HspX promotes the polar localization of mycobacterial protein aggregates

Misfolding of translated proteins occurs in all domains of life. In most cells, misfolded proteins coalesce in discrete aggregates at distinct cellular locations. In many bacteria, including mycobacteria, protein aggregates are located at the cellular pole. Yet the mechanism by which aggregates are sorted to the mycobacterial pole is not known. Here, we show that in Mycobacterium smegmatis, the small heat shock protein HspX plays a critical role in the polar localization of aggregates of a model fluorescent misfolded protein, GLR103. HspX itself has a polar localization, which is dependent on its N-terminal domain. In a strain deleted for hspX, GLR103 is less liable to aggregation and no longer localizes to the pole, and redirecting HspX to the septum radically disrupts the normal polar localization of GLR103 aggregates. To further investigate the role of HspX in native protein aggregation, we performed semi-quantitative mass-spectrometry of mycobacterial protein aggregates in wild-type, hspX-deleted and hspX-overexpressing strains. We identified a subset of proteins that appeared to be HspX-dependent for aggregate formation. Furthermore, we demonstrate that for validated native protein aggregates, sorting to the cellular pole following proteotoxic stress required HspX. In summary, we have identified the cellular function of HspX in Mycobacterium smegmatis as both a pro-aggregase and polar sortase.

and its pro-aggregase function could be recapitulated in vitro using a model substrate 15 . However, whether other sHSPs also perform pro-aggregase functions is not known.
Here, we investigated the role of HspX, a Mycobacterium smegmatis small heat shock protein in the formation and trafficking of protein aggregates. HspX was required for the polar localization of both a model aggregation-prone protein, and several native mycobacterial protein aggregates. Furthermore, we performed semi-quantitative mass-spectrometry of M. smegmatis protein aggregates. We identified a subset of aggregates that appeared to be HspX-dependent, suggesting that HspX acts both as a sortase and/or pro-aggregase for a subset of mycobacterial protein substrates.

Results
HspX is necessary for the polar localization of an aggregation-prone protein in M. smegmatis.
A number of fluorescent reporters have been used to study protein aggregation in bacterial and eukaryotic systems. Many, such as the widely used firefly luciferase 16 require physiological stress such as heat shock to promote aggregation of the reporter 9 , although mCerulean fused to ELK16 could aggregate without heatshock 16 . To construct a new reporter that was aggregation-prone in the absence of heat shock, we mutated a number of conserved residues in green fluorescent protein (GFP). Mutation of the highly conserved aspartate at residue 103 to asparagine resulted in both loss of green fluorescence by two orders of magnitude, and aggregation of the protein when expressed in wild-type Mycobacterium smegmatis. Since the resulting GFP-D103N was no longer fluorescent, we C-terminally fused the protein with a flexible linker and monomeric red fluorescent protein (mRFP), so as to be able to visualize the expressed product, which we named GLR103 (green-linker-red, 103 rd residue mutated), and its corresponding wild-type dual-fluorescent reporter as GLR. We confirmed that when expressed in M. smegmatis, a large proportion of GLR103 formed insoluble aggregates compared with GLR, as determined by differential ultra-centrifugation (Fig. S1a). Measurement of red fluorescence in the different sub-cellular fractions confirmed that GLR103 within the insoluble fraction, which would contain large aggregates, contained the major fluorescent fraction in wild-type cells (Fig. S1b), and therefore the reporter would be suitable for tracking aggregate formation by microscopy.
Expression of wild-type GLR in wild-type M. smegmatis resulted in distribution of the dual-fluorescent protein throughout the cell (Fig. 1a). By contrast, expression of GLR103 resulted in discrete foci of red fluorescence at the pole (Fig. 1a), in keeping with the known polar distribution of protein aggregates in bacteria including mycobacteria 6,7,16,17 . To determine whether the sHSP HspX was involved in aggregation of GLR103, we constructed a strain of M. smegmatis in which hspX was deleted, ∆hspX. Deletion of hspX did not cause a growth defect when the strains were grown under standard laboratory medium and conditions (Fig. S2). Unlike in wild-type M. smegmatis, expression of GLR103 in ∆hspX did not result in polar distribution of the fluorescent protein (Fig. 1b,c). When expressed in ∆hspX, GLR103 was less prone to aggregation (Fig. S1). Complementation of the deletion strain with hspX rescued the localization phenotype, confirming a role for HspX in the polar localization of large protein aggregates (Fig. 1b,c).

The N-terminus of mycobacterial HspX is necessary but not sufficient for its polar localization.
HspX has three domains, a conserved α-crystallin domain, flanked by N-and C-terminal domains [18][19][20][21] . The N-terminus of the M. tuberculosis homolog of HspX was reported to play a role in its chaperone activity, as well as self-oligomerization 20,22 . Furthermore, the N-terminus of a member of the yeast sHSP family, Hsp42, played a crucial role in peripheral aggregate formation 14,23 , and the N-terminal domains of other sHSPs have been reported to mediate substrate interaction and self-assembly 18,19,21 . We therefore decided to test the localization of C-terminally GFP-tagged HspX, with the N-terminal 35 residues deleted (∆N35-HspX-GFP), compared with the full-length protein (HspX-GFP), both expressed on an otherwise HspX-null background. Full-length HspX-GFP localized to the bacterial pole, and was functional since it complemented the hspX-null strain (Fig. S3a). However, deletion of the N-terminus abrogated polar localization of HspX-GFP ( Fig. 2a-c), or of GLR103 (Fig. S3b). This was not due to lower expression levels of the N35 deletion mutant (Fig. S4). Since sHSPs are known to assemble into homo-oligomers [20][21][22]24 , we wished to determine if self-assembly was associated with localization. We C-terminally tagged full-length HspX, ∆N35-HspX, and just the 35 N-terminal residues of HspX (N35) with monomericCherry fluorescent protein (mCherry). We then expressed these constructs +/− full-length HspX-GFP, on an otherwise ∆hspX background. As before with the GFP-tagged constructs, HspX-mCherry, but not ∆N35-HspX-mCherry localized to the bacterial poles, and only full-length HspX-mCherry was found in the insoluble fraction of the cells. The 35 N-terminal residues of HspX were not sufficient for polar localization of N35-mCherry, suggesting that other molecular determinants in the full-length HspX protein were also necessary for its localization (Fig. S5). However, when co-expressed with full-length HspX-GFP, both HspX-mCherry and ∆N35-HspX-mCherry co-localized to the pole (Fig. 2d). These data suggested that full-length and N-terminally deleted HspX were able to co-assemble, and the hetero-oligomers were competent for polar localization (Fig. 2e).
HspX has a functional role in the localization of large mycobacterial protein aggregates. We wished to determine whether HspX played a causal role in polar aggregate formation. Given the generally ascribed chaperone role for sHSPs, one possibility was that HspX associates with misfolded proteins in a non-causal, but consequential role, similarly to the sHSPs IbpA and IbpB in E. coli 6 . Alternatively, HspX may play a directly causal role in the sorting of misfolded protein aggregates to the pole. To resolve these two hypotheses, we decided to deliver HspX to an alternate cellular location by fusing it with SepF, a protein known to localize to the mycobacterial septum 25 . SepF-GFP localized to the septum (Fig. S6a), but did not alter localization of GLR103 (Fig. S6b). SepF-HspX-GFP showed very little expression, for reasons that were not clear, therefore we fused SepF to HspX alone. When co-expressed with GLR103 on an hspX null background, SepF-HspX significantly altered the localization of the GLR103 aggregates to multiple cellular locations, including to the cellular septum (Fig. 3a,b), and   www.nature.com/scientificreports www.nature.com/scientificreports/ aggregates were significantly less associated with the poles (Fig. 3c). Taken together, these data support a critical role for mycobacterial HspX in the sorting and distribution of model protein aggregates to the bacterial pole.

HspX expression promotes the aggregation and polar sorting of a subset of cellular proteins.
To investigate whether HspX played a role in the formation of aggregates of native mycobacterial proteins, we isolated the insoluble protein fraction, representing aggregates, in wild-type M. smegmatis, and in strains deleted for, and over-expressing hspX (Fig. S7) and subjected them to semi-quantitative mass-spectrometry by tandem mass-tags. In total, the insoluble fraction contained 1355 proteins with >1 unique peptides that were identified in all three conditions (Table S1). Only a subset of these proteins varied quantitatively by hspX expression. The vast majority of proteins that varied were over-represented in hspX over-expression, or wild-type M. smegmatis compared with ∆hspX. Of these, 251 varied by over 1.5-fold in the hspX over-expression strain (HspX-OE) compared with ∆hspX (Table S2). Of note, most proteins that were over-represented in HspX-OE were also over-represented in wild-type M. smegmatis compared with ∆hspX, but at a smaller differential ratio (Table S2), suggesting a causal role for HspX in the generation of aggregates in this subset of proteins.
For the model protein GLR103, HspX sorted GLR103 aggregates to the mycobacterial poles. Did HspX perform a similar function in any native proteins? We tested 4 candidates that were identified in the mass spectrometry analysis as present in the insoluble fraction by C-terminally fusing them with fluorescent mApple and visualizing by microscopy (Figs 4 and S8a). When grown under standard laboratory conditions, none of the four proteins formed large, visible aggregates (Fig. 4a). However, following a heat-shock pulse, three of the four candidates, with the exception of EF-Tu (coded by tuf) formed large visible aggregates that were localized at the cellular pole in an HspX-dependent manner (Figs 4b and S8b). Therefore HspX also acts as a polar sortase for native, aggregation-prone mycobacterial proteins.

Discussion
Protein misfolding may occur due to environmental stressors, such as oxidation or heat shock, or due to translational errors from antibiotics, mutations in the translation apparatus, or stochastic mistakes 1,7,[26][27][28][29][30] . The polar localization of protein aggregates in bacteria, including mycobacteria is a long-observed phenomenon, but the molecular mechanisms for sorting misfolded proteins to the poles is less clearly understood. In E. coli, the sHSPs IbpA and IbpB were associated with protein aggregate formation 6,10 , and co-localized with inclusion bodies at the cellular poles. However, the sorting of aggregates in E. coli to the poles appeared to be a passive process, driven predominantly by nucleoid occlusion 17 . In mycobacteria, depletion of the HSP70 DnaK led to a marked increase in aggregate formation, consistent with its known critical role as a disaggregase 16 . This was associated with upregulation and redistribution of the Hsp100 homologue ClpB 31 to discrete foci 16 . ClpB was also shown to associate with irreversibly carbonylated proteins that result from oxidation 7 . However, expression of an enzymatically null ClpB did not associate with aggregates, nor was functional ClpB required for aggregate formation or polar localization 7 , suggesting that ClpB, as a refoldase, was necessary for dealing with aggregate-induced stress, but not the formation or sorting of aggregates to the mycobacterial pole. Similarly to Hsp42 in yeast 13,14 , mycobacterial HspX is required for the polar localization of both a model aggregation-prone protein, GLR103 and native protein aggregates (Figs 1 and 4). Redistribution of HspX results in mis-sorting of GLR103, verifying its active role in sorting this model protein (Fig. 3). Deletion of HspX led a much smaller fraction of GLR103 being found in the insoluble fraction. Therefore, in addition to its function as a sortase for this model protein, it seems likely that HspX acted as a pro-aggregase in this context.
The subdomains of sHSPs, including HspX have been shown to be critically important for their functions. Both N-and C-terminal domains of HspX have been shown to contribute to oligomerization and chaperone functions in vitro [20][21][22]24,32 . The N-terminus of yeast Hsp42 contains both a prion-like domain and an intrinsically disordered domain, which contribute to CytoQ aggregate formation and localization 23 . By expressing differentially-tagged HspX and truncation proteins in M. smegmatis cells lacking endogenous HspX, we showed that N-terminal domain of HspX was necessary, but not sufficient for its polar localization. However, HspX lacking the N-terminal domain could co-localize to the pole when co-expressed with wild-type HspX -suggesting that it is the oligomeric form of the protein that is located at the poles, and that the N-terminally truncated protein can form hetero-oligomers with full-length HspX (Fig. 2). Whether oligomeric HspX is sufficient for polar localization, or requires binding to other partners to sort to the pole is not known.
We performed semi-quantitative mass-spectrometry of M. smegmatis aggregates. A prior study on the role of DnaK in mycobacterial aggregate formation showed that depletion of DnaK appeared to cause large, global changes in mycobacterial aggregates, although formal quantitation was not performed 16 . By contrast, deletion or over-expression of HspX did not quantitatively change the majority of detected insoluble proteins (Table S1). This is perhaps not surprising for two reasons. The experiment was performed in the absence of proteotoxic stress. Also, the function of DnaK as a disaggregase will likely have a greater impact on the quantity and number of aggregated proteins compared with HspX in the absence of proteotoxic stress.
We chose to perform the experiment in the absence of proteotoxic stress to identify "aggregation-prone" client proteins of HspX. Some cellular proteins were under-represented in aggregates when hspX was deleted and were over-represented when hspX was over-expressed (Table S2), suggesting that at least for this subset of proteins, HspX may act as an aggregase. However, as already noted the experiment was performed under standard laboratory growth conditions, and it is possible that with greater proteotoxic stress, the number of HspX client proteins may be increased. We tagged a number of proteins that we identified as present in M. smegmatis aggregates with a fluorescent protein to allow visualization. Under standard laboratory growth conditions, no aggregates were visualized by microscopy: suggesting that the majority of tagged protein molecules had folded appropriately and were soluble, although we could not exclude small aggregates that were not visually distinguishable trafficking to the pole. However, under proteotoxic stress (heat shock pulse), three of the four proteins tested formed large visible Scientific RepoRtS | (2019) 9:14571 | https://doi.org/10.1038/s41598-019-51132-w www.nature.com/scientificreports www.nature.com/scientificreports/ aggregates, suggesting that our dataset probably contains "aggregation-prone" proteins, and formation and/or sorting of these aggregates to the mycobacterial poles was HspX-dependent, confirming its role as an aggregate polar sortase for native mycobacterial proteins.
HspX has been intensively studied. It is expressed abundantly under clinically relevant environmental stressors such as hypoxia due its regulation by the DosRST system [33][34][35][36] , and some tuberculosis vaccine candidates include HspX as an immunogen 37 . The cellular function of HspX, however, has not been characterized to date. www.nature.com/scientificreports www.nature.com/scientificreports/ Deletion of aggregation-associated proteins lead to fitness defects under stress in a number of model systems. In yeast, deletion of Hsp42 led to a specific fitness defect upon cyclic heat shock, but not when strains were grown at 30 °C 15 , and this was dependent on the protein's N-terminal disordered domains 23 . Similarly, ClpB depletion in M. tuberculosis resulted in a defect in recovery from late stationary phase and in mouse lungs 7 . Deletion or chemical inhibition in vitro of HspX in M. tuberculosis decreased viability in hypoxia 36,38 . When mice were infected by an HspX-deficient strain of M. tuberculosis, the strain grew to a greater organ burden compared with the wild-type parent strain, but was more rapidly cleared upon antibiotic treatment, suggesting that HspX played potentially opposing roles in fitness to the host versus antibiotics in that model 39 . By sorting insoluble protein aggregates to a distinct cellular location, HspX acts to compartmentalize the cell in the absence of membrane-bound organelles. As such, HspX may be a key mycobacterial driver of phase separation 40,41 , which has been associated with potentially adaptive functions of aggregate formation 42 .
All genes were cloned by PCR from M. smegmatis genomic DNA or a template plasmid and verified by Sanger sequencing (see Table). The fluorescent protein reporter GLR was constructed by fusion PCR and GLR103 was constructed by site-directed-mutation PCR. HspX-GFP was made by fusing GFP to the C-terminal of HspX by PCR. DeltaN-HspX-GFP was made by deleting the first 35 amino acids of HspX based on HspX-GFP by PCR. HspX-mCherry was made by fusing mCherry to the C-terminal of HspX by PCR. DeltaN-HspX-mCherry was made by deleting the first 35 amino acids of HspX based on HspX-mCherry by PCR. N35-HspX-mCherry was made by fusing mCherry to the N-terminal first 35 amino acids of HspX by PCR. SepF fragment was got by PCR using mc 2 155 strain genomic DNA as template. SepF-HspX was made by fusing HspX to the C-terminus of SepF. Initially, a construct, SepF-HspX-GFP was made, but this failed to localize to the septum for unknown reasons, and therefore the SepF-HspX construct was tested. The mApple fragment was obtained by PCR (with start codon deleted) using pMV261-mApple (a kind gift from Eric J. Rubin) as template and then inserted into plasmid pSML1357 by enzyme digestion (NheI and HindIII) and ligation. Msmeg_6339, murA and tuf (with stop codons deleted) were obtained by PCR using M. smegmatis genomic DNA as template and then inserted into plasmid pSML1357 by enzyme digestion (PacI and NheI) immediately 5′ to mApple to make a fusion protein of Msmeg_6339/MurA/EF-Tu-mApple. Plasmids were transformed M. smegmatis competent cells by electroporation using standard methodology. Growth of bacteria was measured by measuring the optical density of bacteria media at 600 nm wavelength. Plasmids and primers are listed in Tables 1 and 2 respectively. hspX deletion strain construction. The M. smegmatis gene hspX (MSMEG_3932) was deleted through linear dsDNA guided homologous recombineering 30,43,44 . Briefly, a Zeocin resistant cassette flanking the 500 bp upstream and downstream region of MSMEG_3932 was used as an allele exchange substrate (AES). The upstream 500 bp and downstream 500 bp of MSMEG_3932 were amplified from boiled mc 2 155 strain by PCR. The Zeocin resistant cassette was amplified from a plasmid pKM-Zeo-lox (A kind gift from Eric J. Rubin). The three fragments were fused together by overlap-PCR and verified by sequencing. M.smegmatis mc 2 155 transformed with the recombinase expressing plasmid pNIT(KanR)::recET::sacBR was induced to express RecET recombinase overnight with 10 nM Isovaleronitrile (IVN), from which competent cells were prepared and then electroporated with the KO cassette. The transformants were allowed to recover in 7H9 for 4 hours and then grow on LB-agar plates supplemented with 25 μg/mL Zeocin. The strain construction was confirmed by PCR, which verified both integration of the cassette and that it was in the appropriate genomic context. The HspX deleted strain was cured of the recombinase expressing plasmid before downstream applications.

Subcellular fractionation.
Methods were adopted from 16 with a few modifications. Briefly, mycobacterial cultures of interest were pelleted at 4 °C and washed with fractionation buffer (50 mM Tris, 100 mM NaCl, 10% glycerol, 2 mM EDTA, pH 8, supplemented with protease inhibitor cocktails (Roche)). Cell pellets were re-suspended in 1 mL fractionation buffer and subjected to four times of 40 seconds beads beating (MiniBeadbeater-16, 230 V. BioSpec) with 90 seconds intervals on ice in between. Beads and non-lysed cells were discarded by centrifugation at 3,000 × g for 5 minutes at 4 °C, the supernatant was collected and another round of low speed (5,000 × g) centrifugation for 20 minutes at 4 °C was used to remove the residual debris. Supernatants were collected and transferred to 1.5 mL Beckman tubes, and then centrifuged at 180,000Xg for 2.5 hours at 4 °C. Supernatants from this step were collected, labeled as "Soluble" fraction and kept at −20C. Pellets were carefully resuspended in 0.5 mL Scientific RepoRtS | (2019) 9:14571 | https://doi.org/10.1038/s41598-019-51132-w www.nature.com/scientificreports www.nature.com/scientificreports/ ice-cold fractionation buffer supplemented with 1% TritonX-100, allowed to solubilize at 4 °C with gentle shaking for 1 hour. The whole suspension was then centrifuged at 20,000 × g for 2 hours at 4 °C. The supernatant of this fractionation was collected and labeled as "Membrane" fraction, the pellets were dissolved in ice-cold fractionation buffer supplemented with 1% SDS, and labeled as "Pelleted" fraction. All three fractions of different bacterial cultures were loaded onto an SDS-PAGE gel for electrophoresis. The GFP-mRFP (or mutant) proteins were visualized by western blotting using standard protocols and mouse mono-clonal anti-GFP antibody as above.
Quantitative mass-spectrometry. Two independent cultures of each of wild-type M. smegmatis, hspX knock-out and hspX over-expression strains were grown in complete 7H9 medium to late log phase and then lysed and processed as detailed above for "sub-cellular fractionation". The "insoluble" fractions were resuspended and solubilized using 8 M urea, quantified using Bradford assay and then equal quantities loaded onto an SDS gel and subjected to PAGE. Following SDS-PAGE, all the lanes were excised into bands, reduced with 25 mM dithiotreitol and alkylated with 55 mM iodoacetamide. In-gel digestion was then performed using sequencing grade modified trypsin in 50 mM ammonium carbonate buffer at 37 °C overnight. Peptides were extracted twice with 0.1% trifluoroacetic acid in 50% acetonitrile aqueous solution for 30 minutes. Extracts were then partially dried in a speedivac to reduce volume. Tryptic peptides were then redissolved in 50 µl 200 mM tetraethylammonium bromide (TEAB) and then 2 µl TMTsixplex reagent was added to each sample according to the manufacturer's instructions. The reaction mixture was incubated at room temperature for 1 hour, following which 0.5 µl of 5% hydroxylamine (pH 9) was added to the reaction mixture and incubated for 15 minutes to quench the reaction. Sep-Pak C18 Vac cartridges (Waters) were used for desalting. Equal amounts of labelled peptides were combined and then analyzed by LC-MS/MS as detailed below, and as previously described 45 .
For liquid chromatography−tandem mass spectrometry (LC−MS/MS) analysis, the TMT-labeled peptides were separated by a 120 min gradient elution at a flow rate 0.300 μL/min with the EASY-nLC 1000 system which was directly interfaced with a Thermo Orbitrap Fusion mass spectrometer. The analytical column was a 'home-made' fused silica capillary column (75 μm ID, 150 mm length; Upchurch, Oak Harbor, WA) packed with C-18 resin (300 A, 5 μm; Varian, Lexington, MA). Mobile phase A consisted of 0.1% formic acid, and mobile phase B consisted of 100% acetonitrile and 0.1% formic acid. The Orbitrap Fusion mass spectrometer was operated in the data-dependent acquisition mode using Xcalibur3.0 software. There was a single full-scan mass spectrum in the Orbitrap (350-1550 m/z, 120,000 resolution) followed by 3 seconds of data-dependent MS/MS scans in an ion routing multipole at 38% normalized collision energy. MS/MS spectra from each LC−MS/MS run were searched against the Mycobacterium smegmatis database using a Sequest HT algorithm in Proteome Discoverer software (PD, version 1.4). The search criteria were as follows: full tryptic specificity was required; two missed cleavages were allowed; oxidation (M) was set as a variable; carbamidomethylation (C) and TMTsixplex (K and N-terminal) were set as fixed modifications; precursor ion mass tolerance was set at 20 ppm for all MS and 20 mmu for all MS/MS spectra. The peptide false discovery rate (FDR) was estimated using percolator function provided by Proteome Discoverer (PD), and the cut-off score was accepted as 1% based on the decoy database. Relative protein quantification was performed using PD software according to the intensity of the TMT reporter ions. Peptides only assigned to a given protein group were considered as unique. Protein ratios were calculated as the median of all peptide hits belonging to a protein. Quantitative precision was expressed as protein ratio variability. Proteins with unique peptides >1 were included for analysis. The score of a certain protein in wild-type, ΔhspX or hspX over-expression strain was calculated according to the total score and the ratio between samples. 100x oil-immersion lens. Images were processed using FIJI (ImageJ) software. For the analysis, individual cells were picked manually (i.e. by the microscopist, not automated software) and data were extracted as fluorescence intensity per pixel along the cell lengths (or segments). All cells that could be confidently ascribed as individual cells within any chosen visual field were included in the analyses. Since cell lengths varies, the segment lengths were normalized and were displayed as a fractional distance. Quantification of pole and total cellular fluorescence intensity was calculated by the formula: ratio of fluorescence intensity (pole/total) = fluorescence of the region of interest (ROI) in pole-background ROI in pole/fluorescence of the region of interest (ROI) in total cell-background ROI in total cell, i.e. relative fluorescence of a region compared with average cell fluorescence. Statistical analysis of aggregation distribution and plotting of the data were performed with Excel (Microsoft) and R software. The number of bacilli analyzed is indicated in the figure legends. All experiments were performed at least 3 times independently on different days.