions before resection. When we separated individual lesions based on high or low metabolic activity via 18FDG signal intensity, eosinophils were significantly increased relative to neutrophils in PET-high compared with PET-low lung lesions (Fig. 1 F and Fig. S1 J). These data suggest that eosinophils are more abundant in PET FDG-avid areas compared with areas of low metabolic activity and disease. While larger studies are needed to accurately categorize and associate the abundance of specific granulocyte subsets within diverse TB lesion types, our small retrospective study showed that eosinophils can be substantially enriched in individual human TB lesions. Most importantly, the E/N ratio was significantly higher in resected human lung lesion types compared with circulation in the same patients (Fig. 1 G), supporting our hypothesis that peripheral blood eosinophils decrease during active TB disease and enrich at infected tissue sites. While a functional role for eosinophils in human disease remains unclear, our clinical data support the hypothesis that during TB, circulating peripheral blood eosinophils are dynamically regulated based on disease status and that eosinophils can be enriched at the Mtb-infected local tissue site.
To confirm the flow cytometric data, we next examined human TB granuloma sections from an independent autopsy cohort (Fig. 2 A; Blauenfeldt et al., 2018). H&E staining indicated that eosinophils were present primarily in the rim area of the granulomas (Fig. S2 A), whereas diffuse staining for the eosinophil granule protein EPX was observed in the necrotic core of granulomas, suggesting eosinophil degranulation in the center of TB granulomas (Fig. 2 A and Fig. S2 A). To experimentally explore the role of eosinophils, we employed an NHP model in which Mtb infection forms bona fide lung granuloma structures closely resembling those found in humans (Basaraba and Hunter, 2017; Flynn et al., 2015). We initially assessed lung granulomas from Mtb-infected macaques at 7–12 wk after infection for eosinophil infiltration (Fig. S2 B). Like the human TB granulomas, eosinophils were evident in the outer rim area in addition to more diffuse EPX staining in the necrotic core of some granulomas, suggesting degranulated eosinophils (Fig. 2 B; and Fig. S2, C and D). We then asked whether eosinophil recruitment to mycobacterial granulomas was an evolutionarily conserved process. Using the zebrafish model, we assessed eosinophils in mycobacterial granulomas formed after Mycobacterium marinum infection (Parikka et al., 2012; Swaim et al., 2006). Indeed, in agreement with our findings in NHP and human TB granulomas, eosinophils were also found in necrotic granulomas in M. marinum–infected zebrafish (Fig. 2 C). Taken together, our data show that eosinophils are present in mycobacterial granulomas in human, NHP, and zebrafish, suggesting that infiltration of eosinophils into granulomas may be a conserved response to mycobacteria across species.
To measure eosinophil activation and infiltration into granulomas, we examined NHP granulomas in greater detail. To quantify and functionally assess eosinophils at the single-cell level in rhesus macaques, we developed a flow cytometric approach based on EPX (Fig. S2 E). We found that intracellular staining with EPX selectively and specifically stained eosinophils in whole blood (WB) as well as in bronchoalveolar lavage (BAL) fluid (Fig. 3 A). Importantly, when we quantified eosinophils in NHP TB granulomas, we observed a high level of heterogeneity between individual granulomas with eosinophils spanning four orders of magnitude, comprising up to 10% of all CD45+ immune cells with an average of 1.5% in 46 granulomas from seven animals (Fig. 3 B and Fig. S2 F). There was no correlation between individual granuloma bacterial loads and eosinophil abundance (Fig. 3 C). Since diffuse EPX staining in the core of human and NHP granulomas was suggestive of functional activation resulting in degranulation, we next quantified eosinophil degranulation by measuring CD63 surface expression (Fig. 3 D). While blood eosinophils showed little change in CD63 expression, BAL and granuloma eosinophils exhibited significantly increased CD63 expression, with the highest CD63 expression in granulomas (Fig. 3 D). The proportion of CD63+ eosinophils varied widely between individual granulomas, even within the same animal, and was strongly negatively correlated with the abundance of eosinophils (Fig. 3 E), supporting the idea that eosinophil degranulation precedes cell death. Importantly, CD63 expression on eosinophils inversely correlated with bacterial burden in granulomas (Fig. 3 F), suggesting that functional activation of eosinophils in the form of degranulation may impact bacterial growth in Mtb lesions. Collectively, our data from Mtb-infected NHP lesions reveal a high degree of heterogeneity in the eosinophilic response and support the hypothesis that eosinophils migrate into TB granulomas where they degranulate and participate in the local innate immune response against Mtb.
As a next step in examining the function of eosinophils during Mtb infection, we quantified eosinophils after low-dose aerosol Mtb infection in the genetically tractable C57BL/6 mouse model (B6). We measured lung tissue resident eosinophils by flow cytometry using an i.v. labeling technique that distinguishes between cells located in the lung vascular capillary bed (i.v.pos) and cells that migrated into the lung tissue parenchyma (i.v.neg; Anderson et al., 2014). When we quantified the number of lung tissue–resident i.v.neg eosinophils after Mtb infection, we observed an increase after 2–3 mo of Mtb infection compared with either uninfected animals or 30 d postinfection (d30pi; Fig. 4 A; and Fig. S3, A and B). To ask whether eosinophils could be directly harboring Mtb, either by infection or phagocytosis, we assessed bacteria-containing eosinophils using fluorescent Mtb-mCherry and found that eosinophils are not infected with Mtb (Fig. 4 B). In fact, Mtb-infected cells have been identified and described in detail (Cohen et al., 2018; Huang et al., 2018; Lee et al., 2020; Pisu et al., 2020; Rothchild et al., 2019; Wolf et al., 2007), and we, like others, have consistently failed to detect significant numbers of Mtb-containing eosinophils (Lee et al., 2020; Rothchild et al., 2019). Thus, while eosinophils may interact with Mtb or Mtb-infected cells in lungs and TB granulomas, they themselves do not appear to represent a cellular niche for Mtb. Based on the increase in lung parenchymal eosinophils after infection and their role in tissue homeostasis (Lee et al., 2010; Rosenberg et al., 2013; Shah et al., 2020), we asked whether the absence of eosinophils after Mtb infection would lead to noticeable changes in lung pathology. To this end, we infected B6 ΔdblGata mice that lack eosinophils due to a targeted deletion of a high-affinity GATA–binding site in the GATA-1 promoter (Yu et al., 2002). However, we did not observe changes in lung gross pathology 3 mo after Mtb infection in B6 ΔdblGata compared with WT mice (Fig. 4 C). We then examined lung-resident immune cells and found no changes in i.v.negMtb-specific CD4+ and CD8+ T cells (Fig. S3, C–F), NK1.1+ cells (Fig. S3 G), neutrophils (Fig. S3 H), interstitial macrophage/dendritic cell (DC) polarization states (Fig. S3 I), alveolar macrophages (Fig. S3 J), or XCR1+ DC1 (Fig. S3 K) in B6 ΔdblGata mice compared with WT mice 3 mo after Mtb infection. Finally, to explore the impact of eosinophil deficiency in an unbiased fashion, we performed whole-lung transcriptional profiling on 3-mo Mtb-infected B6 ΔdblGata or WT mice alongside day 0 (d0) uninfected controls. Relatively few genes were significantly and differentially expressed when we directly compared transcripts from d90pi B6 ΔdblGata and d90pi WT mouse lungs. 16 differentially expressed genes (DEGs) were upregulated and 9 downregulated in Mtb-infected ΔdblGata lungs compared with Mtb-infected B6 control lungs (Fig. S3 L). Intriguingly, most of these DEGs were not associated with immunity to infection but, instead, were linked to neurological disorders and neuronal pathways (Fig. S3 L, pink annotation). To explore infection-induced changes, we analyzed and compared d90pi B6 ΔdblGata and d90pi WT mouse lungs with d0 WT controls. We identified 194 up- and 262 downregulated DEGs in Mtb-infected ΔdblGata lungs alongside 94 up- and 230 downregulated DEGs in Mtb-infected WT B6 lungs (Fig. S3 M). The top 20 DEGs were again enriched in neuronal-, sensory-, and olfactory-associated genes (pink annotation) in addition to genes related to lipid metabolism (green annotation; Fig. 4 D). Consistent with limited changes in genes linked to host resistance, transcriptional module exploration for type I and type II IFNs (Singhania et al., 2019) confirmed the restricted molecular perturbations in Mtb-infected eosinophil-deficient compared with WT lungs (Table S1). Importantly, gene set enrichment analyses (GSEAs) revealed again downregulation of pathways associated with neuronal processes; neurological disorders; and short-chain fatty acid, endocannabinoid, and arachidonic acid metabolism in the lungs of Mtb-infected ΔdblGata mice (Fig. 4 E and Table S1). Eosinophils produce a variety of bioactive arachidonic acid–derived lipid mediators, some of which have been implicated in host resistance to Mtb infection before (Mayer-Barber and Sher, 2015). Furthermore, endocannabinoids and arachidonic acid derivatives interact with sensory neurons in the airways (Bozkurt, 2019; Chesné et al., 2019). One possible interpretation of the neuronal-associated transcriptional changes could be that they reflect alterations in airway sensory neurons or pulmonary neuroendocrine cells of the lung epithelium. Thus, while our transcriptional profiling of Mtb-infected lungs revealed neuronal-associated pathways as being primarily affected by the lack of eosinophils, the biological significance of these transcriptional changes remains unknown. Moreover, the pulmonary–neuronal axis in host resistance against Mtb infection is largely unexplored, yet there is growing appreciation for the importance of the crosstalk between the nervous and immune systems (Veiga-Fernandes and Mucida, 2016). For instance, a recent study found that Mtb-derived sulfolipids can directly bind and activate nociceptive neurons in the lung (Ruhl et al., 2020). Eosinophils have been further shown to alter parasympathetic nerve function and airway sensory nerve density, thereby affecting reflex bronchoconstriction in asthma (Costello et al., 1997; Drake et al., 2018a; Drake et al., 2018b; Fryer and Wills-Karp, 1991; Kingham et al., 2002). Thus, our findings raise the possibility that eosinophils may be involved in pulmonary tissue–immune–nerve crosstalk during chronic infections.
Finally, to test whether eosinophils could influence the outcome of infection, we monitored pulmonary bacterial loads and survival in B6 ΔdblGata mice after Mtb infection. While lung bacterial loads were similar compared with WT mice for the first month, 2 mo after Mtb infection we consistently measured a significant 0.5–1-log increase in the lungs of B6 ΔdblGata mice (Fig. 5 A). Moreover, ΔdblGata mice on either B6 or Balb/C genetic backgrounds succumbed significantly earlier to Mtb infection (Fig. 5 B). To confirm these results in a second, independent genetic model of eosinophil deficiency, we infected transgenic B6 PHIL mice that express cytocidal diphtheria toxin A under the eosinophil-specific EPX promoter (Lee et al., 2004). Like B6 ΔdblGata mice, B6 PHIL mice succumbed significantly earlier than WT mice to Mtb infection (Fig. 5 C). When infected with a high dose of Mtb, both B6 ΔdblGata (Fig. 5 D) and PHIL (Fig. S3 N) mice displayed significantly increased lung bacterial loads at 2–3 mo and succumbed to Mtb infection–induced disease in an infectious dose-dependent manner (Fig. 5, E and F). Thus, the data from eosinophil-deficient mouse lines show that eosinophils play a functional role in optimal host resistance during Mtb infection.
To conclude, we found that eosinophils are recruited to Mtb lesions in humans, macaques, zebrafish, and mice. In human lung lesions, eosinophils were enriched compared with blood and found in TB granulomas. In NHP granulomas, eosinophils were present and functionally activated to degranulate. We observed diffuse granule protein staining in the necrotic cores of both human and NHP granulomas, indicative of eosinophil degranulation. In settings associated with tissue necrosis, Mtb can survive and propagate extracellularly (Basaraba and Hunter, 2017; Ernst, 2012), perhaps affording eosinophils the opportunity to access and kill extracellular Mtb via their characteristic degranulation and DNA trap formation. Moreover, purified human EPX has been reported to induce Mtb lysis in vitro (Borelli et al., 2003). Alternatively, granulocyte degranulation could facilitate necrotic core formation in granulomas. Thus, the described associations of eosinophils with TB granulomas in human and macaques do not allow for conclusions regarding protective or detrimental roles. In fact, it is likely that the biological function of eosinophils in host resistance to Mtb may be influenced by the stage and dose of infection, disease status, and granulomatous features, all factors that vary considerably between the broad host species examined in this study. Nevertheless, we establish here that eosinophils represent an unanticipated integral part of the lung granulocytic response to Mtb across species.
Importantly, our translational approach in mice uncovered an unexpected, yet functional role of eosinophils during murine Mtb infection. Using multiple eosinophil-deficient mouse lines, we show that eosinophils are required for optimal host survival and bacterial control. While the underlying protective mechanisms in mice are currently unclear, they could involve antibactericidal or immunoregulatory effector functions that directly or indirectly affect Mtb growth. However, we show here that neither phagocytosis nor direct infection with Mtb seem likely to be occurring at significant levels in eosinophils. Additional nondirect protective mechanisms are therefore likely and could involve immunomodulatory cell–cell interactions between eosinophils and Mtb-infected macrophages or other pulmonary immune cells, maintenance of lung barrier function, as well as disease tolerance. Such cell–cell interactions could involve production of lipid mediators, as identified in our GSEA, and type II cytokines. In fact, a recent study in zebrafish showed that noncanonical type II immune IL-4 and IL-13 signals played critical roles in mycobacterial granuloma formation and macrophage epithelialization (Cronan et al., 2021). Additionally, our transcriptional profiling revealed changes in neuronal-associated pathways, raising the intriguing possibility that eosinophils may also interact with nonimmune lung cells such as pulmonary neuroendocrine cells and airway sensory neurons. We therefore propose that eosinophil-mediated protection against Mtb likely includes multiple, nonmutually exclusive mechanisms, which remain to be experimentally explored further in mice and validated in interventional studies in NHPs.
Coinfection with parasitic helminths is common in areas with high TB incidence, and clinical studies have produced contradictory results on the effect of parasitic worm infections on TB outcomes (Babu and Nutman, 2016; Rafi et al., 2012). In some instances of improved clinical outcomes with TB–helminth coinfections (Abate et al., 2015; O’Shea et al., 2018; van Soelen et al., 2012), eosinophils may have contributed to improved outcomes. That said, increased susceptibility to bacterial infections has not been reported in clinical trials of eosinophil-depleting antibodies (Pavord et al., 2012; Rothenberg et al., 2008; Straumann et al., 2010), although longer-term data with examination of patients exposed to Mtb are needed, and it is possible that a protective effect of eosinophils is masked in highly heterogenous patient populations.
Taken together, our study represents a multispecies investigation of the eosinophil response to Mtb. Our data reveal that eosinophilic granulocytes, typically associated with type II inflammatory responses, represent an integral part of the granulocyte response to TB, a disease associated with type I immunity. Thus, our findings open up multiple new lines of investigation into the functional relevance of eosinophils during chronic bacterial lung infection and potential new targets for host-directed therapies for TB.
Materials and methods
Clinical study population in Shanghai, China
Human lung resection material was obtained at the Shanghai Public Health Clinical Center (SPHCC) between 2015 and 2019 following approval by the SPHCC Ethics Committee (#2015-S046-02 and #2019-S009-02). All patients provided written informed consent. Clinically indicated lung or pleural tissue resection surgery was performed at SPHCC as an adjunctive therapy for TB (L. Wang, Y. Song). Patients with HIV, other ongoing infections, or immunodeficiency conditions were excluded. Using sterile instruments, resected lung tissue samples from participants ID#1–4 were divided based on macroscopically apparent diverse tissue pathologies, and a minimum of triplicate samples for each pathological subtype was used for cell isolation and flow cytometric analysis. Participants ID#5–9 consented to 18FDG PET/CT signal scanning 1–6 d before clinically indicated surgery to guide resection and minimize risk of postoperative complications, such as increasing bronchial stump and lung resection margin embedding. After resection, lung samples were further dissected, using sterile scissors, based on standardized uptake values (SUVs) into maximum-signal (PET high, SUVmax >5.0) and low-signal (PET low, similar to background SUV) pieces according to mapped 18FDG PET/CT scanning results. Triplicate samples were generated for cell isolation and flow cytometric analysis for PET-high and -low regions from the same individuals. FACS studies were performed onsite by K.D. Mayer-Barber, Z. Hu, and K.-W. Wong, and samples were acquired on an LSRFortessa flow cytometer (BD Biosciences) and analyzed with FlowJo 10 software (FlowJo, LLC).
Participant ID#1
A 28-yr-old nonsmoker male, sputum smear and culture negative, received repeated ATT for 8 yr; drug resistance was not detected. Treatment before surgery: isoniazid, rifampicin (RIF), pyrazinamide, and ethambutol. Surgical indication: presence of cavities and irreversible TB progression with lung destruction. Postoperative pathological diagnosis: TB with large cavity with aspergilloma, extensive fibrotic lesions with a mixture of local cavities, granulomas, consolidations, and caseous necrosis. AFB was 4+ positive, and culture was positive for Mtb.
Participant ID#2
A 22-yr-old nonsmoker male; sputum smear and culture negative, drug resistance was not detected. Treatment before surgery: 4 mo treatment with isoniazid, RIF, pyrazinamide, ethambutol, and levofloxacin. Surgical indication: enlargement of pleural empyema. Postoperative pathological diagnosis: presence of inflammatory cells, fibrosis, coagulative necrosis, and caseous necrosis. Acid-fast smear staining was 2+ positive, and culture was positive for Mtb.
Participant ID#3
A 25-yr-old nonsmoker female; sputum smear and culture negative at time of surgery; sputum- and culture-confirmed diagnosis of PTB in both lungs in 2010 and repeated ATT over 6 yr; drug resistance to streptomycin, isoniazid, levofloxacin, and para-aminosalicylic acid was detected. Despite isoniazid, RIF, pyrazinamide, and ethambutol regimen treatment for 1 yr, reexamination of chest CT showed bilateral lesions were progressing and sputum culture was positive. Treatment before surgery: cycloserine, amikacin, linezolid, clofazimine, moxifloxacin, isoniazid, RIF, pyrazinamide, and ethambutol. Surgical indication: the presence of multidrug resistance and cavities in the lower right lobe of the lung with complication of pneumothorax. Postoperative pathological diagnosis: tissue infiltration of inflammatory cells, epithelioid granuloma formation, caseous necrosis, and AFB (4+), which was not culture positive.
Participant ID#4
A 32-yr-old nonsmoker female; 5 yr of ATT, with extended drug resistance to streptomycin, isoniazid, RIF, and ofloxacin; sputum was 4+ smear positive and culture positive. Surgical indication: presence of extended drug resistance and consolidations and atelectasis in the lower right lobe of the lung. Postoperative pathological diagnosis: inflammatory cell infiltration, epithelioid granuloma formation, caseous necrosis, and AFB (2+), which was culture positive for Mtb.
Participant ID#5
A 20-yr-old nonsmoker female; drug resistance was not detected, positive sputum smear, anti-TB antibody and T-SPOT. Sputum smear was positive in July 2018. Treatment before surgery: 7-mo ATT with isoniazid, RIF, pyrazinamide, and ethambutol. Surgical indication: presence of tuberculous pleurisy–related tumor with increasing size over preceding 3 mo. Postoperative pathological diagnosis: inflammatory cell infiltration, epithelioid granuloma formation, and coagulation necrosis. Acid-fast smear staining was 4+.
Participant ID#6
A 38-yr-old male smoker (15-yr history); sputum smear and culture negative at time of surgery, no drug resistance 9 yr before surgery, patient received 1-yr ATT with isoniazid, RIF, pyrazinamide, and ethambutol after acid-fast smear positive test and CT showing improved TB lesions after treatment. Symptoms relapsed 3 yr before surgery, and he received ATT with isoniazid, RIF, pyrazinamide, ethambutol, levofloxacin, and azithromycin. Surgical indication: presence of repeated TB-related symptoms, including tuberculous pleuritis and cavities. Postoperative pathological diagnosis: epithelioid granuloma and coagulative necrosis. Tissue sample acid-fast smear staining was negative.
Participant ID#7
A 39-yr-old male smoker (20-yr history) with 5-yr history of TB had no drug resistance and negative sputum smear at time of surgery. He was previously treated with an isoniazid, RIF, pyrazinamide, and ethambutol regimen with positive sputum smear in the local hospital in August 2014 for 1 yr. TB recurred in June 2016, and ATT was repeated for 1 yr. In April 2018, the patient developed hemoptysis, and chest CT suggested left upper lobe TB with cavities. He received a third ATT. Treatment before surgery: 15-mo ATT with isoniazid, RIF, pyrazinamide, and ethambutol. TB antibody IgG was positive. Surgical indication: clinical diagnosis of tuberculous and persistent TB cavity despite ATT. Postoperative pathological diagnosis: tissue infiltration of inflammatory cells, granuloma formation, coagulative necrosis, calcification, focal pulmonary hemorrhage, and local fungal infection. Acid-fast smear staining was negative, and periodic acid–Schiff (PAS) staining was positive.
Participant ID#8
In a 57-yr-old male nonsmoker, sputum smear and anti-TB antibody were positive, and RIF resistance was detected by GeneXpert (GXPT; Cepheid). Treatment before surgery: 1-mo ATT with RIF, pyrazinamide, ethambutol, isoniazid, and levofloxacin (stopped due to cutaneous pruritus) and 1-mo treatment with rifapentine. Surgical indication: presence of lung cavity. Postoperative pathological diagnosis: inflammatory cell infiltration, epithelioid granuloma formation, and coagulation necrosis. Fluid from the postoperative lesion was acid-fast smear 3+, GXPT positive for Mtb, and RIF sensitive.
Participant ID#9
A 23-yr-old female nonsmoker was diagnosed with PTB in 2016 (positive sputum smear). ATT with isoniazid, RIF, pyrazinamide, and ethambutol was discontinued after 6 mo; fatigue, night sweats, and large pleural effusion occurred in 2018, and the sputum smear was positive again. No drug resistance was found prior to ATT. Treatment before surgery: 10-mo ATT with isoniazid, RIF, ethambutol, levofloxacin, bicyclol, and amikacin. Surgical indication: clinical diagnosis of tuberculous hydrothorax and destructive pneumophthisis. Postoperative pathological diagnosis: tissue infiltration of inflammatory cells, epithelioid granuloma formation, fibrosis, and coagulative necrosis. GXPT test on intraoperative endocrine-resected tissue revealed RIF-sensitive Mtb complex.
Clinical study populations in Cape Town, South Africa
For eosinophil studies from Cape Town, South Africa, participants were recruited from the Site B Community Health Centre in Khayelitsha between March 2017 and December 2018. Some study participants are described elsewhere (Du Bruyn et al., 2021) and in Table 1. Briefly, all participants were HIV-uninfected adults (age ≥ 18 yr) who provided written informed consent. The study was approved by the University of Cape Town Human Research Ethics Committee (HREC 050/2015) and was conducted under Diagnostic Microbiology and Infectious Disease protocol no. 15-0047. Those in the active TB group (PTB, n = 48) all tested sputum GXPT (Cepheid) Mtb/RIF and/or sputum Mtb liquid culture positive. All active TB cases were drug sensitive and had received no more than one dose of ATT at the time of baseline blood sampling. Participants in the LTBI group (n = 77) were all asymptomatic with a positive IFN-γ release assay (IGRA; QuantiFERON TB Gold In-Tube), tested sputum GXPT Mtb/RIF negative, and had no clinical or radiographical evidence of active TB. Similarly, participants in the healthy control (HC) group (n = 20) were all asymptomatic without a history of previous TB, tested sputum GXPT Mtb/RIF negative, had no clinical or radiographical evidence of active TB, and tested IGRA negative. Sputum GXPT Mtb/RIF and differential full blood count were performed by the South African National Health Laboratory Services.
Clinical study population in Zhengzhou, China
HIV-uninfected individuals with symptoms indicative of active TB who were administered <2 wk of ATT were enrolled in a natural history study to assess response to chemotherapy (NCT01071603) conducted at the Henan Chest Hospital in Zhengzhou, China, from 2010 to 2012. For detailed descriptions of the clinical cohort, see Mayer-Barber et al. (2014) and others (Trauner et al., 2017; Vinhaes et al., 2019).
Clinical study population in London, UK
The clinical study population of the TB patients in London, UK, who were retrospectively identified by database and case note review at the King’s College Hospital and Newham University Hospital Trust, has been described in detail in Lowe et al. (2013).
Clinical autopsy cohort from Rome, Italy
The autopsy cohort from the National Institute for Infectious Diseases, “Lazzaro Spallanzani,” in Rome, Italy, included neutral-buffered, formalin-fixed paraffin-embedded autopsy lung tissue samples from patients with histologically, histochemically, or PCR-proven confirmed PTB as previously described in Blauenfeldt et al. (2018).
Clinical study population in Bethesda, MD
For WB flow cytometry or in vitro experiments on eosinophils, HCs from the National Institutes of Health (NIH) in Bethesda, MD, were recruited through an institutional review board–approved clinical protocol to obtain normal blood samples (NCT00001846). All participants gave written informed consent.
Rhesus macaques
13 healthy >2-yr-old rhesus macaques (male and female, tuberculin skin test negative) were received from the National Institute of Allergy and Infectious Diseases (NIAID) breeding colony on Morgan Island. Animals were housed in NHP biocontainment racks and maintained in accordance with the Animal Welfare Act; Guide for the Care and Use of Laboratory Animals; and all applicable regulations, standards, and policies in a fully American Association for Accreditation of Laboratory Animal Care International–accredited Animal Biosafety Level 3 vivarium. All procedures were performed using appropriate anesthetics as listed in the NIAID Division of Intramural Research Animal Care and Use Committee–approved animal study proposal LPD25E. Euthanasia methods were consistent with the American Veterinary Medical Association Guidelines on Euthanasia and end point criteria listed in the NIAID Division of Intramural Research Animal Care and Use Committee–approved animal study proposal.
Mtb infections of rhesus macaques
Rhesus macaques from two independent infections provided tissues for this study (Fig. S2 B), and animals DF4A, 0DH, DF8B, and DFNO have previously been reported in another study (Sallin et al., 2018). Individual macaques were infected with 30–50 CFU of either H37Rv or mCherry-expressing Erdman strain of Mtb. For infections, animals were anesthetized, and 2 ml of Mtb-containing saline was instilled bronchoscopically into the right lower lung lobe. Delivered infection doses were confirmed by plating aliquots of instillation solution onto 7H11 agar plates. Colony counts were determined after 21 d.
Mice
B6 mice were purchased from Taconic Farms. B6 PHIL mice (Lee et al., 2004) were obtained from Elizabeth Jacobsen (Mayo Clinic, Phoenix, AZ). B6 ΔdblGata (JAX 33551, B6.129S1(C)-Gata1tm6Sho/LvtzJ) mice were provided by Helene Rosenberg (NIAID, Bethesda, MD). ΔdblGata (JAX 5653, C.129S1(B6)-Gata1tm6Sho/J) on the BALB/cJ background mice were purchased from The Jackson Laboratory. Both male and female mice, 8–24 wk old at the onset of experiments, were used, and experimental groups in individual experiments were age and sex matched. All animals were bred and maintained in an American Association for Accreditation of Laboratory Animal Care–accredited ABSL2 or ABSL3 facility at NIH, and experiments were performed in compliance with an animal study proposal approved by the NIAID Animal Care and Use Committee (protocol LCIM17E).
Mtb infections of mice
For aerosol infection of mice with H37Rv strains of Mtb, animals were placed in a whole-body inhalation exposure system (Glas-Col) and exposed to aerosolized Mtb. Delivery doses (100–300 CFU for standard low dose, 1,000–1,500 CFU for high dose) were confirmed by measuring lung bacterial loads at 2–24 h after exposure in control mice. Bacterial loads were measured in tissue homogenates obtained via digestion and dissociation using gentleMACS (Miltenyi Biotec) or mechanical homogenization using Precellys Evolution (Bertin Instruments). Lung homogenates were serially diluted in PBS/Tween-20 and cultured on Middlebrook 7H11 agar plates supplemented with oleic acid–albumin–dextrose catalase (Difco Laboratories). Colony counts were determined after 21 d. H37Rv-mCherry was provided by Kevin Urdahl (Seattle Children's, Seattle, WA; Cohen et al., 2018).
M. marinum infection and paraffin embedding of zebrafish
All zebrafish husbandry and experiments were approved by the Duke University Animal Care and Use Committee (protocol A122-17-05). Single-cell aliquots of cerulean-expressing M. marinum were made according to the method of Takaki et al. (2013). Zebrafish were anesthetized with 0.016% tricaine and infected with M. marinum–expressing cerulean fluorescent protein at a dose of 400 fluorescent bacteria/animal by intraperitoneal injection. Zebrafish were maintained in spawning tanks with daily water changes. At 2 wk after infection, zebrafish were euthanized by tricaine overdose and the head and tail removed with a scalpel. A small incision was made in the belly, and the zebrafish were fixed for 2 d in 4% paraformaldehyde. The animals were subsequently processed and paraffin embedded by the Duke Research Immunohistology Laboratory.
Histopathology
Tissue samples were fixed with 4% paraformaldehyde, paraffin embedded, sectioned, and stained with either H&E, the Ziehl–Nielsen histochemical method to visualize acid-fast mycobacteria, anti-EPX for the immunohistochemical identification of eosinophils, or PAS for necrosis. Eosinophils were identified via light microscopy using an Olympus BX51 microscope (magnifications of 4×, 10×, 20×, or 40×) and photomicrographs taken using an Olympus DP73 camera. The distribution of eosinophil staining was assessed microscopically using indicated scales; all histological analyses were scored in a blinded manner by a board-certified veterinary pathologist.
Confocal microscopy of rhesus macaque granulomas
Excised granulomas from Erdman-mCherry–fluorescent Mtb-infected rhesus macaques were fixed overnight in 4% paraformaldehyde and subsequently embedded in optimal cutting temperature compound (Thermo Fisher Scientific) and stored at −80°C . Cryostat-cut sections (7 μm) were placed onto microscope slides (Thermo Fisher Scientific) and blocked with 10% FCS in PBS and incubated with anti-laminin-AF647 and anti-human EPX AF488 for 1 h at 37°C in a humidified chamber. Afterward, the slides were washed in PBS for 10 min and mounted with ProLong Gold Antifade Mountant (Thermo Fisher Scientific). The samples were analyzed with a Leica SP8 confocal laser fluorescence microscope and Imaris software (Bitplane).
RNA extraction and sequencing
For RNA extraction, dedicated lung lobes were placed in RNAlater (Invitrogen) and stored at −80°C. RNAlater-stabilized lung lobes were thawed at RT for 20 min, then homogenized in RLT Plus buffer with β-mercaptoethanol (QIAGEN). Total RNA was then isolated from the RLT-homogenized cells using the RNeasy Plus Mini Kit (QIAGEN). RNA quantification, qualification, as well as library preparation and transcriptome sequencing were performed by Novogene Corporation. Briefly, RNA degradation and contamination was monitored on 1% agarose gels, and RNA purity was checked using the NanoPhotometer Spectrophotometer (IMPLEN). RNA integrity and quantitation were assessed using the RNA Nano 6000 Assay Kit of the Bioanalyzer 2100 system (Agilent Technologies). A total amount of 1 μg RNA per sample was used as input material for the RNA sample preparations. Sequencing libraries were generated using NEBNext Ultra RNA Library Prep Kit for Illumina (New England Biolabs) following the manufacturer’s recommendations, and index codes were added to attribute sequences to each sample. Briefly, mRNA was purified from total RNA using poly-T oligo–attached magnetic beads. Fragmentation was performed using divalent cations under elevated temperature in NEBNext First Strand Synthesis Reaction Buffer (5×). First strand cDNA was synthesized using random hexamer primer and M-MuLV Reverse Transcriptase (RNase H−). Second strand cDNA synthesis was subsequently performed using DNA Polymerase I and RNase H. Remaining overhangs were converted into blunt ends via exonuclease/polymerase activities. After adenylation of 3′ ends of DNA fragments, NEBNext Adaptor with hairpin loop structure were ligated to prepare for hybridization. To select cDNA fragments of preferentially 150∼200 bp in length, the library fragments were purified with AMPure XP system (Beckman Coulter). Then 3-μl USER Enzyme (New England Biolabs) was used with size-selected, adaptor–ligated cDNA at 37°C for 15 min followed by 5 min at 95°C before PCR. PCR then was performed with Phusion High-Fidelity DNA Polymerase and Universal and Index Primers. Finally, PCR products were purified (AMPure XP system), and library quality was assessed using the Agilent Bioanalyzer 2100 system. The clustering of the index-coded samples was performed on a cBot Cluster Generation System using PE Cluster Kit cBot-HS (Illumina) according to the manufacturer’s instructions. After cluster generation, the library preparations were sequenced on an Illumina NovaSeq 6000 platform, and paired-end 150-bp reads were generated. The entire gene expression dataset is available in Gene Expression Omnibus under accession no. GSE165871.
Transcriptional analysis
For all samples, low-quality bases were removed and adapters trimmed using Trimmomatic v0.32 (Bolger et al., 2014). After the quality check, sequences were aligned to the Mus musculus genome (GRCm38.p6 version 67) with STAR 2.7.0 (Dobin et al., 2013). After mapping, the output was converted to count tables with the tximport package (Soneson et al., 2015) from R 4.0.2. The stratification was performed as follows: d0 WT B6 group was composed of five samples: d0WT1–d0WT5, d0 ΔdblGata group (d0Gata1–4), d90 WT B6 group (d90WT1–4), and d90 ΔdblGata group (d90Gata1–5). Count gene expression matrix was examined by the DESeq2 package (Love et al., 2014) from R 4.0.2 to identify DEGs using the following the comparisons: d90 ΔdblGata versus d90 WT B6, d0 ΔdblGata versus d0 WT B6, d90 ΔdblGata versus d0 WT B6, and d90 WT B6 versus d0 WT B6. Changes in gene expression levels were considered significant when statistical test values (false discovery rate [FDR]–adjusted P value) were <0.05 and the fold change/difference higher than ±1.4. Candidate DEGs were visualized by volcano plots and Venn diagrams with the EnhancedVolcano and VennDiagram packages from R 4.0.2. The obtained DEGs were scanned by Reactome (Yu and He, 2016) and Kyoto Encyclopedia of Genes and Genomes (http://www.genome.jp/kegg/) pathway database using the compareCluster package (Yu et al., 2012) from R 4.0.2. The gene list of IFN-inducible (type I) and IFN-γ–inducible (type II) lung modules were retrieved from Singhania et al. (2019). The sample clustering and classification were assessed using heatmaps and applied in the variance-stabilizing transformation gene expression values.
Cell isolations for flow cytometry: Peripheral blood
Rhesus macaque and human blood samples were collected in EDTA tubes, and WB was used for flow cytometry. Briefly, antibody cocktails were directly added to 200 μl of WB and stained at 37°C for 20 min, after which samples were washed twice in 4 ml 1% FCS/PBS. Fixable live/dead cell stain (Molecular Probes; Invitrogen) was used according to the manufacturer’s protocol. Pellets were then fixed and permeabilized with the eBioscience Foxp3 Transcription Factor Staining Buffer Kit (Life Technologies) for at least 1 h followed by intracellular antigen staining for 30 min at 4°C. Cells were washed and samples acquired on a FACSymphony (BD Biosciences) at NIH or an LSRFortessa (BD Biosciences) at SPHCC. FACS data were analyzed using FlowJo 10 (FlowJo, LLC).
Cell isolations for flow cytometry: BAL and granulomas from rhesus macaques
Rhesus macaque BAL samples were passed through a 100-μm cell strainer, pelleted, and counted for analysis. Granulomas were individually resected from the lungs, and samples used for flow cytometry analysis were pushed through a 100-μm cell strainer. Aliquots from all samples were serially diluted and plated on 7H11 agar plates for CFU quantification. Alternatively, resected granulomas were fixed in 4% paraformaldehyde for later paraffin embedding. Samples were stained with surface antibody cocktails for 20 min at 4°C followed by fixable live/dead cell stain (Molecular Probes; Invitrogen). Cells were then fixed and permeabilized (eBioscience Transcription Factor Staining Buffer Kit) for at least 1 h followed by intracellular antigen staining for 30 min at 4°C. Cells were washed and samples acquired on a FACSymphony (BD Biosciences) at NIH. FACS data were analyzed using FlowJo 10 (FlowJo, LLC).
Cell isolations for flow cytometry: Human and murine lung tissue
Mice were i.v. injected 3 min before euthanasia with 5 μg/mouse APC- or BV711-labeled CD45 (30-F11) as previously reported (Anderson et al., 2014). Lungs from infected mice were digested using a lung cell isolation buffer and dissociated via gentleMACS (Miltenyi Biotec) according to the manufacturer’s instructions. Digested lung was fully dispersed by passage through a 100-μm cell strainer, and an aliquot was removed for bacterial load measurements. Cells were then washed and purified using a 37% Percoll step. Samples were stained with surface antibody cocktails for 20 min at 4°C followed by fixable live/dead cell stain (Molecular Probes; Invitrogen). Cells were then fixed and permeabilized (eBioscience Transcription Factor Staining Buffer Kit) for at least 1 h followed by intracellular antigen staining for 30 min at 4°C. Cells were washed and samples acquired on a FACSymphony (BD Biosciences) at NIH. FACS data were analyzed using FlowJo 10 (FlowJo, LLC).
The surgically resected human lung tissue samples were weighed and manually aliquoted into replicates using sterile scissors inside a sterile 50-ml tube. Aliquots were transferred into Trizol for RNA isolation or subjected to single-cell isolation. Briefly, sample replicates for single-cell isolation were digested with 100 U/ml of collagenase IV (Sigma) and 50 U/ml of Benzonase (Sigma) in 8 ml of incomplete RPMI-1640 medium (without FBS) for 45 min at 37°C in a shaker. Cells were then filtered through 100-μm cell strainers (Miltenyi Biotec), and the remainder of tissue pieces were gently squashed by a syringe and washed with 4 ml ice-cold PBS containing 50% FBS to stop digestion. Next, the cell suspension was centrifuged and subjected to Debris Removal Solution (Miltenyi Biotec) according to the manufacturer’s protocol. Samples were stained with surface antibody cocktails for 20 min at 4°C followed by fixable live/dead cell stain (Molecular Probes; Invitrogen). Cells were then fixed and permeabilized (eBioscience Transcription Factor Staining Buffer Kit) for at least 1 h followed by intracellular antigen staining for 30 min at 4°C. Cells were washed and samples acquired on an LSRFortessa (BD Biosciences) at SPHCC. FACS data were analyzed using FlowJo 10 (FlowJo, LLC).
Flow cytometry reagents
Fluorochrome-labeled antibodies against mouse, human, or rhesus macaque antigens used for flow cytometric analysis are as follows: I-A/I-E (clone M5/114.15.2), Ly6G (1A8), CD11c (HL3 and N418), CD45.2 (104), TCRb (H57-597), NK1.1 (PK136), CD11b (M1/70), CD45 (30-F11), CD68 (FA-11), Ly6C (AL-21 and HK1.4), Siglec-F (E50-2440), CD193 (J073E5), Arg1 (A1exF5), NOS2 (CXNFT), CD4 (L3T4), CD8a (53-6.7), Foxp3 (FJK-16s), CD193 (5E8.4), EPX (AHE-1), CD62L (MEL-14), CD14 (M5E2), CD11c (3.9), CD45 (D058-1283), CD66abce (TET2), CD15 (HI98), CD68 (Y1/82A), HLA-DR (L243), CD16 (3G8), CD45 (2D1), CD193 (83101), CD66b (G10F5), CD62L (REG-56), CD206 (15–2.2), CD123 (6H6), and Siglec-8 (7C9). ESAT-61–17 or TB10.44–11 tetramers were produced by the NIAID tetramer core facility (Emory University, Atlanta, GA).
Statistical analyses
The statistical significance of differences between data groups was calculated using GraphPad Prism 8. The Mann–Whitney test and Wilcoxon matched pairs test (PET low versus high, otherwise paired samples) were used for comparison of group means, log-rank Mantel–Cox test was used for survival data, and Spearman correlation test was used unless otherwise indicated in the figures or figure legends.
Online supplemental material
Fig. S1 shows additional clinical data and study designs. Fig. S2 shows human and NHP granuloma structures with eosinophil distribution and EPX staining specificity. Fig. S3 shows depletion efficiency in eosinophil-deficient mouse strains after Mtb infection and additional transcriptional and immunological parameters assessed in ΔdblGata and increased bacterial loads in B6 PHIL mice. Table S1 lists RNA sequencing analysis results of DEGs, modules, and GSEA.
Acknowledgments
We thank the clinical study participants and medical staff at the Khayelitsha Site B Community Health Centre in Cape Town, the SPHCC affiliated with Fudan University in Shanghai, and the Henan Chest Hospital in Zhengzhou. We also thank Drs. Y. Belkaid, A. O’Garra, S.P. Babu, M. Lionakis, and A. Sher for data discussions and feedback on the manuscript. We are grateful to Ray Y. Chen for assistance with clinical protocol, all Tuberculosis Imaging Program members, and the staff of the NIAID ABSL2 and ABSL3 facilities.
This work was supported in part by the Division of Intramural Research, National Institute of Allergy and Infectious Diseases (K.D. Mayer-Barber, D.L. Barber, C.E. Barry III, L.E. Via) and the National Natural Science Foundation of China (grant no. 81770010 to K.-W. Wong). R.J. Wilkinson is supported by the National Institutes of Health (grant no. U01AI115940), Wellcome Trust (grant nos. 104803 and 203135), and Cancer Research UK, UK Research and Innovation, and Wellcome Trust (FC0010218). B.B. Andrade is supported by the National Council for Scientific and Technological Development (senior fellowship) and by the Intramural Research Program of the Oswaldo Cruz Foundation, Brazil. This research was funded in whole or in part by Wellcome Trust. For the purpose of Open Access, the author has applied a CC-BY public copyright license to any Author Accepted Manuscript (AAM) version arising from this submission.
Author contributions: Conceptualization: Y. Song, A.D. Klion, K.-W. Wong, and K.D. Mayer-Barber. Methodology: A.C. Bohrer, E. Castro, Z. Hu, L. Wang, Tuberculosis Imaging Program, Y. Song, K.-W. Wong, and K.D. Mayer-Barber. Investigation: A.C. Bohrer, E. Castro, Z. Hu, C.E. Tocheny, M. Assmann, C. Nelson, P.J. Baker, H. Ma, L. Wang, W. Zilu, S. Sakai, E. du Bruyn, C. Riou, K.D. Kauffman, Tuberculosis Imaging Program, I.N. Moore, D.M. Lowe, K.-W. Wong, and K.D. Mayer-Barber. Resources: A.R. Martineau, D.M. Lowe, L. Wang, Tuberculosis Imaging Program, F. Del Nonno, L. Petrone, D. Goletti, M.R. Cronan, R.J. Wilkinson, C.E. Barry III, L.E. Via, D.L. Barber, Y. Song, K.-W. Wong, and K.D. Mayer-Barber. Data analysis and curation: A.C. Bohrer, E. Castro, Z. Hu, E. du Bruyn, C. Riou, C.E. Tocheny, M. Assmann, Tuberculosis Imaging Program, H. Ma, A.T.L. Queiroz, B.B. Andrade, and K.D. Mayer-Barber. Writing-original draft: K.D. Mayer-Barber. Writing-review and editing: all authors. Visualization: A.C. Bohrer, A.T.L. Queiroz, and K.D. Mayer-Barber. Supervision: B.B. Andrade, D.L. Barber, L.E. Via, R.J. Wilkinson, Y. Song, K.-W. Wong, and K.D. Mayer-Barber. Funding acquisition: B.B. Andrade, D.L. Barber, L.E. Via, K.-W. Wong, and K.D. Mayer-Barber.
References
Competing Interests
Disclosures: D.M. Lowe reported personal fees from Merck, non-financial support from CSL Behring, and non-financial support from Fujifilm Chemical Co. outside the submitted work. No other disclosures were reported.
Author notes
A.C. Bohrer, E. Castro, and Z. Hu contributed equally to this paper.