The iron-dependent repressor YtgR regulates the tryptophan salvage pathway through a bipartite mechanism of transcriptional control in Chlamydia trachomatis

During infection, pathogens are starved of essential nutrients such as iron and tryptophan by host immune effectors. Without conserved global stress response regulators, how the obligate intracellular bacterium Chlamydia trachomatis arrives at a physiologically similar “persistent” state in response to starvation of either nutrient remains unclear. Here, we report on the iron-dependent regulation of the trpRBA tryptophan salvage pathway in C. trachomatis. Iron starvation specifically induces trpBA expression from a novel promoter element within an intergenic region flanked by trpR and trpB. YtgR, the only known iron-dependent regulator in Chlamydia, can bind to the trpRBA intergenic region upstream of the alternative trpBA promoter to repress transcription. Simultaneously, YtgR binding promotes the termination of transcripts from the primary promoter upstream of trpR. This is the first description of an iron-dependent mechanism regulating prokaryotic tryptophan biosynthesis that may indicate the existence of novel approaches to gene regulation and stress response in Chlamydia.

Stephens et al., 1998; Thomson et al., 2008). Therefore, outside of TrpR-mediated 117 repression, the complete detail of trpRBA regulation remains poorly elucidated and 118 previous reports have indicated the possibility of more complex mechanisms of 119 regulation (Brinkworth, Wildung, & Carabeo, 2018). 120 In evaluating alternative regulatory modes of the trpRBA operon, an interesting 121 consideration is the pleiotropic effects induced by IFN-g stimulation of infected cells. 122 IFN-g is involved in many processes that limit iron and other essential biometals to 123 intracellular pathogens as a component of host nutritional immunity (Cassat & Skaar,124 2013; Hood & Skaar, 2012). Chlamydia have a strict iron dependence for normal 125 development, evidenced by the onset of persistence following prolonged iron limitation 126 (Raulston, 1997). Importantly, Chlamydia presumably acquire iron via vesicular 127 interactions between the chlamydial inclusion and slow-recycling transferrin (Tf)-128 containing endosomes (Ouellette & Carabeo, 2010). IFN-g is known to down-regulate 129 transferrin receptor (TfR) expression in both monocytes and epithelial cells with

trachomatis. 147
Consistent with the highly reduced capacity of the chlamydial genome, it is likely 148 that C. trachomatis has a limited ability to tailor a specific response to each individual 149 stress. In the absence of identifiable homologs for most global stress response 150 regulators in C. trachomatis, we hypothesized that primary stress responses to 151 pleiotropic insults may involve mechanisms of regulatory integration, whereby important 152 molecular pathways are co-regulated by stress-responsive transcription factors such 153 that they can be utilized across multiple host-mediated stresses. Here, we report on the 154 unique iron-dependent regulation of the trpRBA operon in Chlamydia trachomatis. We 155 propose a model of iron-dependent transcriptional regulation of trpRBA mediated by the 156 repressor YtgR binding specifically to the IGR, which may have implications for how C. 157 trachomatis responds to immunological and environmental insults. Such a mechanism 158 of iron-dependent regulation of Trp biosynthesis has not been previously described in 159 any other prokaryote and adds to the catalog of regulatory models for Trp biosynthetic 160 operons in bacteria. Further, it reveals a highly dynamic mode of regulatory integration 161 within the trpRBA operon, employing bipartite control at the transcription initiation and 162 termination steps. 163

Results 164
Brief iron limitation via 2,2-bipyridyl treatment yields iron-starved, but non-165 persistent Chlamydia trachomatis. To identify possible instances of regulatory 166 integration between iron and Trp starvation in C. trachomatis, we optimized a stress 167 response condition that preceded the development of a characteristically persistent 168 phenotype. We reasoned that in order to effectively identify regulatory integration, we 169 would need to investigate the bacterium under stressed, but not aberrant, growth 170 conditions such that we could distinguish primary stress responses from abnormal 171 growth. To specifically investigate the possible contribution of iron limitation to a broader 172 immunological (e.g. IFN-g-mediated) stress, we utilized the membrane-permeable iron 173 chelator 2,2-bipyridyl (Bpdl), which has the advantage of rapidly and homogeneously 174 starving C. trachomatis of iron (Thompson & Carabeo, 2011). We chose to starve C. 175 trachomatis serovar L2 of iron starting at 12 hrs post-infection (hpi), or roughly at the 176 beginning of mid-cycle growth. At this point the chlamydial organisms represent a 177 uniform population of replicative RBs that are fully competent, both transcriptionally and 178 translationally, to respond to stress. We treated infected HeLa cell cultures with 100 µM 179 Bpdl or mock for either 6 or 12 hours (hrs) to determine a condition sufficient to limit iron 180 to C. trachomatis without inducing hallmark persistent phenotypes. We stained infected 181 cells seeded on glass coverslips with convalescent human sera and analyzed 182 chlamydial inclusion morphology under both Bpdl-and mock-treated conditions by laser 183 point-scanning confocal microscopy ( Figure 1A). Following 6 hrs of Bpdl treatment, 184 chlamydial inclusions were largely indistinguishable from mock-treated inclusions, 185 containing a homogeneous population of larger organisms, consistent with RBs in mid-186 cycle growth. However, by 12 hrs of Bpdl treatment, the inclusions began to display 187 signs of aberrant growth: they were perceptibly smaller, more comparable in size to 18 188 hpi, and contained noticeably fewer organisms, perhaps indicating a delay in RB-to-EB there was no statistically distinguishable difference in genome copy number when 192 compared to the equivalent mock-treated time-point. However, by 12 hrs of treatment, 193 genome copy number was significantly reduced 4.7-fold in the Bpdl-treated group 194 relative to mock-treatment (p = 0.0033). We then assayed the transcript expression of 195 two markers for persistence by reverse transcription quantitative PCR (RT-qPCR): the 196 early gene euo, encoding a transcriptional repressor of late-cycle genes ( Figure 1C), 197 and the adhesin omcB, which is expressed late in the developmental cycle ( Figure 1D). 198 Characteristic persistence would display elevated euo expression late into infection, and 199 suppressed omcB expression throughout development. We observed that at 6 hrs of 200 Bpdl treatment, there was no statistically distinguishable difference in either euo or 201 omcB expression when compared to the mock-treatment. Still at 12 hrs of Bpdl 202 treatment, euo expression was unchanged. However, omcB expression was 203 significantly induced following 12 hrs of Bpdl-treatment (p = 0.00015). This was 204 unexpected, but we note that omcB expression has been shown to vary between 205 chlamydial serovars and species when starved for iron (Pokorzynski et al., 2017). 206 Collectively, these data indicated that 6 hrs of Bpdl treatment was a more suitable time-207 point at which to monitor iron-limited stress responses. 208 We additionally assayed these same metrics following 6 or 12 hrs of Trp 209 starvation by culturing cells in either Trp-replete or Trp-deplete DMEM-F12 media 210 supplemented with fetal bovine serum (FBS) pre-dialyzed to remove amino acids. We 211 observed no discernable change in inclusion morphology out to 12 hrs of Trp starvation 212 infected HeLa cells at the indicated times post-infection under iron-replete (blue) and -247 deplete (red) conditions. Chlamydial genome copy number was quantified by qPCR.

248
Chlamydial genome replication is stalled following 12 hours of Bpdl treatment, but not 6. 249 N=2. (C) Total RNA was harvested from infected HeLa cells at the indicated times post-250 infection under iron-replete (teal) and -deplete (orange) conditions. The transcript 251 abundance of hallmark persistence genes euo and (D) omcB were quantified by RT-252 qPCR and normalized against genome copy number. Only at 12 hours of Bpdl 253 treatment is omcB expression significantly affected. N=3 for 12+6, N=2 for 12+12.

A B C D
We next sought to determine whether our brief 6-hr Bpdl treatment was sufficient 258 to elicit a transcriptional iron starvation phenotype. We chose to analyze the expression 259 of three previously identified iron-regulated transcripts, ytgA (  pathway, was not observed to significantly respond to our brief iron limitation condition, 282 suggesting that it is not a component of the primary iron starvation stress response in C. 283 trachomatis. As expected, the transcript expression of dnaB, a replicative DNA helicase, 284 was not altered by our iron starvation condition, consistent with its presumably iron-285 independent regulation (Brinkworth et al., 2018). Overall, these data confirmed that our The transcript abundance of iron-regulated ytgA, (B) ahpC, (C) devB and (D) non-iron 324 regulated dnaB were quantified by RT-qPCR and normalized against genome copy 325 number. The transcript expression of ytgA and ahpC were significantly elevated 326 following 6-hour Bpdl treatment, indicative or iron starvation to C. trachomatis. N=3. 327 Statistical significance was determined by One-Way ANOVA followed by post-hoc 328 pairwise t-tests with Bonferroni's correction for multiple comparisons. * = p < 0.05, ** = p 329 < 0.01, *** = p < 0.005. To decouple Trp limitation from iron limitation and assess their relative 353 contribution to regulating a critical pathway for responding to IFN-g-mediated stress, we 354 monitored the transcript expression of the trpRBA operon under brief Trp or iron 355 starvation by RT-qPCR. When starved for Trp for 6 hrs, we observed that the 356 expression of trpR, trpB and trpA were all significantly induced greater than 10.5-fold 357 relative to 12 hpi (p = 0.00077, 0.025 and 9.7e-5, respectively; Figure 3B). All three 358 ORFs were also significantly elevated relative to the equivalent mock-treated time-point 359 is nonetheless interesting to note that only 6 hrs of media replacement is sufficient to 366 markedly up-regulate trpRBA expression. This suggests that C. trachomatis has a 367 highly attuned sensitivity to even moderate changes in Trp levels. under Trp-replete (black) and -deplete (red) conditions. The transcript expression of 419 trpRBA operon was quantified by RT-qPCR and normalized against genome copy 420 number. All three ORFs are significantly induced relative to 12 hpi following Trp 421 starvation. N=2. (C) Total RNA and gDNA were harvested from infected HeLa cells at 422 the indicated times post-infection under iron-replete (blue) and -deplete (red) conditions. 423 The transcript expression of trpRBA operon was quantified by RT-qPCR and normalized 424 against genome copy number. Only trpB and trpA expression was significantly induced 425 relative to 12 hpi. N=3. Statistical significance was determined by One-Way ANOVA 426 followed by post-hoc pairwise t-tests with Bonferroni's correction for multiple 427 comparisons. * = p < 0.05, ** = p < 0.01, *** = p < 0.005. Supplement 1A). Using this approach, we analyzed four conditions: 12 hpi, 18 hpi, 12 444 hpi + 6 hrs of Bpdl treatment, and 12 hpi + 6 hrs of Trp-depletion ( Figure 4A). We 445 observed three RACE products that migrated with an apparent size of 1.5, 1.1 and 1.0 446 kilobases (kb). At 12 and 18 hpi, all three RACE products exhibited low abundance, 447 even following the nested PCR amplification. This observation was consistent with the 448 expectation that the expression of the trpRBA operon is very low under normal, iron and 449 Trp-replete conditions. However, we note that the 6-hr difference in development did 450 appear to alter the representation of the 5' cDNA ends, which may suggest a stage-451 specific promoter utilization within the trpRBA operon. In our Trp starvation condition, 452 we observed an apparent increase in the abundance of the 1.5 kb RACE product, which 453 was therefore presumed to represent the primary TSS upstream of trpR, at nucleotide  If iron depletion was inducing trpBA expression independent of trpR, we 461 reasoned that we would observe specific enrichment of trpB sequences in our 5'-RACE 462 cDNA samples relative to trpR sequences. We again utilized RT-qPCR to quantify the 463 abundance of trpB transcripts relative to trpR transcripts in the 5'-RACE total RNA 464 samples ( Figure 4B). In agreement with our model, only under iron starved conditions 465 did we observe a significant enrichment of trpB relative to trpR (p < 0.01). Additionally, 466 we observed that at 12 and 18 hpi in iron-replete conditions, the ratio of trpB to trpR was 467 approximately 1.0, suggesting non-preferential basal expression across the three 468 putative TSSs. Another factor contributing to this ratio is the synthesis of the full-length 469 trpRBA polycistron. In support of this, the trpB to trpR ratio remained near 1.0 under the 470 Trp-starved condition, which would be expected during transcription read-through of the 471 whole operon. The apparent lack of preferential promoter utilization as described above 472 could be attributed to the relatively low basal expression of the operon at 12 and 18 hpi 473 under Trp-and iron-replete conditions, thus precluding quantitative detection of 474 differential promoter utilization in this assay. 475 To determine the specific location of the 5' cDNA ends within the trpRBA operon, 476 we isolated the 5'-RACE products across all conditions by gel extraction and cloned the 477 products into the pRACE vector supplied by the manufacturer. We then sequenced the 478  -IGR3). Surprisingly, we observed that none of these 610 fragments alone were capable of producing a significant repression phenotype in our 611 reporter system. This finding indicated that while the operator site was necessary for 612 YtgR repression, it alone was not sufficient. Together, these data indicated that YtgR 613 could bind to the trpBA promoter element and that this binding was dependent upon an 614 intact AT-rich palindromic sequence, likely representing an YtgR operator, but that 615 further structural elements in the trpRBA IGR may be necessary for repression. 616 Nonetheless, we demonstrated the existence of a functional YtgR binding site that 617 conferred iron-dependent transcriptional regulation to trpBA, independent of the major 618 trpR promoter. implying that the terminated transcript species are rare. We therefore turned to more 677 sensitive and quantitative methods to interrogate possible transcript termination within 678 the trpRBA IGR. 679 To identify transcription termination sites (TTSs) in the trpRBA operon in C.  To identify the specific TTS locations, we gel extracted the four distinct 3'-RACE 696 bands across all conditions and cloned them into the pRACE sequencing vector as was 697 done for the 5'-RACE experiments. We then sequenced the inserted RACE products 698 and mapped them to the C. trachomatis L2 434/Bu genome ( Figure 6B). This revealed a 699 highly dynamic TTS landscape within the trpRBA IGR, which has not previously been 700 investigated (For a full description of mapped 3'-RACE products, see Supplementary  701 File 2). The 0.20 kb RACE product mapped to the 3'-end of the trpR CDS, with a mean

Iron limitation promotes transcription read-through at the YtgR operator site. 754
Given the observation that transcripts terminated at the YtgR operator site, we 755 hypothesized that YtgR binding may promote transcription termination at this locus. 756 Conversely, we hypothesized that inactivating YtgR DNA-binding by Bpdl treatment 757 would allow transcription to read through the YtgR operator site. To quantitatively 758 analyze the possibility that iron-depletion, and thus dissociation of YtgR, may facilitate 759 transcription read-through at the operator site, we utilized RT-qPCR to monitor the 760 abundance of various amplicons across the trpRBA operon ( Figure 7A). We quantified 761 these data in relation to a "read-through" normalization amplicon that, based on 5'-and 762 3'-RACE data, should only be represented when the full-length trpRBA message is 763 transcribed ( Figure 7A). The representation of a specific mRNA species relative to the 764 full-length transcript should therefore be interpretable through a simple ratio of the 765 experimental amplicon to the "read-through" amplicon. If an mRNA species is poorly 766 represented relative to the full-length transcript, the ratio should be approximately 1.0. 767 Conversely, if that species is over-represented relative to the full-length transcript, the 768 ratio should exceed 1.0 ( Figure 7B; left). Therefore, as each amplicon is increasingly 769 represented as a part of the full-length transcript, the ratio of the specific amplicon to the 770 normalization amplicon should approach 1.0 ( Figure 7B; right). 771 We first analyzed an amplicon from nucleotide 511,416 -531 to monitor the 772 relative abundance of transcript species associated with transcription initiating at the 773 trpR promoter ( Figure 7C). We observed that the representation of this amplicon was 774 not significantly altered following iron limitation relative to 12 hpi, suggesting that the 775 depletion of iron was not affecting initiation of transcription at the trpR promoter. 776 Interestingly, at 18 hpi, the representation ratio of this amplicon significantly shifted 777 further away from 1.0 (p = 0.00358), indicating that at 18 hpi this amplicon is 778 represented less as a component of read-through transcription relative to 12 hpi. As 779 expected, under Trp-deplete conditions, the representation ratio shifted significantly 780 closer to 1.0 (p = 0.00064), consistent with read-through transcription of the full-length 781 trpRBA message. 782 We then preformed the same analysis on an amplicon from nucleotide 511,639 -783 764, immediately upstream of the TTS at the YtgR operator site to monitor condition-784 dependent read-through at this site ( Figure 7D). We again observed that at 18 hpi, the 785 representation ratio was significantly increased (p = 0.01046), and following Trp-786 depletion, the ratio was significantly decreased (p = 0.00023), as expected. Notably, 787 and consistent with our hypothesis, we observed that the representation ratio of this 788 amplicon was also significantly closer to 1.0 following iron limitation (p = 0.00407), 789 suggesting that transcription read-through was increased at this site under iron limited 790 conditions. Indeed, if YtgR is dissociating from the operator site during iron depletion, a 791 greater proportion of transcripts would be expected to read-through this locus. 792 Finally, we analyzed an amplicon from nucleotide 513,856 -968, at the very 3'-793 end of trpA to asses changes in terminal transcription under our experimental conditions 794 ( Figure 7E). At 18 hpi, we observed a significant increase in the representation ratio of 795 this amplicon (p = 0.00476), which is likely attributable to both basal levels of alternative 796 transcription from the IGR as well as poor transcription read-through of the full-length 797 message. Following 6 hrs of Bpdl treatment, we also observed a significant increase in 798 the representation ratio of this amplicon (p = 0.01510), which supports the finding that 799 trpBA is being preferentially transcribed under this condition, distinct from the full-length 800 trpRBA transcript. We were only able to detect a marginal decrease in the 801 representation of this amplicon under Trp-depleted conditions (p = 0.07942), which may 802 suggest that the very 3'-end of trpRBA is under-represented relative to our 803 normalization amplicon, which falls within the middle of the operon. In fact, recent work 804 has reported on the relatively poor representation of 3'-end mRNAs in Chlamydia 805 Schematic representation of RT-qPCR analysis. On the left, a biological interpretation of 839 the ratio used to determine the relative transcript abundance is provided. When a 840 termination product (red) is poorly represented, e.g. read-through is high, the ratio of 841 specific amplicon (blue) to full-length transcript (grey) should be close to 1.0. When the 842 termination product is abundant, e.g. read-through is low, then the ratio should exceed 843 1.0. On the right, a graphical demonstration of this concept is provided using mock data. 844 The within the IGR which they suggested was responsible for trpBA expression (Carlson et 908 al., 2006). In these studies, we were unable to confirm the presence of the previously 909 suggesting that some other Trp-dependent mechanism may control transcription from 922 this site (Akers & Tan, 2006). Ultimately, our approach of investigating more immediate 923 responses to stress revealed previously unreported mechanisms functioning to regulate 924 Trp biosynthesis in C. trachomatis, underscoring the value of transient as opposed to 925 sustained induction of stress. 926 Another mechanism of regulation reported to control the chlamydial trpRBA 927 operon is Trp-dependent transcription attenuation. Based on sequence analysis, a 928 leader peptide has been annotated within the trpRBA IGR (Merino & Yanofsky, 2005). 929 Presumably, this functions analogously to the attenuator in the E. coli trpEDCBA 930 operon; Trp starvation causes ribosome stalling at sites of enriched Trp codons such 931 that specific RNA secondary structures form to facilitate RNAP read-thru of downstream 932 sequences -in this case, trpBA (Yanofsky, 1981  For C. trachomatis, iron limitation may therefore serve as a critical signal in the 984 LGT, inducing the expression of trpBA such that Trp is stockpiled from available indole, 985 allowing the pathogen to counteract impending IFN-g-mediated Trp starvation. We 986 suggest the possibility that iron limitation in the LGT may be a significant predictor of 987 successful pathogen colonization in the UGT. Unfortunately, testing these hypotheses in 988 cell culture models of infection presents a significant challenge. Evaluating rescue of 989 chlamydial growth in the presence of indole to specifically assess the iron-dependent 990 role of trpBA requires simultaneous Trp and iron depletion. The former ensures indole 991 utilization by the bacteria, and the latter de-represses YtgR-regulated trpBA expression. 992 In theory, this is feasible, but in practice the combined stress rapidly induces aberrant 993 development, muddying results obtained from such studies (data not shown). Ideally, 994 genetic approaches could be employed to distinguish the regulatory effects of YtgR 995 independent of TrpR. However, the genetic manipulation of trans-acting factors (e.g. 996 YtgR) will presumably have unpredictable off-target effects. Genetically altering cis-997 acting factors -such as operator sequences -is more feasible, but at present we lack 998 the information necessary to rationally mutate these sequences in C. trachomatis to 999 interrogate these questions. The tight regulatory coordination at both the transcription 1000 initiation and termination steps would likely mean any mutation in the cis-acting 1001 sequences would affect both processes indiscriminately. Furthermore, in vivo infection 1002 models present challenges: attempting to answer these questions will likely require the 1003 use of in vivo non-human primate studies, as mouse models of Chlamydia-infection do 1004 not recapitulate immune-mediated Trp starvation (Nelson et al., 2005). Ultimately, these 1005 limitations do not undermine the biological significance of an iron-dependent mode of 1006 regulating Trp salvage, given the critical role played by this pathway during infection. 1007 Finally, and of note, the expression of the ribonucleotide diphosphate-reductase-1008 encoding nrdAB was also recently shown to be iron-regulated in C. trachomatis 1009 (Brinkworth et al., 2018). The regulation of nrdAB is known to be mediated by the 1010 presumably deoxyribonucleotide-dependent transcriptional repressor NrdR, encoded 1011 distal to the nrdAB locus (Case, Akers, & Tan, 2011). As NrdR activity is not known to 1012 be modulated by iron availability, this raises the intriguing possibility that here too a 1013 unique iron-dependent mechanism of regulation may integrate the chlamydial stress 1014 response to promote a unified response across various stress conditions. Future studies 1015 may require more metabolomics-based approaches to thoroughly dissect the integration 1016 of these stress responses, as transcriptome analyses alone often miss broader, 1017 pathway-oriented metabolic coordination. Ultimately, these studies point towards a need 1018 to carefully re-evaluate the molecular stress response in Chlamydia, with greater 1019 emphasis on the use of targeted approaches and treatment protocols that induce stress, 1020 but not persistence. We anticipate that the rapid progress of the field in recent years will 1021 cDNA was generated using either SuperScript® IV Reverse Transcriptase (RT-qPCR; 1110 ThermoFisher Scientific) or SMARTScribe™ Reverse Transcriptase (RACE and RACE-1111 specific qRT-PCR); Takara Bio, Kusatsu, Shiga Prefecture, Japan) essentially as 1112 described by the respective manufacturers. For cDNA generated for RT-qPCR, 650 ng 1113 of total RNA was used as a template in a 20 µL total reaction volume. For every RT 1114 reaction, a "no-RT" control, generated from 350 ng of total RNA template in a 10 µL total 1115 volume, was included. For 5'-RACE, cDNA was generated from 250 ng of total RNA 1116 using random primers in a 10 µL total volume and further processed in the RACE 1117 workflow. cDNA was stored at -20° C. 1118 1119 gDNA was harvested from C. trachomatis-infected HeLa cell monolayers by scraping 3 1120 wells of a 6-well plate in ice-cold PBS + 10% Proteinase K (ThermoFisher Scientific). 1121 Samples were then pooled and split into two technical replicates for analysis of genome 1122 copy number by qPCR. gDNA was isolated using the DNeasy Blood and Tissue Kit 1123 following manufacturer protocols (QIAGEN, Hilden, Germany). gDNA was stored at -20° 1124 C until further use.

1126
Reverse Transcription Quantitative Polymerase Chain Reaction (RT-qPCR) 1127 cDNA (or gDNA in qPCR), prepared as described above, was diluted 1:10 or 1:100 in 1128 nuclease-free H2O depending on the experimental condition being assayed (e.g. analysis to ensure that a single product was generated. For each primer set, a standard 1137 curve was generated using purified C. trachomatis L2 gDNA from EB preparations 1138 diluted from 2 x 10 -3 to 2 x10 0 ng per reaction. Ct values generated from each 1139 experimental reaction were then fit to standard curves (satisfying an efficiency of 1140 95±5%) for the respective primer pair and from the calculated ng quantities, transcript or 1141 genome copy number was calculated as follows: Where df = dilution factor and the number of copies/ng DNA is calculated based on the 1148 size of the C. trachomatis L2 genome assuming that the molar mass per base pair is 1149 650 (g/mol)/bp (note that this value should be the same for any single-copy ORF on the 1150 genome). All quantifications of genome copy number were determined using the ahpC 1151 overnight at 37° C in LB liquid broth containing 50 µg/mL carbenicillin and plasmids catabolite repression of expression vectors), 50 µg/mL carbenicillin and 15 µg/mL