Growth characteristics of pneumococcus vary with the chemical composition of the capsule 1 and with environmental conditions

34 Background : Pneumococcus, a bacterium that typically resides in the nasopharynx, is exposed to 35 a variety of temperature and oxygen levels in the upper respiratory tract and as it invades the lung, 36 tissues, and blood. The response to these variations likely varies by strain and could influence the 37 fitness of a strain and its virulence. We sought to determine the effect of environmental variability 38 on the growth characteristics of pneumococcus and to evaluate correlations between variability in 39 growth characteristics between strains and biological and epidemiological characteristics. 40 41 Methods : We evaluated the effect of temperature and oxygen on the growth of 256 pneumococcal 42 isolates representing 53 serotypes, recovered from healthy carriers and from disease patients. 43 Strains were grown at a range of temperatures anaerobically or in ambient air with and without 44 catalase and were monitored by reading the optical density. Regression models were used to 45 evaluate bacterial and environmental factors associated with characteristics of the growth curves. 46 47 Results: Most isolates grew to the maximal density at the temperature of the nasopharynx (~33C) 48 and under aerobic conditions (with catalase). Maximum density achieved was positively associated 49 with the presence of N-acetylated sugars in the capsule and negatively associated with the presence 50 of uronic acids. Reaching a greater density at an early time point was positively associated with 51 the prevalence of serotypes among healthy carriers in the pre-vaccine period. 52 53 Discussion : Environmental variability affects the growth of pneumococcus, with notable 54 differences between isolates and by serotype. Such variability could be influenced by 55 characteristics of the capsule and might affect virulence and transmissibility. 56

Streptococcus pneumoniae (pneumococcus) is an opportunistic pathogen that resides in the 59 human nasopharynx. The nasopharynx is considered to be the reservoir of transmission between 60 individuals [1]. Pneumococci are diverse, with >90 serotypes (defined by the capsule 61 polysaccharide) and has tremendous genetic variation, resulting from recombination. Serotypes 62 vary in their prevalence among healthy carriers and in the likelihood that they will cause severe 63 disease [2]. As conjugate vaccines against 7, 10, and 13 pneumococcal serotypes have been 64 introduced, the vaccine-targeted serotypes have declined in frequency among healthy carriers and 65 as causes of disease, while serotypes not targeted by the vaccine have increased in importance 66 (serotype replacement). Next-generation conjugate vaccines are under development that target 67 additional serotypes, and it is likely that these vaccines will lead to further serotype replacement. 68 Understanding the factors that influence the fitness of these non-vaccine strains could help to 69 anticipate future patterns of serotypes replacement and could aid in the design of more optimal 70 vaccines.

72
The success of pneumococcus in the nasopharynx and the likelihood that it causes disease is likely 73 driven, in part, by how it responds to variations in its local environment. In different anatomical 74 sites within the human host, pneumococci are exposed to variable temperature and oxygen levels. 75 In the nasopharynx, considered its main niche, the average temperature is around 33°C, with some 76 differences between children and adults [3][4][5][6]. The core body temperature, which would be 77 encountered during invasion into tissues, is 37°C. The temperature in the lungs is constantly 78 changing based on the temperature of inhaled air but is generally lower than 37°C [7]. During 79 infection by pneumococci or during viral co-infection (such as influenza or RSV), both external 80 and internal temperature increases [8][9][10][11][12]. Oxygen levels also vary within the host. In the 81 . CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/416040 doi: bioRxiv preprint first posted online Sep. 13, 2018; nasopharynx, bacteria on top of the mucus layer are exposed to almost ambient air (20% O2). 82 Pneumococci in biofilms in the nasopharynx encounter lower levels of oxygen [13]. Entering the 83 lower respiratory tract or the middle ear, pneumococci are exposed to micro-aerophilic conditions 84 and to almost anaerobic conditions when present in blood and the cerebrospinal fluid (CSF) [14-85 17]. Likewise, mucus production during infection (i.e., due to influenza or RSV) can block the air 86 passage and form micro-aerophilic (around 5% O2) or even anaerobic microenvironments [15,16]. growth characteristics and prevalence of serotypes in the nasopharynx of healthy children [18,19].

92
However, the effect of variation in temperature and oxygen on the growth of different strains and 93 serotypes has not been systematically explored.

95
The aim of this study was to investigate how environmental variability in temperature and oxygen 96 influences the growth characteristics of pneumococci and how the responses to these variations 97 correlates with the biological characteristics of the strains and the observed epidemiology of the 98 serotypes. Using a diverse set of clinical and nasopharyngeal isolates, as well as capsule-switch 99 and capsule-knockout variants generated in the lab, we quantified how the growth characteristics 100 of pneumococci in vitro vary under a range of temperatures and in aerobic and anaerobic 101 conditions. Using statistical models, we estimated the variation in these growth characteristics 102 associated with serotype (after adjusting for isolate-specific effects that could reflect culture 103 . CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/416040 doi: bioRxiv preprint first posted online Sep. 13, 2018; history or other characteristics) and evaluated the relationship between the serotype growth 104 characteristics and relevant serotype-specific epidemiological and biological variables.

108
In total, we performed more than 4,900 growth curves on 256 different pneumococcal strains, 109 representing 53 different serotypes (Figure 1) The maximum density of the strains was similar at 30-35°C, and lower densities were 118 achieved at 37-39°C ( Figure 2A); maximum density was greater in aerobic conditions (with 119 catalase) than in anaerobic conditions. The growth rates were fastest at 35-37°C, with slower 120 growth at lower or higher temperatures ( Figure 2B); growth was faster in aerobic conditions (with 121 catalase) than anaerobic conditions. The length of the lag phase increased with temperature 122 ( Figure 2C).

124
Variation in growth characteristics associated with serotype 125 . CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/416040 doi: bioRxiv preprint first posted online Sep. 13, 2018; We quantified variation in maximum density achieved, growth rate, and length of lag phase 126 associated with serotype. A number of serotypes differed from the reference (serotype 14) in 127 maximum density achieved, length of the lag phase, and the density at an early time point ( Figure   128 3). We therefore evaluated characteristics between these serotype-specific averages and 129 characteristics of the capsules. The presence of uronic acid (GlcA/GalA) in the capsular 130 polysaccharide was associated with lower densities of growth and longer lag phases (P=0.001, 131 Figure 3). The presence of N-acetylated sugar in the capsule was associated with higher density 132 (P=0.002), but not with growth rate or length of the lag phase (Figure 4). The strength of these 133 effects did not differ appreciably between aerobic and anaerobic conditions. 134 We next sought to determine whether this variability in growth characteristics was due to 135 serotype or due to other genetic variability. Some of the serotypes in our collection were 136 represented by multiple genetic lineages (MLST types). There did appear to be variation in growth 137 characteristics associated with serotype that was similar across multiple isolates and MLST 138 lineages (Figures S1, S2). However, there was not enough diversity in our sample to do a formal 139 analysis. We also evaluated the growth characteristics of capsule-knockout strains as well as 140 several capsule-switch variants. While the results were ambiguous, they suggested that there was 141 an effect of capsule production on maximum density, and this effect was more pronounced during 142 anaerobic growth compared with aerobic growth with catalase ( Figure S3-S6).

144
Effect of oxygen on growth varies by serotype 145 We considered whether the effect of oxygen on growth patterns varied by serotype ( Figure 5). 146 Overall, isolates producing certain capsules (e.g., 2, 4, 13, 35B) grew to a higher density under 147 aerobic conditions (with catalase), while isolates producing other capsules (e.g., 6B, 8, 9N, 12F) 148 . CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/416040 doi: bioRxiv preprint first posted online Sep. 13, 2018; did not show a difference between aerobic and anaerobic growth ( Figure 5). This pattern was 149 similar for maximum growth rate. For lag phase, there was little difference among the serotypes 150 in how they responded to the presence of oxygen, with only the serotype 6C and 13 isolates 151 exhibiting a longer lag phase in aerobic conditions.

153
Relationship of growth characteristics to serotype-specific characteristics 154 Finally, we evaluated associations between the characteristics of the growth curves and serotype-155 specific epidemiological characteristics. There was a correlation between the density of a serotype 156 at an early time point and the prevalence of the serotype among healthy carriers during the pre-157 vaccination period in several settings (Figure 6). The association was marginally stronger when 158 looking at aerobic growth compared to anaerobic growth. There was no association between 159 prevalence in carriage and maximum density achieved, length of the lag phase, or maximum 160 growth rate. There was also no notable association between these growth characteristics and the 161 invasiveness of the serotypes or the case fatality ratio.

163
Discussion 164 We provide novel information about how the response to environmental conditions can differ 165 between pneumococcal isolates. Testing a diverse set of strains, pneumococci grew to the highest 166 density under conditions that mimic the normal environment of the nasopharynx in terms of 167 temperature and oxygen level. Important differences in these patterns were observed between 168 isolates. While, some of this variability could be due to the culturing history of the individual 169 isolates, patterns of variation were apparent by serotype. The presence of specific polysaccharide 170 components in the capsule was associated with higher or lower density. And the density of growth 171 . CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/416040 doi: bioRxiv preprint first posted online Sep. 13, 2018; at an early time point was correlated with higher prevalence of those serotypes among healthy 172 carriers.

174
Serotype is a major determinant of pneumococcal epidemiology and biology. We identified a novel 175 association between the capsule composition and the growth phenotype of the isolates, with 176 serotypes containing uronic acids in the capsule growing to a lower density and serotypes with N-177 acetylated sugars growing to a higher density. Further work is needed to understand the mechanism 178 underlying these patterns, but the reliance on metabolic pathways shared between central 179 metabolism and capsule production could influence the growth phenotypes. This could result from 180 metabolic fluxes that restrict growth or through feedback loops that lead to increased metabolic 181 activity. When comparing growth in aerobic conditions (with catalase) with growth in anaerobic 182 conditions, the benefit of oxygen varied by serotype (Figure 4, 5). Serotypes 2, 4, 13, 23B, 35B 183 and 38 grew better with additional catalase than in anaerobic conditions, whereas serotypes 6B, 8, 184 9N, 12F, 22F and 42 grew similarly in both environments. These patterns did not correlate with 185 capsule characteristics. We tested a relatively small number of isolates per serotype, so it is 186 possible that non-capsular genetic variations or differences in the culture history of the isolates 187 could influence the observed responses to oxygen.

189
As the nasopharynx is the normal habitat of pneumococcus, we had hypothesized that 190 pneumococci would grow optimally at temperatures in the low-30C range, similar to the 191 temperature of the nasopharynx. Indeed, temperature played an important role in terms of the 192 maximal density achieved and how quickly the isolates started growing (lag phase). Isolates grew 193 to the highest density and had the shortest lag at temperatures resembling those of the nasopharynx;

194
. CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/416040 doi: bioRxiv preprint first posted online Sep. 13, 2018; these patterns were particularly pronounced for those isolates obtained from the nasopharynx.  CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/416040 doi: bioRxiv preprint first posted online Sep. 13, 2018; Some of the differences observed in growth phenotypes between carriage and disease isolates 217 could reflect opaque/transparent phase variation [25]. Opaque variants are generally isolated from 218 IPD and have increased capsule production and decreased production of certain surface proteins.

219
Phenotypically, the presence of oxygen accentuates differences in capsule production between 220 opaque and transparent variants [24]. This effect could be mediated via the pathways involved in 221 converting pyruvate to acetyl-CoA, an important biochemical precursor for capsule production for 222 many serotypes [26]. Variations in the use of this pathway between serotypes or the efficiency of 223 this pathway between lineages could influence some of the patterns that were observed in the 224 growth curves.

226
This study had certain limitations. For the growth curves, we used BHI broth which is an artificial 227 growth medium that differs in nutrient composition from the host. We evaluated several minimal 228 media but found that growth was generally poor, making comparisons between strains difficult.

229
While we tested a large number of strains representing many serotypes, some serotypes were only 230 represented by a single isolate (i.e., 11B, 12F, 13). This could limit the generalizability of serotype-231 specific findings in these instances, making it difficult to make inferences about whether variability 232 was due to serotype, site of isolation, or lineage effects. The strains used in this study were largely 233 a convenience sample from clinical studies. The genetic diversity of pneumococcus makes it 234 difficult to draw conclusions about the cause of differences between strains. The growth curves 235 with the capsule-knockout strains and capsule-switch variants suggests that the capsule itself could 236 influence these phenotypes. We did not perform any gene expression studies, which could be 237 highly influenced by environmental conditions [27]. Further work could explore the genetic basis 238 (both capsular and non-capsular factors) for the differences in growth phenotypes between strains.

239
. CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/416040 doi: bioRxiv preprint first posted online Sep. 13, 2018; In conclusion, we demonstrate that the growth characteristics of pneumococcus are influenced by 241 environmental variations, that the effect of these variations depend on strains, and that the optimal   CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. Murrieta, CA). Strains were routinely grown at 37°C and 5% CO2 overnight on tryptic soy agar 266 plates supplemented with 5% sheep blood (TSAII) (Thermo Fisher Scientific). Growth in broth CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/416040 doi: bioRxiv preprint first posted online Sep. 13, 2018; Each growth curve was blanked by subtracting the OD600 reading at t=0 for that well. In 284 instances where the t=30 minutes measurement was lower than the t=0 measurement due to 285 measurement error at the first time point, the OD600 at t=30m was subtracted instead. 286 We extracted three characteristics from each of the growth curves: maximum density

306
. CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/416040 doi: bioRxiv preprint first posted online Sep. 13, 2018;307 Linkage to serotype-specific characteristics 308 We sought to evaluate the link between serotype-specific growth curve characteristics and 309 previously published serotype-specific characteristics (capsule structure, disease severity, 310 invasiveness, pre-vaccine carriage prevalence) [37][38][39][40]. We first fit a linear mixed effects model 311 as described above that had dummy variables for serotype (reference: serotype 14) and controlled 312 for temperature (categorical), presence of oxygen, site of isolation, and interactions between site 313 of isolation and presence of oxygen and temperature and presence of oxygen. We then extracted 314 the serotype-specific regression coefficients, which represent variation in the growth 315 characteristics compared to serotype 14. These coefficients were used in second stage analyses to 316 assess correlations with serotype-specific biological and epidemiological variables.

317
Associations with the presence of uronic acids or N-acetylated sugars were assessed using 318 a Monte Carlo approach, where the mean maximum density was compared in the serotypes with 319 or without the sugar components. The sugar component labels were randomly scrambled 999 320 times, and the difference was calculated for each of these samples. P-values were calculated by 321 comparing the observed difference in means with the distribution obtained from the resampling.

322
The association with carriage prevalence was assessed with Poisson regression. In a series of linear 323 regressions, we evaluated the association case-fatality rate, complexity of the capsular 324 polysaccharide (carbons/polysaccharide repeat) and invasiveness (log transformed).

326
Data availability 327 . CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/416040 doi: bioRxiv preprint first posted online Sep. 13, 2018; The raw data and an R Markdown file are available in a github repository 328 (https://github.com/weinbergerlab/GrowthVariation) and can be used to fully re-create the 329 analyses presented here. Hungary.

337
. CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. . CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/416040 doi: bioRxiv preprint first posted online Sep. 13, 2018;  CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.

Figure legends
The copyright holder for this preprint . http://dx.doi.org/10.1101/416040 doi: bioRxiv preprint first posted online Sep. 13, 2018; a lower density on average, while serotypes with N-acetylated sugars in the capsule grew to a 469 higher density. P-values are calculated with Monte Carlo resampling. CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/416040 doi: bioRxiv preprint first posted online Sep. 13, 2018; random effects are effectively an average residual for the isolate, after adjusting for    CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/416040 doi: bioRxiv preprint first posted online Sep. 13, 2018; . CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/416040 doi: bioRxiv preprint first posted online Sep. 13, 2018; 507 . CC-BY 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.