The success of Mycobacterium tuberculosis (Mtb) as a human pathogen results from its ability to avoid, resist, or subvert immunological mechanisms that are effective against most pathogens. Mtb has a remarkable ability to withstand the microbicidal arsenal of host macrophages, including the low pH of the lysosome and a diverse array of toxic reactive oxygen species (ROS) produced by the host respiratory burst. However, relatively little is known about the molecular mechanisms of resistance to these toxic insults.

Upon phagocytosis of a pathogen, macrophages and neutrophils unleash a potent microbicidal arsenal, including a respiratory burst that leads to the rapid release of ROS that kill and degrade most microbes (El-Benna et al., 2009; Nathan, 2006; Segal, 2005). The production of ROS is initiated when NADPH oxidase assembles on pathogen-containing phagosomes and initiates a series of reactions that produce superoxide, hydrogen peroxide, hyperchlorous acid, lipid peroxides, and other toxic reactive molecules. Importantly, Mtb is highly resistant to oxidative stress, enabling the bacterium to survive within macrophages and neutrophils and establish chronic infection (Chan et al., 1992). Although some oxidative defense mechanisms have been characterized (Ehrt and Schnappinger, 2009; Nambi et al., 2015; Voskuil et al., 2011), the Mtb genome encodes a large number of enzymes that could participate in defense against oxidative stress that remain uncharacterized. For example, Mtb encodes ~15 peroxidases, a class of enzymes with diverse functions that form the front line of oxidative defense in bacteria (Mishra and Imlay, 2012; Singh and Eltis, 2015). It is unclear why Mtb encodes such a diversity of peroxidases, and little is known about what individual roles these enzymes may play in both resistance to oxidative stress and in normal bacterial physiology. One possibility is that specific peroxidase genes may participate in defense against different types of ROS generated during the respiratory burst. One of the most intriguing peroxidases encoded by Mtb is DyP, a member of the dye-decolorizing peroxidase family of proteins. DyP enzymes are unique amongst peroxidases in that some family members have been shown to have optimal activity at pH 4–5 (Brown et al., 2012; Singh and Eltis, 2015; Uchida et al., 2015). Furthermore, DyP is the only Mtb peroxidase predicted to be packaged inside a proteinaceous organelle known as a bacterial nanocompartment (Contreras et al., 2014).

Bacterial cells were long thought to lack compartmentalization of function. However, the identification of organelle-like structures including microcompartments, anammoxosomes, magnetosomes, and, most recently, encapsulin nanocompartments has revolutionized our understanding of bacterial cell biology (Ferousi et al., 2017; Kerfeld et al., 2018; Uebe and Schüler, 2016). Characterized nanocompartments are proteinaceous shells that are 24–45 nm in diameter and comprised of 60–240 subunits of a single protomer (Giessen et al., 2019; Nichols et al., 2017). This shell surrounds (‘encapsulates’) an enzymatic cargo protein (Nichols et al., 2017). Although putative encapsulin systems have been identified in >900 bacterial and archaeal genomes (Giessen and Silver, 2017; Nichols et al., 2020), very little is known about their physiological function. Based on genomic organization, encapsulin systems are often predicted to compartmentalize enzymes involved in oxidative stress defense, iron storage, and anaerobic ammonium oxidation (Giessen and Silver, 2017). However, the function of an encapsulin system has only ever been demonstrated in a single bacterial species, Myxococcus xanthus (McHugh et al., 2014), and the physiological role of nanocompartments is largely unexplored.

Here, we demonstrate that an Mtb nanocompartment containing DyP is crucial for resisting oxidative stress at low pH. Mutants lacking the nanocompartment are highly attenuated when exposed to H₂O₂ at pH 4.5, the pH of the lysosome. We show that this attenuation is linked to fatty acid metabolism in the bacteria. Further, we show that encapsulation of the DyP enzyme promotes its function, the first report to demonstrate a role for encapsulation in endogenous bacterial physiology. Finally, we show that the Mtb nanocompartment system is required for full growth in macrophages, and for resistance to the antibiotic pyrazinamide (PZA), demonstrating the importance of this system for a globally significant pathogen.

The Mtb genome encodes the predicted encapsulin gene Rv0798c/Cfp29 (Contreras et al., 2014) in a two-gene operon with Rv0799c, the dye-decolorizing peroxidase DyP (Figure 1A). Overexpression of the predicted Mtb encapsulin gene in Escherichia coli was previously shown to result in the formation of nanocompartment-like structures (Contreras et al., 2014). Three potential cargo proteins for the nanocompartment were proposed based on a putative shared encapsulation targeting sequence: DyP, FolB, and BrfB. In E. coli, overexpression of each protein with Cfp29 resulted in encapsulation. However, this study did not address whether Mtb produces endogenous nanocompartments or identify the specific function of these compartments in Mtb biology. A transposon screen has identified Cfp29 as a gene required for growth in mice (Zhang et al., 2013) and Cfp29 has long been known as an immunodominant T cell antigen in both mice and human TB patients (Weldingh and Andersen, 1999). Taken together, these results suggest that Mtb may produce an encapsulin nanocompartment that is important for pathogenesis.

Figure 1 with 1 supplement see all

Download asset Open asset

To confirm the previous finding that heterologous expression of Rv0798c and Rv0799c in a host species results in the assembly of an encapsulin system, we expressed these genes in E. coli and isolated nanocompartments. Clarified protein lysates from E. coli were purified by ultracentrifugation and size-exclusion chromatography. Assembled encapsulin nanocompartments are distinguishable by their high molecular weight (Nichols et al., 2017). Indeed, a fraction from the purification contained a high molecular weight species > 260 kDa observable on an SDS-PAGE gel (Figure 1—figure supplement 1A). Fractions containing putative nanocompartments were pooled and imaged using transmission electron microscopy, which revealed the presence of icosahedral structures with the expected diameter of ~25 nm (Figure 1B and C).

To determine whether Mtb produces nanocompartments under normal laboratory growth conditions, we performed an ultracentrifugation-based nanocompartment isolation strategy using wild-type H37Rv strain bacteria grown to mid-log phase (Figure 1—figure supplement 1B). Mass spectrometry analysis of the nanocompartment fraction identified both the encapsulin protein Cfp29 and the peroxidase DyP (Figure 1D). TEM analysis confirmed the presence of nanocompartment particles ~25 nm in diameter (Figure 1E). Interestingly, Rv1762c, a protein of unknown function, was consistently identified in purified nanocompartment preparations from Mtb (Figure 1D). We were unable to identify either FolB or BrfB in nanocompartments from Mtb, suggesting that these proteins are not endogenous substrates for encapsulation under normal laboratory growth conditions for Mtb.

Cfp29 (‘culture filtrate protein 29’) was originally identified in the supernatants of Mtb cells grown in axenic culture (Rosenkrands et al., 1998). As Cfp29 lacks a secretory signal sequence and is part of a large macromolecular complex, it is unclear how a nanocompartment could be actively secreted. However, nanocompartment structures are remarkably stable and it is possible that nanocompartments released from dying bacteria accumulate in culture as they are highly resistant to proteolysis/degradation (Cassidy-Amstutz et al., 2016; Nichols et al., 2017).

DyP proteins are known to have low pH optima (Contreras et al., 2014; Uchida et al., 2015). To demonstrate that Mtb DyP has a similarly low pH optimum, SUMO-tagged unencapsulated DyP was purified from E. coli (Figure 1—figure supplement 1C), and the peroxidase activity was evaluated at a range of pH values using the ABTS (2,2'-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid)) dye-decolorizing assay previously used to characterize DyP proteins (Ahmad et al., 2011; Contreras et al., 2014). Similar to the Vibrio cholerae DyP (Uchida et al., 2015), purified Mtb DyP had increased enzymatic activity in low pH environments, with the greatest efficacy at pH 4.0, the lowest pH tested (Figure 1F). We next tested the ability of encapsulated DyP to degrade ABTS across a range of pH values. Similar to free DyP, the encapsulated enzyme had the highest activity at pH ~4.0 (Figure 1G). Taken together, these data demonstrate the stability and functionality of Mtb nanocompartments under acid stress, a condition that mimics the host lysosomal environment. The Mtb DyP protein was previously shown to function as a bona fide dye-decolorizing peroxidase (Contreras et al., 2014). Because peroxidases consume H₂O₂, they often participate in defense against oxidative stress (Mishra and Imlay, 2012). During macrophage infection, the phagocytes initiate an oxidative burst that exposes Mtb to H₂O₂ (El-Benna et al., 2009). We therefore reasoned that DyP-containing nanocompartments may function to protect Mtb from H₂O₂-induced stress. To test this hypothesis, we created a mutant strain lacking both genes from the nanocompartment operon (Figure 1A, Δoperon). Lysates from Δoperon mutants were used for nanocompartment purification (data not shown) and western blot analysis using an antibody for Cfp29 as a probe (Figure 1—figure supplement 1D). Data from these analyses revealed that Δoperon mutants cannot produce viable nanocompartments. In addition, we isolated a transposon mutant (DyP::Tn) containing an insertion in DyP, a mutation that eliminated expression of both Cfp29 and DyP (Figure 1—figure supplement 1E).

To test whether DyP-containing nanocompartments are required for defense against H₂O₂, we exposed wild-type and DyP::Tn Mtb to increasing concentrations of H₂O₂ and monitored bacterial survival. Mutants lacking DyP nanocompartments were more susceptible to H₂O₂ when compared with wild-type bacteria as measured by OD₆₀₀ (Figure 2A) and by plating for colony-forming units (CFU) (Figure 2B). However, the phenotype was relatively modest. We next reasoned that DyP nanocompartments might be required to resist oxidative stress in acidic environments that more closely mimic the in vivo environment. During infection, Mtb bacilli encounter the low pH of the phagolysosome and the ability to tolerate low pH is required for Mtb survival in both infected macrophages and mice (Vandal et al., 2008). We therefore tested the susceptibility of DyP mutants to a combination of H₂O₂ and acid stress (pH 4.5). Wild-type Mtb was able to withstand these conditions and did not significantly decrease in number over a 3-day exposure period (Figure 2C). In contrast, the Δoperon mutant was resistant to each stressor individually, but was highly susceptible to H₂O₂ at pH 4.5 (Figure 2C). Thus, DyP-containing nanocompartments are required to protect bacteria from oxidative stress at low pH.

Figure 2

Download asset Open asset

We next sought to determine whether encapsulation of DyP is important for its activity in bacterial cells. To accomplish this, DyP::Tn mutants were complemented with constructs expressing DyP alone (pDyP), Cfp29 encapsulin alone (pCfp29), or both proteins (pOperon). As expected, expression of DyP did not restore production of nanocompartments, whereas expression of Cfp29 alone resulted in the formation of empty nanocompartment structures (Figure 1—figure supplement 1E). Coexpression of both proteins resulted in the formation of DyP-containing nanocompartments in the DyP::Tn mutant (Figure 1—figure supplement 1E). We exposed the full set of complemented mutants to H₂O₂ at pH 4.5 and determined survival after 3 days by plating for CFU. As expected, DyP::Tn was attenuated for survival under these conditions when compared to wild-type bacteria (Figure 2D). Restoring expression of the nanocompartment shell protein in the absence of DyP had no effect on bacterial survival (Figure 2D). Complementation by overexpression of unencapsulated DyP was sufficient to confer almost wild-type levels of resistance to oxidative and acid stress (Figure 2D). However, complementation with both Cfp29 and DyP resulted in enhanced resistance to these stressors when compared to mutants complemented with DyP alone (Figure 2D), suggesting that encapsulation of DyP enhances its function.

In the mutant complemented with DyP alone, it is possible that overexpression of the peroxidase compensates for the lack of encapsulation. We therefore examined the importance of DyP encapsulation in mutant strains in which we deleted Cfp29 only (Δcfp29). Deletion of Cfp29 resulted in an ~0.5 log decrease in bacterial viability after 3 days of exposure to H₂O₂ at pH 4.5, confirming that encapsulation of the peroxidase in the shell protein is important for full protection (Figure 2E). As expected, greater attenuation was observed in the mutant lacking both DyP and Cfp29 (Δoperon). Complementation of the Δcfp29 mutant fully restored resistance (Figure 2F). Taken together, these data demonstrate that encapsulation of DyP has a modest, yet significant impact on protection against oxidative stress in low pH conditions.

We next sought to test whether nanocompartment mutants have altered redox homeostasis in the presence of oxidative stress. We transformed wild-type and mutant strains with Mrx1-roGFP, a fluorescent reporter of redox potential in mycobacteria (Bhaskar et al., 2014). Remarkably, the Δoperon mutant had an intracellular environment at baseline that was more oxidizing than wild-type bacteria exposed to H₂O₂ (Figure 2G), indicating a failure of redox homeostasis in mutants lacking the DyP encapsulin system. This oxidizing cellular environment was further enhanced by treatment with H₂O₂ (Figure 2G). We did not observe a difference in the redox state of Δcfp29. Thus, mutants lacking DyP-containing nanocompartments exhibit significant dysregulation of redox homeostasis.

The Mtb genome encodes ~15 putative peroxidase genes, including DyP. To determine whether other bacterial peroxidases participate in the defense against oxidative stress at low pH, we performed a transposon sequencing (Tn-seq) screen (Griffin et al., 2012; van Opijnen and Camilli, 2013). To do so, we created a transposon library containing ~100K individual transposon mutants and exposed this library to H₂O₂ at pH 4.5 for 3 days, at which point the surviving bacteria were plated. Transposon gene junctions were amplified and sequenced from the recovered bacteria, and the sequencing data were analyzed using TRANSIT (DeJesus et al., 2015; Supplementary file 1). These data revealed that of the 18 putative enzymes encoded by Mtb that could be involved in oxidative defense, including peroxidases, catalases, and superoxide dismutases, only two genes were required for survival in the culture conditions—DyP and KatG (Figure 3A). KatG encodes a catalase-peroxidase that is important for defense against oxidative stress in the host (Li et al., 1998; Pym et al., 2002). Coincidentally, KatG is also required for activation of isoniazid (INH), a central component of anti-TB therapy. Approximately 10% of TB cases are caused by INH-resistant bacteria, many of which have loss-of-function mutations in KatG (Seifert et al., 2015). Importantly, KatG variants that do not activate INH often display a concomitant decrease in peroxidase and catalase activity (DeVito and Morris, 2003; Mo et al., 2004; Wei et al., 2003). Both the high proportion of KatG mutant bacteria in the human population and studies of virulence in animals suggest that KatG mutants are still capable of growth in vivo (Nieto et al., 2016; Pym et al., 2002). The fact that both DyP and KatG mutants are important for survival in oxidative stress at low pH suggests that the DyP nanocompartment may provide redundancy that compensates for a loss of KatG in INH-resistant strains of Mtb.

Figure 3

Download asset Open asset

In the Tn-seq data, we observed that mutants involved in lipid or cholesterol metabolism were also attenuated when exposed to H₂O₂ at pH 4.5 (Figure 3B). Mtb biology is highly linked to lipid biology; Mtb granulomas are rich in free fatty acids (Kim et al., 2010) and the bacteria utilize lipids as a source of nutrition both during macrophage infection ex vivo and growth in vivo (Marrero et al., 2010; Russell, 2011; Russell et al., 2019). In particular, mutants containing transposon insertions in three independent subunits of the Mce1 complex, which is required for import of fatty acids (Nazarova et al., 2019), were susceptible to oxidative stress under acidic conditions (Figure 3B). The standard Mtb culture medium, 7H9, contains bovine serum albumin (BSA), which has binding sites for fatty acids and may also serve as a cholesterol shuttle in serum (Sankaranarayanan et al., 2013; van der Vusse, 2009). We hypothesized that fatty acids carried by albumin may become toxic following exposure to oxidative stress and low pH and that import of fatty acids into the bacteria might be a protective mechanism. . We therefore tested whether fatty acids are required for the susceptibility of Δoperon mutants using exposure to acid and oxidative stress in Sauton’s broth, a minimal medium that lacks BSA, and found that in the absence of fatty acids the Δoperon mutants persisted to the same degree as wild-type bacteria (Figure 3C). To test whether lipids bound to BSA mediate the susceptibility of DyP mutants to acid and oxidative stress, we cultured wild-type and Δoperon Mtb in media constituted with fatty acid-free BSA and exposed the bacteria to H₂O₂ at pH 4.5 for 3 days. In the absence of BSA-bound lipids, we found that the Δoperon mutant was not susceptible to oxidative stress at low pH (Figure 3D). 7H9 medium prepared with lipid-free BSA was then reconstituted with oleic acid, an abundant lipid found in mammalian systems (Abdelmagid et al., 2015). Reconstitution of fatty acid-free medium with oleic acid restored toxicity to nanocompartment mutants under conditions of acid and oxidative stress (Figure 3D). Taken together, these data suggest that the phagosomal lipids used by Mtb as a carbon source may be toxic to bacteria lacking functional nanocompartments under conditions of oxidative stress.

We considered the possibility that lipids present in the medium might disrupt Mtb membranes, resulting in altered pH homeostasis and a drop in cytosolic pH. Interestingly, in acidified medium we found that the intracellular pH of wild-type bacteria dropped from ~pH 7.5 to ~pH 6.2, and this decrease was dependent on the presence of albumin-bound fatty acids (Figure 3E). These data confirm published results demonstrating that the cytosolic pH of Mtb drops significantly during acid exposure (Vandal et al., 2008). A lower cytosolic pH would possibly impair the function of many other bacterial enzymes important for resisting oxidative stress and place increasing importance on enzymes that can function in acidic environments, such as DyP (Uchida et al., 2015). Taken together, these data may explain why functional nanocompartments are critical for bacterial survival when Mtb is exposed to acid and oxidative stress in a fatty acid-rich environment.

Our in vitro findings that DyP nanocompartments are important for defense against a combination of oxidative stress, low pH, and fatty acids suggested that this system may be important for Mtb survival in the phagosomal environment. Therefore, we sought to determine whether DyP nanocompartments are required for bacterial growth in host cells. We found that the Δoperon strain was consistently attenuated for growth in murine bone marrow-derived macrophages (Figure 4A). We attempted to complement this phenotype; however, the limited dynamic range between wild-type and mutant in macrophages made clean interpretation of the complementation results difficult. As these genes are predicted to be in a two-gene operon with no other predicted genes in close proximity, polar effects are extremely unlikely. However, to confirm that the phenotype of the mutants in macrophages is due to the loss of the two-gene encapsulin operon, we repeated the experiment with the independently derived Dyp:Tn mutant and also observed attenuated growth in macrophages across repeated experiments (Figure 4B).

Figure 4

Download asset Open asset

One of the antibiotics used to treat TB infection, PZA, is most effective in low pH environments, such as the phagolysosome. PZA is thought to disrupt the cell wall of Mtb, which leads to acidification of the bacterial cytosol. A recent study in human TB patients demonstrated that the pH of necrotic lung cavities is ~5.5, a finding that may also explain the efficacy of PZA in treating Mtb infection (Kempker et al., 2017). Since DyP is critical for protection of Mtb against oxidative stress at low pH, we considered whether nanocompartments may mediate Mtb resistance to PZA. To test bacterial susceptibility to PZA, we treated wild-type and Δoperon Mtb with PZA in the presence of H₂O₂ at pH 5.5. We found that mutants lacking the DyP nanocompartment were sensitive to concentrations of PZA that did not impact growth of the wild-type bacteria (Figure 4C). We next set out to test whether mutants lacking the DyP operon are susceptible to PZA treatment in vivo. We infected BALB/C mice with wild-type and mutant bacteria by the aerosol route with a dose of ~250 bacteria instilled on day 0 of infection. PZA treatment began on day 14 and continued until day 35, at which time we enumerated CFU in the lungs. We did not find that Δoperon mutants were more attenuated than wild-type bacteria in mouse lungs at 35 days post infection, at least under the conditions examined. However, Δoperon mutants were significantly more susceptible to treatment with PZA in vivo (Figure 4D), demonstrating that the DyP encapsulin operon is required for endogenous resistance to PZA treatment.

Here, we report that a protein nanocompartment in Mtb encapsulates the dye-decolorizing peroxidase DyP. By demonstrating that the encapsulin system is important for resisting oxidative stress at low pH, we provide the first evidence that encapsulin systems contribute to oxidative stress in a bacterial pathogen, and only the second demonstration of a functional encapsulin system in any species. Furthermore, we expand our understanding of the importance of encapsulation in enzyme function by demonstrating that the protein shell encapsulating DyP enhances its function in vivo. Finally, we demonstrate that the DyP nanocompartment system is essential for the ability of Mtb to grow in host macrophages and to resist killing by the antibiotic PZA. These findings are the first demonstration that a nanocompartment system is important for virulence in any bacterial pathogen.

Bacterial protein-based compartments are emerging as functional equivalents of eukaryotic organelles that can sequester enzymatic reactions from the environment of the cytosol. Bacteria appear to have two predominant classes of protein based organelles, the relatively well-understood bacterial microcompartment (Kerfeld et al., 2018) and the more recently discovered bacterial nanocompartment (Nichols et al., 2017). Rather than using a lipid membrane for compartmentalization, both bacterial microcompartments and nanocompartments consist of a porous protein shell that surrounds enzymatic proteins. Bacterial microcompartments that enable utilization of specialized nutrients including ethanolamine and 1,2 propane diol have previously been implicated in pathogenesis of E. coli and Salmonella enterica, respectively (Jakobson and Tullman-Ercek, 2016). Nanocompartments, which are 10 times smaller than a typical bacterial microcompartment, have been identified bioinformatically in over 900 bacterial species; however, very few studies have addressed the role that nanocompartments play in bacterial physiology. Bioinformatic analysis and the heterologous expression of nanocompartment systems from diverse bacteria in E. coli have suggested that these systems might function in iron mineralization, oxidative and nitrosative stress resistance, and anaerobic ammonium oxidation (Giessen and Silver, 2017). However, the only nanocompartment that has been shown to be important for survival under conditions of stress is the ferritin-like protein containing compartment in M. xanthus, which is required for resistance to H₂O₂ (McHugh et al., 2014). Our demonstration that an encapsulin system is required for resistance to oxidative stress in Mtb significantly broadens the potential role and impact of nanocompartment systems in bacterial physiology.

A previous report examined the Mtb nanocompartment by heterologous expression of the encapsulin gene in E. coli. Coexpression with three potential substrates identified bioinformatically based on a putative signal sequence suggested that three proteins can be encapsulated by the Mtb nanocompartment: DyP, FolB, and BrfB (Contreras et al., 2014). Here, we show for the first time that Mtb produces this nanocompartment. However, of the three potential cargo proteins, only DyP was identified in encapsulin systems purified from logarithmically growing bacteria. Why is the DyP protein encapsulated? There is almost no understanding of the role of encapsulation in bacterial physiology despite the widespread presence of encapsulins across bacterial species. In this study, we found that DypB alone can protect the bacteria from oxidative stress at low pH. It is possible that in some circumstances, particularly when the stressor is not maximal, encapsulation of DyP may not be necessary. We expect that further investigation of this nanocompartment system in Mtb and in other bacterial species will reveal additional roles of encapsulins in the bacterial physiology.

We showed that Mtb mutants lacking the DyP encapsulin system have an altered redox balance at baseline, with a significantly more oxidative cytosol than wild-type. However, under standard growth conditions, the mutants are not defective. While mutants exhibit a small defect for survival when exposed to H₂O₂ at pH 6.5, they are highly attenuated at low pH in the presence of H₂O₂. Interestingly, the attenuation of the mutants under these conditions requires the presence of fatty acids bound to BSA in the culture medium. Why is the DyP encapsulin system required for growth under these conditions? It is possible that this system is required for detoxifying an oxidative species other than H₂O₂ itself, such as a lipid peroxide. It is also possible that the low pH compromises the integrity of the cell wall such that oxidative damage is sustained, or that the low pH magnifies the importance of DyP, which has a low pH optimum. Nevertheless, although the exact function of the DyP encapsulin system remains to be elucidated, this system is clearly important for pathogenesis as mutants fail to grow in host macrophages. Intriguingly, encapsulin systems have been shown to confer stability to the enzymatic cargo under harsh conditions, including heat and protease exposure (Nichols et al., 2017). Cfp29 was originally identified as a component of short-term culture filtrate present in broth of growing cells, leading to the supposition that this protein is secreted (Rosenkrands et al., 1998). Given that nanocompartments spontaneously assemble inside cells (Cassidy-Amstutz et al., 2016; Nichols et al., 2017; Snijder et al., 2016), it is unlikely that a structure the size of a ribosome could be secreted. We speculate that nanocompartments are released from dying cells, and that their extreme stability promotes their accumulation in axenic bacterial culture. Indeed, this stability may partially explain why Cfp29 is a known immunodominant antigen for T cells (Rosenkrands et al., 1998). Taken together, these data suggest the intriguing possibility that nanocompartments released from dying cells in vivo could continue to provide defense against oxidative stress for the surviving bacteria.

We found that only two peroxidases, KatG and DyP, are required for survival in the presence of H₂O₂ at pH 4.5. These findings raise the possibility that these peroxidases have redundant function in vivo, possibly explaining the lack of attenuation we observed in Δoperon mutants in BALB/C mice at 35 days post infection. It is also possible that Δoperon mutants are not attenuated at this timepoint in BALB/C mice as the Tn-seq screen that previously identified Cfp29 as being required for growth in mouse lungs used the C57BL/6 strain (Zhang et al., 2013). KatG mutants are highly prevalent in the clinic as loss of function in KatG prevents activation of the INH prodrug. Importantly, many KatG mutants that fail to activate INH are also defective for catalase activity (DeVito and Morris, 2003; Mo et al., 2004; Wei et al., 2003). This raises the possibility that the DyP nanocompartment system could partially compensate for a loss of KatG catalase activity, enabling these drug resistant mutants to survive in vivo. In addition to a potential role for DyP in the context of INH treatment and INH resistant infections, we found that DyP encapsulin mutants are susceptible to treatment with PZA, an antibiotic known to require acidic conditions for efficacy. Thus, we predict that this encapsulin system may limit the efficacy of PZA treatment of patients.

In summary, we have demonstrated that Mtb produces a functional nanocompartment containing the peroxidase DyP. This system is essential for resisting oxidative stress at low pH, conditions that resemble the host lysosome. In addition, the DyP encapsulin system is required for growth in host macrophages and for resistance to the antibiotic PZA. Previously encapsulin systems have been primarily studied in the context of heterologous expression in E. coli and for bioengineering or structural studies. Our findings demonstrate the significance of encapsulin nanocompartments in bacterial physiology in the globally significant pathogen Mtb.

All data generated or analysed during this study are included in the manuscript and supporting file.

1. 1. Abdelmagid SA
   2. Clarke SE
   3. Nielsen DE
   4. Badawi A
   5. El-Sohemy A
   6. Mutch DM
   7. Ma DWL

   (2015) Comprehensive profiling of plasma fatty acid concentrations in young healthy Canadian adults

   PLOS ONE 10:e0116195.

   https://doi.org/10.1371/journal.pone.0116195

   • Google Scholar
2. 1. Ahmad M
   2. Roberts JN
   3. Hardiman EM
   4. Singh R
   5. Eltis LD
   6. Bugg TDH

   (2011) Identification of DypB from Rhodococcus jostii RHA1 as a lignin peroxidase

   Biochemistry 50:5096–5107.

   https://doi.org/10.1021/bi101892z

   • PubMed
   • Google Scholar
3. 1. Bhaskar A
   2. Chawla M
   3. Mehta M
   4. Parikh P
   5. Chandra P
   6. Bhave D
   7. Kumar D
   8. Carroll KS
   9. Singh A

   (2014) Reengineering redox sensitive GFP to measure mycothiol redox potential of Mycobacterium tuberculosis during infection

   PLOS Pathogens 10:e1003902.

   https://doi.org/10.1371/journal.ppat.1003902

   • PubMed
   • Google Scholar
4. 1. Brown ME
   2. Barros T
   3. Chang MCY

   (2012) Identification and characterization of a multifunctional dye peroxidase from a lignin-reactive bacterium

   ACS Chemical Biology 7:2074–2081.

   https://doi.org/10.1021/cb300383y

   • PubMed
   • Google Scholar
5. 1. Cassidy-Amstutz C
   2. Oltrogge L
   3. Going CC
   4. Lee A
   5. Teng P
   6. Quintanilla D
   7. East-Seletsky A
   8. Williams ER
   9. Savage DF

   (2016) Identification of a Minimal Peptide Tag for in Vivo and in Vitro Loading of Encapsulin

   Biochemistry 55:3461–3468.

   https://doi.org/10.1021/acs.biochem.6b00294

   • PubMed
   • Google Scholar
6. 1. Chan J
   2. Xing Y
   3. Magliozzo RS
   4. Bloom BR

   (1992) Killing of virulent Mycobacterium tuberculosis by reactive nitrogen intermediates produced by activated murine macrophages

   The Journal of Experimental Medicine 175:1111–1122.

   https://doi.org/10.1084/jem.175.4.1111

   • PubMed
   • Google Scholar
7. 1. Contreras H
   2. Joens MS
   3. McMath LM
   4. Le VP
   5. Tullius MV
   6. Kimmey JM
   7. Bionghi N
   8. Horwitz MA
   9. Fitzpatrick JAJ
   10. Goulding CW

   (2014) Characterization of a Mycobacterium tuberculosis nanocompartment and its potential cargo proteins

   The Journal of Biological Chemistry 289:18279–18289.

   https://doi.org/10.1074/jbc.M114.570119

   • PubMed
   • Google Scholar
8. 1. DeJesus MA
   2. Ambadipudi C
   3. Baker R
   4. Sassetti C
   5. Ioerger TR

   (2015) TRANSIT--A Software Tool for Himar1 TnSeq Analysis

   PLOS Computational Biology 11:e1004401.

   https://doi.org/10.1371/journal.pcbi.1004401

   • PubMed
   • Google Scholar
9. 1. DeVito JA
   2. Morris S

   (2003) Exploring the structure and function of the mycobacterial KatG protein using trans-dominant mutants

   Antimicrobial Agents and Chemotherapy 47:188–195.

   https://doi.org/10.1128/AAC.47.1.188-195.2003

   • PubMed
   • Google Scholar
10. 1. Ehrt S
    2. Guo XV
    3. Hickey CM
    4. Ryou M
    5. Monteleone M
    6. Riley LW
    7. Schnappinger D

    (2005) Controlling gene expression in mycobacteria with anhydrotetracycline and Tet repressor

    Nucleic Acids Research 33:e21.

    https://doi.org/10.1093/nar/gni013

    • PubMed
    • Google Scholar
11. 1. Ehrt S
    2. Schnappinger D

    (2009) Mycobacterial survival strategies in the phagosome: defence against host stresses

    Cellular Microbiology 11:1170–1178.

    https://doi.org/10.1111/j.1462-5822.2009.01335.x

    • PubMed
    • Google Scholar
12. 1. El-Benna J
    2. Dang PMC
    3. Gougerot-Pocidalo MA
    4. Marie JC
    5. Braut-Boucher F

    (2009) p47phox, the phagocyte NADPH oxidase/NOX2 organizer: structure, phosphorylation and implication in diseases

    Experimental & Molecular Medicine 41:217–225.

    https://doi.org/10.3858/emm.2009.41.4.058

    • PubMed
    • Google Scholar
13. 1. Ferousi C
    2. Lindhoud S
    3. Baymann F
    4. Kartal B
    5. Jetten MS
    6. Reimann J

    (2017) Iron assimilation and utilization in anaerobic ammonium oxidizing bacteria

    Current Opinion in Chemical Biology 37:129–136.

    https://doi.org/10.1016/j.cbpa.2017.03.009

    • PubMed
    • Google Scholar
14. 1. Giessen TW
    2. Silver PA

    (2017) Widespread distribution of encapsulin nanocompartments reveals functional diversity

    Nature Microbiology 2:17029.

    https://doi.org/10.1038/nmicrobiol.2017.29

    • PubMed
    • Google Scholar
15. 1. Giessen TW
    2. Orlando BJ
    3. Verdegaal AA
    4. Chambers MG
    5. Gardener J
    6. Bell DC
    7. Birrane G
    8. Liao M
    9. Silver PA

    (2019) Large protein organelles form a new iron sequestration system with high storage capacity

    eLife 8:e46070.

    https://doi.org/10.7554/eLife.46070

    • PubMed
    • Google Scholar
16. 1. Griffin JE
    2. Pandey AK
    3. Gilmore SA
    4. Mizrahi V
    5. McKinney JD
    6. Bertozzi CR
    7. Sassetti CM

    (2012) Cholesterol catabolism by Mycobacterium tuberculosis requires transcriptional and metabolic adaptations

    Chemistry & Biology 19:218–227.

    https://doi.org/10.1016/j.chembiol.2011.12.016

    • PubMed
    • Google Scholar
17. 1. Jakobson CM
    2. Tullman-Ercek D

    (2016) Dumpster Diving in the Gut: Bacterial Microcompartments as Part of a Host-Associated Lifestyle

    PLOS Pathogens 12:e1005558.

    https://doi.org/10.1371/journal.ppat.1005558

    • PubMed
    • Google Scholar
18. 1. Kempker RR
    2. Heinrichs MT
    3. Nikolaishvili K
    4. Sabulua I
    5. Bablishvili N
    6. Gogishvili S
    7. Avaliani Z
    8. Tukvadze N
    9. Little B
    10. Bernheim A
    11. Read TD
    12. Guarner J
    13. Derendorf H
    14. Peloquin CA
    15. Blumberg HM
    16. Vashakidze S

    (2017) Lung Tissue Concentrations of Pyrazinamide among Patients with Drug-Resistant Pulmonary Tuberculosis

    Antimicrobial Agents and Chemotherapy 61:e00226-17.

    https://doi.org/10.1128/AAC.00226-17

    • PubMed
    • Google Scholar
19. 1. Kerfeld CA
    2. Aussignargues C
    3. Zarzycki J
    4. Cai F
    5. Sutter M

    (2018) Bacterial microcompartments

    Nature Reviews. Microbiology 16:277–290.

    https://doi.org/10.1038/nrmicro.2018.10

    • PubMed
    • Google Scholar
20. 1. Kim MJ
    2. Wainwright HC
    3. Locketz M
    4. Bekker LG
    5. Walther GB
    6. Dittrich C
    7. Visser A
    8. Wang W
    9. Hsu FF
    10. Wiehart U
    11. Tsenova L
    12. Kaplan G
    13. Russell DG

    (2010) Caseation of human tuberculosis granulomas correlates with elevated host lipid metabolism

    EMBO Molecular Medicine 2:258–274.

    https://doi.org/10.1002/emmm.201000079

    • PubMed
    • Google Scholar
21. 1. Larsen MH
    2. Biermann K
    3. Jacobs, WR

    (2007) Laboratory Maintenance of Mycobacterium tuberculosis

    Current Protocols in Microbiology 6:s6.

    https://doi.org/10.1002/9780471729259.mc10a01s6

    • Google Scholar
22. 1. Li Z
    2. Kelley C
    3. Collins F
    4. Rouse D
    5. Morris S

    (1998) Expression of katG in Mycobacterium tuberculosis is associated with its growth and persistence in mice and guinea pigs

    The Journal of Infectious Diseases 177:1030–1035.

    https://doi.org/10.1086/515254

    • PubMed
    • Google Scholar
23. 1. Long JE
    2. DeJesus M
    3. Ward D
    4. Baker RE
    5. Ioerger T
    6. Sassetti CM

    (2015) Identifying essential genes in Mycobacterium tuberculosis by global phenotypic profiling

    Methods in Molecular Biology 1279:79–95.

    https://doi.org/10.1007/978-1-4939-2398-4_6

    • PubMed
    • Google Scholar
24. 1. Marrero J
    2. Rhee KY
    3. Schnappinger D
    4. Pethe K
    5. Ehrt S

    (2010) Gluconeogenic carbon flow of tricarboxylic acid cycle intermediates is critical for Mycobacterium tuberculosis to establish and maintain infection

    PNAS 107:9819–9824.

    https://doi.org/10.1073/pnas.1000715107

    • PubMed
    • Google Scholar
25. 1. McHugh CA
    2. Fontana J
    3. Nemecek D
    4. Cheng N
    5. Aksyuk AA
    6. Heymann JB
    7. Winkler DC
    8. Lam AS
    9. Wall JS
    10. Steven AC
    11. Hoiczyk E

    (2014) A virus capsid-like nanocompartment that stores iron and protects bacteria from oxidative stress

    The EMBO Journal 33:1896–1911.

    https://doi.org/10.15252/embj.201488566

    • PubMed
    • Google Scholar
26. 1. Mishra S
    2. Imlay J

    (2012) Why do bacteria use so many enzymes to scavenge hydrogen peroxide?

    Archives of Biochemistry and Biophysics 525:145–160.

    https://doi.org/10.1016/j.abb.2012.04.014

    • PubMed
    • Google Scholar
27. 1. Mo L
    2. Zhang W
    3. Wang J
    4. Weng XH
    5. Chen S
    6. Shao LY
    7. Pang MY
    8. Chen ZW

    (2004) Three-dimensional model and molecular mechanism of Mycobacterium tuberculosis catalase-peroxidase (KatG) and isoniazid-resistant KatG mutants

    Microbial Drug Resistance 10:269–279.

    https://doi.org/10.1089/mdr.2004.10.269

    • PubMed
    • Google Scholar
28. 1. Nambi S
    2. Long JE
    3. Mishra BB
    4. Baker R
    5. Murphy KC
    6. Olive AJ
    7. Nguyen HP
    8. Shaffer SA
    9. Sassetti CM

    (2015) The Oxidative Stress Network of Mycobacterium tuberculosis Reveals Coordination between Radical Detoxification Systems

    Cell Host & Microbe 17:829–837.

    https://doi.org/10.1016/j.chom.2015.05.008

    • PubMed
    • Google Scholar
29. 1. Nathan C

    (2006) Neutrophils and immunity: challenges and opportunities

    Nature Reviews. Immunology 6:173–182.

    https://doi.org/10.1038/nri1785

    • PubMed
    • Google Scholar
30. 1. Nazarova EV
    2. Montague CR
    3. Huang L
    4. La T
    5. Russell D
    6. VanderVen BC

    (2019) The genetic requirements of fatty acid import by Mycobacterium tuberculosis within macrophages

    eLife 8:e43621.

    https://doi.org/10.7554/eLife.43621

    • PubMed
    • Google Scholar
31. 1. Nichols RJ
    2. Cassidy-Amstutz C
    3. Chaijarasphong T
    4. Savage DF

    (2017) Encapsulins: molecular biology of the shell

    Critical Reviews in Biochemistry and Molecular Biology 52:583–594.

    https://doi.org/10.1080/10409238.2017.1337709

    • PubMed
    • Google Scholar
32. Preprint

    1. Nichols RJ
    2. LaFrance B
    3. Phillips N
    4. Oltrogge LM
    5. Valentin-Alvarado LE
    6. Bischoff A
    7. Nogales E
    8. Savage DF

    (2020) Discovery and Characterization of a Novel Family of Prokaryotic Nanocompartments Involved in Sulfur Metabolism

    bioRxiv.

    https://doi.org/10.1101/2020.05.24.113720

    • Google Scholar
33. 1. Nieto LM
    2. Mehaffy C
    3. Creissen E
    4. Troudt J
    5. Troy A
    6. Bielefeldt-Ohmann H
    7. Burgos M
    8. Izzo A
    9. Dobos KM

    (2016) Virulence of Mycobacterium tuberculosis after Acquisition of Isoniazid Resistance: Individual Nature of katG Mutants and the Possible Role of AhpC

    PLOS ONE 11:e0166807.

    https://doi.org/10.1371/journal.pone.0166807

    • PubMed
    • Google Scholar
34. 1. Pym AS
    2. Saint-Joanis B
    3. Cole ST

    (2002) Effect of katG mutations on the virulence of Mycobacterium tuberculosis and the implication for transmission in humans

    Infection and Immunity 70:4955–4960.

    https://doi.org/10.1128/IAI.70.9.4955-4960.2002

    • PubMed
    • Google Scholar
35. 1. Rosenberg OS
    2. Dovala D
    3. Li X
    4. Connolly L
    5. Bendebury A
    6. Finer-Moore J
    7. Holton J
    8. Cheng Y
    9. Stroud RM
    10. Cox JS

    (2015) Substrates Control Multimerization and Activation of the Multi-Domain ATPase Motor of Type VII Secretion

    Cell 161:501–512.

    https://doi.org/10.1016/j.cell.2015.03.040

    • PubMed
    • Google Scholar
36. 1. Rosenkrands I
    2. Rasmussen PB
    3. Carnio M
    4. Jacobsen S
    5. Theisen M
    6. Andersen P

    (1998) Identification and characterization of a 29-kilodalton protein from Mycobacterium tuberculosis culture filtrate recognized by mouse memory effector cells

    Infection and Immunity 66:2728–2735.

    https://doi.org/10.1128/IAI.66.6.2728-2735.1998

    • PubMed
    • Google Scholar
37. 1. Russell DG

    (2011) Mycobacterium tuberculosis and the intimate discourse of a chronic infection

    Immunological Reviews 240:252–268.

    https://doi.org/10.1111/j.1600-065X.2010.00984.x

    • PubMed
    • Google Scholar
38. 1. Russell DG
    2. Huang L
    3. VanderVen BC

    (2019) Immunometabolism at the interface between macrophages and pathogens

    Nature Reviews. Immunology 19:291–304.

    https://doi.org/10.1038/s41577-019-0124-9

    • PubMed
    • Google Scholar
39. 1. Sankaranarayanan S
    2. de la Llera-Moya M
    3. Drazul-Schrader D
    4. Phillips MC
    5. Kellner-Weibel G
    6. Rothblat GH

    (2013) Serum albumin acts as a shuttle to enhance cholesterol efflux from cells

    Journal of Lipid Research 54:671–676.

    https://doi.org/10.1194/jlr.M031336

    • PubMed
    • Google Scholar
40. 1. Segal AW

    (2005) How neutrophils kill microbes

    Annual Review of Immunology 23:197–223.

    https://doi.org/10.1146/annurev.immunol.23.021704.115653

    • PubMed
    • Google Scholar
41. 1. Seifert M
    2. Catanzaro D
    3. Catanzaro A
    4. Rodwell TC

    (2015) Genetic mutations associated with isoniazid resistance in Mycobacterium tuberculosis: a systematic review

    PLOS ONE 10:e0119628.

    https://doi.org/10.1371/journal.pone.0119628

    • PubMed
    • Google Scholar
42. 1. Singh R
    2. Eltis LD

    (2015) The multihued palette of dye-decolorizing peroxidases

    Archives of Biochemistry and Biophysics 574:56–65.

    https://doi.org/10.1016/j.abb.2015.01.014

    • PubMed
    • Google Scholar
43. 1. Snijder J
    2. Kononova O
    3. Barbu IM
    4. Uetrecht C
    5. Rurup WF
    6. Burnley RJ
    7. Koay MST
    8. Cornelissen JJLM
    9. Roos WH
    10. Barsegov V
    11. Wuite GJL
    12. Heck AJR

    (2016) Assembly and Mechanical Properties of the Cargo-Free and Cargo-Loaded Bacterial Nanocompartment Encapsulin

    Biomacromolecules 17:2522–2529.

    https://doi.org/10.1021/acs.biomac.6b00469

    • PubMed
    • Google Scholar
44. 1. Uchida T
    2. Sasaki M
    3. Tanaka Y
    4. Ishimori K

    (2015) A Dye-Decolorizing Peroxidase from Vibrio cholerae

    Biochemistry 54:6610–6621.

    https://doi.org/10.1021/acs.biochem.5b00952

    • PubMed
    • Google Scholar
45. 1. Uebe R
    2. Schüler D

    (2016) Magnetosome biogenesis in magnetotactic bacteria

    Nature Reviews. Microbiology 14:621–637.

    https://doi.org/10.1038/nrmicro.2016.99

    • PubMed
    • Google Scholar
46. 1. van der Vusse GJ

    (2009) Albumin as fatty acid transporter

    Drug Metabolism and Pharmacokinetics 24:300–307.

    https://doi.org/10.2133/dmpk.24.300

    • PubMed
    • Google Scholar
47. 1. van Opijnen T
    2. Camilli A

    (2013) Transposon insertion sequencing: a new tool for systems-level analysis of microorganisms

    Nature Reviews. Microbiology 11:435–442.

    https://doi.org/10.1038/nrmicro3033

    • PubMed
    • Google Scholar
48. 1. Vandal OH
    2. Pierini LM
    3. Schnappinger D
    4. Nathan CF
    5. Ehrt S

    (2008) A membrane protein preserves intrabacterial pH in intraphagosomal Mycobacterium tuberculosis

    Nature Medicine 14:849–854.

    https://doi.org/10.1038/nm.1795

    • PubMed
    • Google Scholar
49. 1. Voskuil MI
    2. Bartek IL
    3. Visconti K
    4. Schoolnik GK

    (2011) The response of mycobacterium tuberculosis to reactive oxygen and nitrogen species

    Frontiers in Microbiology 2:105.

    https://doi.org/10.3389/fmicb.2011.00105

    • PubMed
    • Google Scholar
50. 1. Wei CJ
    2. Lei B
    3. Musser JM
    4. Tu SC

    (2003) Isoniazid activation defects in recombinant Mycobacterium tuberculosis catalase-peroxidase (KatG) mutants evident in InhA inhibitor production

    Antimicrobial Agents and Chemotherapy 47:670–675.

    https://doi.org/10.1128/AAC.47.2.670-675.2003

    • PubMed
    • Google Scholar
51. 1. Weldingh K
    2. Andersen P

    (1999) Immunological evaluation of novel Mycobacterium tuberculosis culture filtrate proteins

    FEMS Immunology and Medical Microbiology 23:159–164.

    https://doi.org/10.1111/j.1574-695X.1999.tb01235.x

    • PubMed
    • Google Scholar
52. 1. Zhang YJ
    2. Reddy MC
    3. Ioerger TR
    4. Rothchild AC
    5. Dartois V
    6. Schuster BM
    7. Trauner A
    8. Wallis D
    9. Galaviz S
    10. Huttenhower C
    11. Sacchettini JC
    12. Behar SM
    13. Rubin EJ

    (2013) Tryptophan biosynthesis protects mycobacteria from CD4 T-cell-mediated killing

    Cell 155:1296–1308.

    https://doi.org/10.1016/j.cell.2013.10.045

    • PubMed
    • Google Scholar

Author details

1. Katie A Lien

   Department of Molecular and Cell Biology, Division of Immunology and Pathogenesis, University of California, Berkeley, Berkeley, United States

   Contribution

   Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Writing – original draft, Writing – review and editing

   Contributed equally with

   Kayla Dinshaw

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-2721-0887

2. Kayla Dinshaw

   Department of Molecular and Cell Biology, Division of Immunology and Pathogenesis, University of California, Berkeley, Berkeley, United States

   Contribution

   Data curation, Formal analysis, Investigation, Writing – original draft, Writing – review and editing

   Contributed equally with

   Katie A Lien

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-5354-746X

3. Robert J Nichols

   Department of Molecular and Cell Biology, Division of Biochemistry, Biophysics and Structural Biology, University of California, Berkeley, Berkeley, United States

   Contribution

   Formal analysis, Investigation, Methodology, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-8476-0554

4. Caleb Cassidy-Amstutz

   Department of Molecular and Cell Biology, Division of Biochemistry, Biophysics and Structural Biology, University of California, Berkeley, Berkeley, United States

   Contribution

   Formal analysis, Investigation, Methodology

   Competing interests

   No competing interests declared

5. Matthew Knight

   Department of Plant and Microbial Biology, University of California, Berkeley, Berkeley, United States

   Contribution

   Investigation

   Competing interests

   No competing interests declared

6. Rahul Singh

   Department of Microbiology and Immunology, The University of British Columbia, Vancouver, Canada

   Contribution

   Methodology, Resources, Writing – review and editing

   Competing interests

   No competing interests declared

7. Lindsay D Eltis

   Department of Microbiology and Immunology, The University of British Columbia, Vancouver, Canada

   Contribution

   Conceptualization, Methodology, Resources, Writing – review and editing

   Competing interests

   No competing interests declared

8. David F Savage

   Department of Molecular and Cell Biology, Division of Biochemistry, Biophysics and Structural Biology, University of California, Berkeley, Berkeley, United States

   Contribution

   Conceptualization, Investigation, Methodology, Supervision, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-0042-2257

9. Sarah A Stanley

   1. Department of Molecular and Cell Biology, Division of Immunology and Pathogenesis, University of California, Berkeley, Berkeley, United States
   2. School of Public Health, Division of Infectious Diseases and Vaccinology, University of California, Berkeley, Berkeley, United States

   Contribution

   Conceptualization, Data curation, Formal analysis, Funding acquisition, Methodology, Project administration, Supervision, Writing – original draft, Writing – review and editing

   For correspondence

   sastanley@berkeley.edu

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-4182-9048

• 2,309

  views

• 409

  downloads

• 21

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.