Competition Among Gardnerella Subgroups From the Human Vaginal Microbiome

Gardnerella spp. are hallmarks of bacterial vaginosis, a clinically significant dysbiosis of the vaginal microbiome. Gardnerella has four subgroups (A, B, C, and D) based on cpn60 sequences. Multiple subgroups are often detected in individual women, and interactions between these subgroups are expected to influence their population dynamics and associated clinical signs and symptoms of bacterial vaginosis. In the present study, contact-independent and contact-dependent interactions between the four Gardnerella subgroups were investigated in vitro. The cell free supernatants of mono- and co-cultures had no effect on growth rates of the Gardnerella subgroups suggesting that there are no contact-independent interactions (and no contest competition). For contact-dependent interactions, mixed communities of 2, 3, or 4 subgroups were created and the initial (0 h) and final population sizes (48 h) were quantified using subgroup-specific PCR. Compared to the null hypothesis of neutral interactions, most (69.3%) of the mixed communities exhibited competition. Competition reduced the growth rates of subgroups A, B, and C. In contrast, the growth rate of subgroup D increased in the presence of the other subgroups. All subgroups were able to form biofilm alone and in mixed communities. Our study suggests that there is scramble competition among Gardnerella subgroups, which likely contributes to the observed distributions of Gardnerella spp. in vaginal microbiomes and the formation of the multispecies biofilms characteristic of bacterial vaginosis.


Introduction
replicate experiments, we grew the four subgroups alone (n = 4; A, B, C, D), and in all possible 145 combinations of two (n = 6; AB, AC, AD, BC, BD, CD), three (n = 4; ABC, ABD, ACD, BCD) 146 and four subgroups (n = 1; ABCD) for a total of 15 different combinations. Each of the 15 147 combinations was replicated three times in the wells of a 96-well tissue culture plate (i.e. 4 148 experiments *15 combinations * 3 replicates per combination = 180 replicates). The members of 149 each community were allowed to interact for a period of 48 hours and the abundance of the 150 constituent subgroups was estimated at the start (0 h) and the end (48 h) of this period using 151 subgroup-specific quantitative real-time PCR (qPCR). Prior to the interaction assay, each of the 152 four subgroups was grown alone at 37 °C anaerobically in BHI with 0.25% maltose and 10% 153 horse serum for a period of 12 h and then mixed in BHI+ 0.25% maltose. Immediately prior to 154 combining the subgroups to create the mixed communities, a sub-sample was taken from each of 155 the four cultures to determine the abundance of each subgroup at the time point of 0 h using the 156 subgroup-specific qPCR. To create the mixed communities, equal volumes of each isolate 9 incubated for 30 minutes to solubilize the biofilm. The bottom of each well was scraped with a 168 pipette tip and the suspension was pipetted up and down several times before transferring it to a 169 bead tube. The biofilm solubilization step was repeated to maximize biofilm collection. To 170 enhance lysis, 100 μl of chaotropic agent FB was added to the bead tubes. The bead tubes were 171 incubated at 65˚C for 5 minutes in a water bath, and then vortexed using a multitube vortexer at 172 maximum speed for 15 min. Later steps were performed following the manufacturer's 173 instructions. 174 Subgroup-specific quantitative real-time PCR was performed using previously published primers 175 and probes [8] (Table S2). Amplification was performed in 10 μl reactions containing 1× iQ 176 Supermix (BioRad, Mississauga, ON), 800 nM of each primer, 100 nM of TaqMan probe, and 2 177 μl of template. The qPCRs were performed using a CFX Connect (BioRad, Mississauga, ON) 178 instrument. The qPCR results were reported as target copy number per PCR reaction (2 µl of 179 template DNA extract). Each sample was assayed in duplicate reactions with the appropriate 180 standard curve comprised of plasmids containing probe targets (10 2 to 10 9 plasmid copies per 181 reaction). Thermocycling conditions were: initial denaturation at 95˚C for 3 min, 40 cycles of 182 95˚C for 15 s, and annealing/extension at 60˚C (subgroups A, B, and D) or 63.3˚C (subgroup C) 183 for 40 s. Each plate contained a no template control, DNA extraction controls, and non-target 184 subgroup templates as negative controls. For each qPCR reaction, the genome copy number was 185 calculated using the standard curve. The qPCR assay was repeated for samples with a difference 186 in Cq value > 1 between duplicate wells. 187 188 The contact-independent interactions were analyzed using Kruskal-Wallis nonparametric one- 189 way ANOVA with Dunn's post hoc test (Prism 8, Graphpad Software). abundance of a mixed community was higher than the expected null abundance, the interaction 206 was classified as a facilitation. If the observed abundance of a mixed community was lower than 207 the predicted null abundance, it was classified as a competition (Fig. 1). The null hypothesis of 208 no interaction predicts that due to random measurement error, 50% of the interactions should be 209 positive (facilitation) and 50% of the interactions should be negative (competition). A proportion 210 test was used to determine whether the observed prevalence of facilitation and competition were 211 significantly different from the 50/50 expectation. This approach is a general test of the nature of 212 interactions between Gardnerella subgroups and does not consider that each subgroup may be 213 affected differently by competitors. 214 To test whether the subgroups were affected differently by the number of competitors, the 215 growth rates of the Gardnerella subgroups were analyzed using linear mixed effect models 216 (LMMs). The residuals of the growth rates were considered as normally distributed. The fixed 217 effects were subgroup (four levels: A, B, C, and D), the number of competitors (0, 1, 2, and 3), 218 and their interaction. The random effects were the 3 replicates of each community nested in the 4 219 different experiments (i.e. total of 12 replicates for each community). This approach does not 220 consider the identity of the competitors. We used R (v 1.1-21) to analyze the data; the LMM 221 models were run using the lmer() function in the R package lme4. 222 To test whether the identity of the competitors mattered, the growth rate of each subgroup was 223 analyzed separately using LMMs. The four subgroups had to be analyzed separately, because the 224 identity of the competitors differs for each subgroup. The fixed effects were the identities of the 225 competitors. For example, for the growth rate of subgroup A, the competitors included B, C, D, 226 BC, BD, CD, and BCD. The random effects structure was the same as before.

229
Effect of Gardnerella culture supernatant on growth and biofilm formation 230 The initial optical density (OD) of all the focal strains was ~ 0.05 and they grew in both NYC III 231 and BHI + 1% glucose, except for one subgroup C strain, NR001, which did not grow in BHI+ 232 1% glucose (Fig S11). The OD at 48 h of these strains varied from as low as 0.05 (after 233 subtracting initial OD) to 0.80. There was no effect of CFS on overall growth or planktonic 234 growth of focal strains, nor on biofilm formation (P > 0.05 for all comparisons) (Fig S1-S14). 235 The type of medium, however, influenced mode of growth: NYC III had more planktonic 236 growth, whereas BHI + 1% glucose had more biofilm growth. Increasing the concentration of 237 CFS from 10% to 20% had no effect on the growth pattern of the Gardnerella subgroups (data 238 not shown).

239
Validation of qPCR assays 240 Prior to performing the co-culture experiments, the efficiency of each subgroup-specific qPCR 241 assay and the limits of detection and quantification were determined, since these values had not 242 been reported previously [8]. The amplification efficiency for subgroups A, B, C, and D were 243 99.9%, 107.4%, 110%, and 98.2%, respectively. The lowest concentration at which all subgroups 244 were detected was 1 target copy per qPCR reaction. However, the lower limit of quantification 245 (LOQ) was different for each subgroup. The LOQ for subgroups A, B, C and D were 1, 10, 100, 246 and 1 copy per reaction, respectively. competitors on the growth rate differs between subgroups. Growth rates of subgroups A, B, and 271 C decreased significantly (p <0.0001) in mixed communities (Fig. 2a,2b,2c). In contrast, the 272 growth rate of subgroup D increased significantly (p <0.0001) in mixed communities (Fig. 2d). 273 Thus, subgroups A, B, and C experienced competition in mixed communities, whereas subgroup 274 D experienced facilitation. Regardless of the identity of the community, subgroup C always has a 275 higher intrinsic growth rate than the other subgroups (Fig. 2).

287
Mixing of different bacterial species often leads to increased biofilm formation. We therefore 288 investigated whether mixing of Gardnerella subgroups would enhance biofilm formation. If the 289 amount of biofilm formed by a mixture of subgroups was greater than the amount formed by the 290 best individual biofilm former of that mixture, the interaction was considered synergistic. If the 291 amount of biofilm formed by a mixture was less than the amount formed by the worst individual 292 biofilm former of that mixture, the interaction was considered antagonistic [31]. Here biofilm 293 formation by mixed subgroups was almost always less than the individual biofilm formation by 294 the best biofilm-forming subgroup (Fig. 4 a, b). Only one co-culture of subgroups A and D was 295 higher than the individual biofilm formation of both strains (Fig. 4b). A proportion test found 296 that these proportions were not significantly (p >0.05) higher than 50/50 null expectation. The 297 results of this experiment show that overall biofilm biomass is not enhanced by mixing of 298 subgroups.

299
Effect of Gardnerella co-culture supernatant on individual subgroups 300 Our initial experiments showed that the CFS of singleton cultures had no impact on the growth 301 of isolates from other subgroups, but negative interactions were frequently observed in co-302 cultures. To determine if effectors were secreted as a result of contact, we derived CFS from 303 pairwise co-cultures and prepared media conditioned with 10% co-culture supernatant. We grew 304 four representative isolates of all four subgroups in media with and without co-culture CFS and 305 measured optical density to monitor growth and used a CV assay for quantification of biofilm 306 formation ( Fig S15). No significant differences in the amount or mode of growth were observed 307 with exposure to co-culture CFS (p > 0.05).  [14,33,34]. Our current study was designed to determine the types of interactions 318 that occur between isolates from different cpn60-defined subgroups of Gardnerella, and to 319 discover if multiple subgroups can be incorporated into biofilms.

321
A contest is a direct, interference competition where the secretion of small molecules (secondary 322 metabolites or toxins) by one organism inhibits the growth of other organisms in an environment 323 [10-12, 19, 21, 35, 36]. Cell-free supernatant (CFS) is the first place to look for any such 324 secreted small molecules that could affect the growth of other bacterial species or strains. 325 Gardnerella isolates have been shown to inhibit the growth of some vaginal lactobacilli in a 326 contact-independent manner [37,38], and both inhibitory and stimulatory effects of Gardnerella 327 CFS on the growth of a range of vaginal microbiota have been documented [39]. These previous 328 reports, however, have involved relatively few isolates and no information regarding the 329 Gardnerella species involved was provided. In the present study, we detected no effect on the 330 amount or mode of growth of Gardnerella isolates when they were exposed to the CFS from 331 other isolates grown in isolation (Figs S1-S14). Since effector molecules are often only secreted 332 when their producers are in contact with other bacterial species [39,40], we also tested whether 333 CFS from co-culture combinations (where competition had been observed in co-culture assays) 334 affected the growth of Gardnerella strains. We found no effect of co-culture CFS on growth, 335 which further supports the conclusion that there is no contest or direct interference competition 336 between Gardnerella subgroups (Fig 4). Similarly, no enhancement of growth was observed, 337 which would have been expected if there was nutritional synergy or cross-feeding among 338 Gardnerella spp. as has been demonstrated for G. vaginalis and Prevotella bivia [41].

340
When Gardnerella isolates from different subgroups were co-cultured, all of them were present 341 in both planktonic and biofilm fractions of each tested community, indicating that no subgroup 342 was completely dominant or excluded over the 48-hour observation period. Competition between 343 subgroups was common, with 70% of the observed interactions classified as competitive. 344 Although intrinsic growth rates differed among the four subgroups ( Fig. 2), subgroups A, B and 345 C all showed a reduced growth rate as the number of competitors increased ( Fig. 2a-c). 346 Interestingly, subgroup D experienced facilitation in co-cultures because its growth rate 347 increased with increasing numbers of competitors (Fig. 2d). Subgroup D also had a negative 348 effect on the growth rates of other subgroups ( Fig. 3a-d). Taken together, these co-culture 349 observations are consistent with a non-interfering, exploitative competition, which is also called 350 scramble competition [11,42]. Scramble competitions result in the dominance of the competitor 351 with the greatest ability to exploit a shared resource (e.g. nutrients), and a general reduction in 352 the overall fitness of all members of a mixed community that share this resource [13,29,42]. 353 One possible explanation for the distinct behaviour of subgroup D is that it has different 354 nutritional requirements than the other subgroups and thus remains unaffected when others 355 compete for the same nutrient resources. It might also represent a "social cheater" [11,43]; an 356 opportunistic member of the community that occupies a distinct niche and benefits from the  [19,29]. In the current study, no enhancement of biofilm biomass was detected using a 383 CV assay when different Gardnerella subgroups were co-cultured (Fig 4), which is consistent 384 with the non-interfering, exploitative competition we observed in the co-cultures. Importantly, 385 our results show that all subgroups of Gardnerella can participate in biofilms, and thus 386 contribute to the formation of this defining feature of bacterial vaginosis, regardless of their 387 individual arsenals of "virulence factors". 388 Overall, our experiments suggest that competition is common in mixed communities of 389 Gardnerella subgroups and that these negative interactions are likely due to niche overlap and