Human variation impacting MCOLN2 restricts Salmonella Typhi replication by magnesium deprivation

Summary Human genetic diversity can reveal critical factors in host-pathogen interactions. This is especially useful for human-restricted pathogens like Salmonella enterica serovar Typhi (S. Typhi), the cause of typhoid fever. One key defense during bacterial infection is nutritional immunity: host cells attempt to restrict bacterial replication by denying bacteria access to key nutrients or supplying toxic metabolites. Here, a cellular genome-wide association study of intracellular replication by S. Typhi in nearly a thousand cell lines from around the world—and extensive follow-up using intracellular S. Typhi transcriptomics and manipulation of magnesium availability—demonstrates that the divalent cation channel mucolipin-2 (MCOLN2 or TRPML2) restricts S. Typhi intracellular replication through magnesium deprivation. Mg2+ currents, conducted through MCOLN2 and out of endolysosomes, were measured directly using patch-clamping of the endolysosomal membrane. Our results reveal Mg2+ limitation as a key component of nutritional immunity against S. Typhi and as a source of variable host resistance.

Correspondence dennis.ko@duke.edu In brief Gibbs et al. conducted a cellular GWAS of nearly a thousand lymphoblastoid cells to identify genetic variation affecting the MCOLN2 (TRPML2) divalent cation channel as the primary genetic determinant of variation in Salmonella Typhi intracellular replication. MCOLN2 deprives S. Typhi of Mg 2+ , serving as an important mechanism of nutritional immunity.

INTRODUCTION
Genome-wide association studies (GWASs) are a powerful method to identify common genetic variants associated with risk, resistance, or other quantitative measures of infectious disease. 1 However, connecting variants identified by whole-organism GWAS to disease pathogenesis is often challengingespecially when it is unclear how the identified variants affect nearby genes or how these genes relate to the disease under investigation. To solve this, studies look among the disease-linked variants for those that associate with expression of nearby genes (called expression quantitative trait loci [eQTLs]), which can provide important clues, especially during stimulation with pathogens 2 or pathogen-associated molecular patterns. 3,4 A complimentary approach to connect variant and disease is GWAS of cellular traits, such as our Hi-HOST (high-throughput human in vitro susceptibility testing) 5,6 platform, which associates genetic variation with quantifiable cellular traits, such as invasion, 7 inflammation, 8 and intracellular pathogen replication. As a further benefit, cellular GWASs provide control of environmental and pathogen variation, which boosts statistical power by reducing noise. Used together, cellular GWASs of eQTLs can connect genetic variants to both altered gene expression and cellular process, explaining how the identified variation impacts clinical outcomes and facilitating subsequent mechanistic studies.
Here, we used this approach to study susceptibility to the human-restricted enteric pathogen Salmonella enterica ser. Typhi (S. Typhi), which relies on a permissive niche inside immune cells to cause the life-threatening syndrome of typhoid fever. 9 We discovered that the interferon-inducible 10 host cation channel, mucolipin-2 (MCOLN2 or TRPML2), is critical for nutritional immunity against S. Typhi. In the dynamic competition between the host and bacteria, nutritional immunity is the ongoing effort of the host cell to restrict Salmonella replication by depriving it of key nutrients [11][12][13] or delivering toxic metabolites. 14 Nutritional immunity is well-characterized in the Salmonella-host competition for iron, [15][16][17][18] and has recently expanded to encompass competition for other key trace metal ions, such as zinc [19][20][21] and manganese. 22,23 Here, we demonstrate that MCOLN2 deprives S. Typhi of magnesium (Mg 2+ ), playing a major role in Mg 2+ -based nutritional immunity for Salmonella replicating inside human cells.

RESULTS
We identified common human single-nucleotide polymorphisms (SNPs) associated with S. Typhi intracellular replication, using Hi-HOST screening and family-based GWAS analysis 24 of 961 lymphoblastoid cell lines (LCLs) (EBV-immortalized B cells) from eight populations ( Figure 1A; Table S1). LCLs are a powerful in vitro model because they are karyotypically normal, and B cells are a natural site of Salmonella replication in vivo. 25 Because intracellular replication, or host cell permissivity, is a demonstrated proxy for Salmonella virulence in whole organisms, 26 we used variable LCL permissivity to screen for human susceptibility or resistance factors. In this screen, we defined permissivity as the ratio of bacterial burden at 24 h to 3.5 h based on median green fluorescence intensity of live intracellular S. Typhi. This analysis revealed a single genome-wide significant locus (lead SNP is rs10873679, p = 6 3 10 À9 ) on chromosome 1 ( Figure 1B). A quantile-quantile plot demonstrated no overall inflation of the test statistic, with primarily rs10873679-linked SNPs deviating from the neutral distribution ( Figure 1C). The association signal covers two genes in the mucolipin subfamily, MCOLN2 and MCOLN3 ( Figure 1D). Mucolipins are a family of three inward rectifying divalent cation channels that localize to endolysosomal membranes and regulate vesicular trafficking. 27 The minor (globally less common) C-allele of rs10873679 is associated with more intracellular S. Typhi replication ( Figures  1E and S1). To link this to cellular physiology, we examined expression of MCOLN2 and MCOLN3 in RNA sequencing (RNA-seq) of 1000 Genomes LCLs. 28 The C-allele associated with less MCOLN2 expression (Figure 2A; p < 2 3 10 À16 ), while rs10873679 was not associated with a significant difference in MCOLN3 expression ( Figure 2B). In confirmation, the C-allele also associated with reduced MCOLN2 protein abundance in a quantitative mass spectrometry analysis of HapMap LCLs 29 ( Figure 2C; p = 0.01). In this same analysis, MCOLN3 protein was only detected in 9 LCLs. This was insufficiently powered to draw a conclusion, although there was no evidence for association of MCOLN3 protein with rs10873679 genotype with these limited numbers. Together, the rs10873679 C-allele's association with both more S. Typhi replication and less MCOLN2 mRNA and protein suggested that MCOLN2 restricts S. Typhi intracellular replication.
Strengthening this model, MCOLN2 is upregulated in human macrophages after treatment with M1 polarizing LPS and IFNg, 30 which indicates that MCOLN2 is part of the host response. Similarly, we observed MCOLN2 induction after S. Typhi infection ( Figure 2D). If MCOLN2 is a restriction factor, we expected that ablating MCOLN2 expression would increase intracellular Salmonella replication. Knocking down MCOLN2, but not MCOLN3, increased S. Typhi intracellular replication ( Figure 2E), without affecting bacterial invasion or pyroptosis ( Figure S2). This phenotype generalized to other human immune cells, as knocking down MCOLN2 in THP-1 monocytes by RNAi (Figure 2F) or knocking out the gene using CRISPR-Cas ( Figure 2G) resulted in an even greater increase in S. Typhi replication than in LCLs. In fact, MCOLN2 knockout in THP-1s increased S. Typhi replication from 1-to 1.5-fold to 3-to 4-fold at 24 h, a large 2.5-fold increase in bacterial replication.
The rs10873679 locus was also associated with intracellular replication of S. Typhimurium (p = 8.1 3 10 À7 ; Figure S3), a serovar used to model enteric fever in mice as S. Typhi is humanrestricted; however, the impact of reducing MCOLN2 expression is much smaller with S. Typhimurium ( Figure 2H). This demonstrates that, while MCOLN2 is a key restriction factor for S. Typhi (knockout results in $150% more replication), it is an accessory factor for controlling S. Typhimurium (knockout results in $20% more replication). We confirmed lack of a large effect with S. Typhimurium by infecting susceptible C57BL/6J mice with Mcoln2 knocked out 31 via intraperitoneal injection-which avoids restriction by stomach acid or variance introduced by gut microbiota-and quantified S. Typhimurium burden in the spleen 4 days post infection ( Figure 2I). This revealed no significant difference in S. Typhimurium burden between Mcoln2 genotypes, despite a modest trend of higher burden in Mcoln2 À/À mice, which is not surprising given the small in vitro phenotype. This serovar difference could be explained by bacterial differenceonly S. Typhi has the capacity to take advantage of a changed niche after MCOLN2's removal-or a differential host response, in which the more immunogenic S. Typhimurium induces additional restriction factors that prevent it from fully exploiting MCOLN2 knockout. Regardless, our data demonstrate that MCOLN2 is a strong restriction factor for the human-specific serovar S. Typhi in cells, which underscores the value of cellular GWAS for identifying human-specific host-pathogens interactions.
To determine how MCOLN2 reduces S. Typhi replication, we used the intracellular bacteria as reporters of their own environment. We conducted transcriptomics at 16 h post infection (hpi), near maximum divergence of replication inside wild-type vs. MCOLN2 À/À THP-1s and prior to restriction in wild-type THP-1s ( Figures 3A and 3B). While >2,600 bacterial genes were detected, and expression of one-quarter of the bacterial transcriptome significantly changed between late-log inoculum and 16 hpi, differences between bacteria within wild-type and MCOLN2 À/À cells were more modest with expression of no individual bacterial gene passing significance threshold after correction for multiple testing (Table S2). Therefore, we used gene set enrichment analysis to identify S. Typhi processes that were upregulated in MCOLN2-containing wild-type THP-1s. We generated a list of 15 gene sets of physiological processes associated with virulence or divalent cation transport ( Figure 3C; Table S3). Only genes regulated by the PhoP/Q two-component system 32 were significantly enriched (NES = À1.81 with FDR q = 0.004) in bacteria living inside wild-type THP-1s compared with MCOLN2 À/À THP-1s ( Figure 3D). While S. Typhi within MCOLN2 À/À cells upregulate PhoP/Q targets (10.7-fold more expression than late-log), induction is greater in bacteria inside wild-type cells (13.3-fold).
To determine if PhoP/Q signaling contributes to replication in MCOLN2 knockout cells, we infected THP-1s with the Ty800 DphoPQ strain, 33 which revealed that most ($75%) of the increased replication inside MCOLN2 À/À requires intact PhoPQ signaling ( Figure 3E). Chief among PhoP/Q targets is the SPI-2 T3SS, which injects effectors to maintain Salmonella's intracellular niche. Removing an essential component of the SPI-2 T3SS basal body (ssaT) to prevent any secretion caused no change in S. Typhi replication within wild-type THP-1s (compare blue bars in Figure 3F). This contrasts with S. Typhimurium 34,35 but is consistent with past S. Typhi literature. 36 In contrast, replication is reduced in MCOLN2 À/À cells, suggesting that roughly half of the PhoP/Q-dependent increase in S. Typhi replication depends on SPI-2 effectors ( Figure 3F). This indicates the SPI-2 independence of S. Typhi replication in THP-1 monocytes is actually an MCOLN2-dependent host response that suppresses the fitness advantage provided by S. Typhi's SPI-2 effectors.
Our results demonstrate that S. Typhi replicating inside MCOLN2 À/À monocytes upregulate PhoP targets, which significantly boosts replication. However, in wild-type cells, the even greater induction of PhoP targets is not sufficient to increase replication, so we speculated that the PhoP upregulation was a symptom of a restrictive condition enhanced by MCOLN2. Three potentially restricting conditions in the SCV lead to more PhoP activity: PhoP/Q is repressed by high Mg 2+37 and activated by cationic antimicrobial peptides 38 or acidification. 39,40 Since MCOLN2 is a divalent cation channel, PhoP/Q was most likely responding to reduced Mg 2+ concentrations, which, along with Zn 2+ , are limited in SCVs. 41,42 Indeed, the PhoP-activated Mg 2+ importers mgtA and mgtB were both upregulated more in bacteria inside wild-type THP-1s ( Figures 3D and S4). Therefore, the transcriptomics and phoPQ mutant infection data suggested a simple hypothesis: MCOLN2 deprives S. Typhi of Mg 2+ .
To test this, we repleted Mg 2+ 2 h after infecting and measured bacterial replication ( Figure 4A). Mg 2+ supplementation disproportionately benefited bacterial replication inside wild-type THP-1s (1.6-fold in wild-type vs. 1.2-fold in knockout THP-1s; interaction p = 0.002). Similar results were also observed with S. Typhimurium ( Figure 4B). While our transcriptomics could also support a role for Zn 2+ , zinc repletion did not have interactions with the MCOLN2 genotype, meaning it was similarly toxic to S. Typhi inside both MCOLN2 À/À and wild-type THP-1 cells ( Figure 4C; interaction p = 0.3). This agrees with previous findings that high concentrations of zinc are toxic to Salmonella. 43 However, S. Typhi inside MCOLN2 knockout cells are not more susceptible to Zn 2+ repletion, so we infer that MCOLN2 does not help S. Typhi resist zinc toxicity. Together, these data demonstrate that intracellular replication is held back by magnesium starvation and not zinc toxicity.
This Mg 2+ starvation model is supported by whole-endolysosome patch-clamp measurements. While previous studies using whole-cell patch-clamping have demonstrated that MCOLN1 is permeable to most monovalent and divalent cations, 44 there has been no direct evidence showing that MCOLN2 conducts Mg 2+ from the lumen of endolysosomes into the cytosol. To determine if human MCOLN2 can conduct Mg 2+ , it was expressed in HEK293 cells, and endolysosomal organelles were isolated for direct patch-clamping using a previously established approach. [45][46][47] While no significant Mg 2+ currents were seen in non-transfected endolysosomes ( Figure 4D), application of an MCOLN2-specific small-molecule agonist, ML2-SA1, 46 evoked inward Mg 2+ currents on intact endolysosomes isolated from MCOLN2-expressing cells ( Figure 4E). We also observed Mg 2+ currents after administration of PI(3,5)P 2 , a putative endogenous agonist 48 ( Figure 4F). This is especially intriguing in light of Salmonella's known manipulation of phosphoinositides through the sopB effector 49,50 and our previous finding that a host protein that regulates PI(3,5)P 2 is associated with Salmonella invasion and typhoid fever risk. 7 These results demonstrate that MCOLN2 conducts Mg 2+ and is capable of serving as a channel for Mg 2+ out of endolysosomes and into the cytosol.  The repletion and electrophysiological evidence is further bolstered by genetic interaction of MCOLN2 with Salmonella Mg 2+ transporters. The importance of Mg 2+ acquisition for Salmonella replication is underscored by its trio of Mg 2+ uptake proteins: one constitutive, CorA, and two inducible, MgtA and MgtB. If knocking out MCOLN2 increases Mg 2+ availability, we theorized that these transporters would be necessary to uptake that extra Mg 2+ and therefore essential for the enhanced replication inside MCOLN2 À/À host cells. To test this, we generated a double knockout (DmgtADmgtB), which lacks the high-affinity Mg 2+ importers used in low-Mg 2+ environments, like the % 10 mM concentration in the SCV. 41 In confirmation of our hypothesis, the double importer mutant is killed, instead of replicating, inside THP-1s ( Figure 4G). Knocking out MCOLN2 provides less of an advantage to the double importer mutant (increasing replication 60% in DD vs. 150% in wild-type S. Typhi; interaction p < 0.001). This corroborates the Mg 2+ repletion and suggests that most of the enhanced replication in MCOLN2 À/À THP-1s depends on increasing Mg 2+ availability.
To test if this magnesium-MCOLN2 interaction occurs in vivo, we infected susceptible mice with a 1:1 ratio of double knockout (DmgtADmgtB or DD) and wild-type S. Typhimurium by intraperitoneal injection. In theory, the increased Mg 2+ availability in Mcoln2 À/À mice would change the competitive index (CI) between double mutant and wild-type S. Typhimurium. As expected, DmgtADmgtB S. Typhimurium is greatly attenuated compared with the wild type in C57BL/6J (CI = 0.024), and the double Mg 2+ -importer mutant's attenuation is significantly more pronounced in Mcoln2 À/À mice (CI = 0.007; Figure 4H). Based on our cellular findings, one would expect the reduced CI in Mcoln2 À/À mice to be driven by more replication of the wild-type bacteria that can take advantage of increased Mg 2+ availability in Mcoln2 À/À mice; instead, we observed no significant change in wild-type bacteria replication in Mcoln2 À/À mice (DD inÀ/À /DD in+/+ = 0.48 with p = 0.4) accompanied with significantly less replication of double mutant bacteria in Mcoln2 À/À mice (DD inÀ/À /DD in+/+ = 0.14 with p = 0.002; Figure 4I). The genetic interaction of a magnesium importer mutant with murine host Mcoln2 genotype (p = 0.0002) leads us to conclude that murine Mcoln2, like human MCOLN2, affects Mg 2+ accessibility by Salmonella during infection. However, the comparative growth disadvantage of the S. Typhimurium double importer mutant in Mcoln2 knockout mice contrasting with the comparative growth advantage of wild-type S. Typhi in MCOLN2 +/+ human THP-1 cells (see Figure 4G) suggests that mucolipin-2's impact on Mg 2+ availability during infection depends on context, likely including Salmonella serovar and host species as well as the infected cell type or tissue. Despite these differences, the in vivo and in vitro data concur that mucolipin-2 changes Salmonellae replication by altering their access to Mg 2+ .
In the simplest version of our model, removing human MCOLN2 increases Mg 2+ availability to S. Typhi, which relieves a nutrient limitation and directly increases bacterial replication. However, $1/3 of the increased bacterial replication inside MCOLN2 knockout cells is not explained by manipulating Mg 2+ availability or uptake. We theorized that this putatively Mg 2+ -independent replication boost in MCOLN2 À/À cells could still be PhoP regulated, as we had already identified other PhoP-targets, namely SPI-2 T3SS, which further benefit S. Typhi replication when MCOLN2 is knocked out. To test this, we repleted Mg 2+ after infecting THP-1s with S. Typhi DphoPQ ( Figure 4J). Mg 2+ increased replication of DphoPQ bacteria, as it partially overcomes the inability to fully upregulate mgtA and mgtB. Furthermore, the combined Mg 2+ repletion and phoPQ deletion removed any discernable difference in S. Typhi replication between MCOLN2 genotypes. Thus, enhanced bacterial replication in the absence of MCOLN2 depends on both Mg 2+ -independent effects of PhoPQ and PhoPQ-independent effects of Mg 2+ availability ( Figure 4K).

DISCUSSION
In this report, we directly connect expression of the divalent cation channel MCOLN2 with variable immune cell permissivity to S. Typhi. For S. Typhimurium, intracellular replication regulates outcomes in mouse models of enteric fever, 26,51 and, therefore, S. Typhi replication likely also correlates with disease outcome in humans. Unfortunately, there is no published GWAS of typhoid severity or clinical outcome and only one study on typhoid fever onset, which identified an association between the MHC region and susceptibility. 52 Thus, determining the clinical significance of rs10873679 in humans awaits well-powered studies for this disease phenotype. Our findings also underscore that, despite great insights gleaned from mouse models of S. Typhimurium infection, studies of genetic diversity using humanspecific pathogens in human cells provide unique insight.
Furthermore, we showed that MCOLN2 ablation reduced the low-Mg 2+ stress faced by intracellular S. Typhi based on lowered expression of Mg 2+ -regulated PhoP targets (including key Mg 2+ transporters) and reduced benefit of Mg 2+ repletion. Thus, the divalent cation channel MCOLN2 exerts restriction pressure on S. Typhi inside human monocytes by reducing Mg 2+ availability, which is similar to how the divalent cation transporter Slc11a1 (Nramp1) is proposed to restrict S. Typhimurium inside murine macrophages. 53 It is worth noting that C57BL/6J mice are highly susceptible to Salmonella due to a deleterious mutation in Slc11a1, which means divalent cation transport in their immune cells is already disrupted in a way that advantages S. Typhimurium replication. 54 It is possible that future work will find a greater or different effect of Mcoln2 in mice with functional Slc11a1. Despite the similarity of proposed mechanisms for the effects of MCOLN2 and Slc11a1, transport of Mg 2+ by Slc11a1 has never been demonstrated nor has human SLC11A1 ever been shown to restrict Salmonella replication. This underscores the importance of our discovery that MCOLN2 is a bona fide Mg 2+ channel between endolysosomes and the cytosol, as it bolsters our genetic and functional evidence of Mg 2+ -based nutritional immunity (F) S. Typhi DssaT has no effect on intracellular replication in WT THP-1s and partially accounts for the requirement of phoPQ to achieve maximal replication in MCOLN2 À/À THP-1s. Ten replicates from three experiments. p values in (E) and (F) are from Sídá k's comparison of MCOLN2 +/+ to MCOLN2 À/À following two-way ANOVAs finding significant main effects and interaction (all p < 0.0001). Statistics in ( Article ll OPEN ACCESS against intracellular Salmonella. Thus, our multi-disciplinary approach to understanding human variation, which revealed the first common human genetic difference that regulates intracellular resistance to Salmonella, has also led to the identification of the critical host factor that restricts S. Typhi by Mg 2+ deprivation.
Identifying human MCOLN2 as a host factor that drastically reduces Salmonella replication by lowering Mg 2+ availability highlights the key role played by Mg 2+ in nutritional immunity. This builds on a line of work identifying the sophisticated regulatory network in S. Typhimurium that allows it to respond to the low-Mg 2+ environment of the SCV. 37, 55 Notably, these investigations into Salmonella response to low Mg 2+ have been conducted with non-typhoidal S. Typhimurium. While much of this regulatory system is likely preserved in S. Typhi, the much greater sensitivity of S. Typhi to MCOLN2 ablation suggests that some component of this low Mg 2+ response is not conserved between the serovars. Future studies investigating this difference could reveal key serovar-specific virulence strategies.
Our finding that MCOLN2 restricts S. Typhi also explains why it is an ISG, despite previous findings that it increases macrophage susceptibility to endocytosed viruses including influenza A virus (Orthomyxoviridae) and yellow fever virus (Flaviviridae). 56 The induction of MCOLN2 expression in activated immune cells therefore provides two mechanisms whereby this channel could regulate infection-Ca 2+ currents regulating endocytic events and Mg 2+ currents affecting Mg 2+ acquisition. This identifies the MCOLN2 locus as a possible site of balancing selection be-tween different infectious disease pressures-viruses that use the endocytic pathway for entry might select for people with less MCOLN2 expression, while Salmonellae infections might select for people with more MCOLN2 expression. This balancing selection could explain the wide distribution of both rs10873679 alleles in populations around the world, and, ultimately, it highlights the persistent and complex power of infectious disease as an evolutionary pressure shaping human evolution.

Limitations of the study
Our genetic association work in this study is limited to LCLs. Therefore, the association of rs10873679 with S. Typhi replication will need to be examined in other cell types, with varying immune cell polarization, and ultimately in human populations. Similarly, the functional studies of MCOLN2 were consistent in LCLs and THP-1 monocytes but have not been extended to other cell types. The MCOLN2 patch-clamp experiments were conducted using overexpression in HEK293 cells, and there may be differences with endogenous expression in immune cells. As noted above, the effects of MCOLN2 varies across different Salmonella enteria serovars, and future studies will need to define the mechanistic underpinnings of these differences.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

DECLARATION OF INTERESTS
The authors declare no competing interests.

INCLUSION AND DIVERSITY
We worked to ensure diversity in experimental samples through the selection of the cell lines. We support inclusive, diverse, and equitable conduct of research.

RESOURCE AVAILABILITY
Lead contact Further information, as well as plasmids and bacterial strains generated for this study, are available by request from the lead contact, Dennis C. Ko (dennis.ko@duke.edu).

Materials availability
Plasmids and bacterial strains, as listed in the key resources table, are available upon request.
Data and code availability Intracellular replication data for the 961 LCL samples can be found in Table S1, and GWAS summary statistics are available for download at the Duke Research Data Repository (Duke Research Data Repository: https://doi.org/10.7924/r4x92bd76). Intracellular S. Typhi RNA-seq data are available in GEO (GEO: GSE222194). The analyses of this data-differential gene expression and gene set enrichment analysis (GSEA)-are available in Table S2. In other replication experiments, the value of each biological replicate is shown as dots on top of bar graphs.

Human cells
Lymphoblastoid cell lines (LCLs; EBV-immortalized B cells) were from the Coriell Institute. MCOLN2 À/À and matched wild-type THP-1 cell pools were generated by Synthego using guide 5 0 -TTTTGGTTTAAGTAACCAGC-3 0 (PAM is TGG) to target the start of