Scott T. Kelley, [email protected] Classification: Major- · 169 (n=8/group). The mice were divided...

1

Exposure to a Healthy Gut Microbiome Protects Against Reproductive and Metabolic 1 Dysregulation in a PCOS Mouse Model 2 3 Pedro J. Torres

1, Bryan S. Ho

2, Pablo Arroyo

2, Lillian Sau

2, Annie Chen

2, Scott T. Kelley

1, 4

Varykina G. Thackray2, 3

5 1Department of Biology, San Diego State University, San Diego, CA, USA 6

2Department of Obstetrics, Gynecology and Reproductive Sciences, University of California, 7

San Diego, La Jolla, CA, USA 92093 8 3Corresponding author: Varykina G. Thackray, Department of Obstetrics, Gynecology and 9

Reproductive Sciences, University of California, San Diego, La Jolla, CA, USA 92093; 619-200-10 4156; [email protected] 11 12 Pedro J. Torres, [email protected] 13 Bryan S. Ho, [email protected] 14 Pablo Arroyo, [email protected] 15 Lillian Sau, [email protected] 16 Annie Chen, [email protected] 17 Scott T. Kelley, [email protected] 18 19 Classification: Major-Biological Sciences, Minor-Physiology 20 21 Keywords: polycystic ovary syndrome, hyperandrogenism, gut microbiome 22 23 Competing interests: The authors declare no conflict of interest. 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46

not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 19, 2018. ; https://doi.org/10.1101/472688doi: bioRxiv preprint

mailto:[email protected]

https://doi.org/10.1101/472688

2

Abstract 47 48 Polycystic ovary syndrome (PCOS) is a common endocrine disorder affecting approximately 49 10% of reproductive-aged women worldwide. Diagnosis requires two of the following: 50 hyperandrogenism, oligo/anovulation and polycystic ovaries. In addition to reproductive 51 dysfunction, many women with PCOS display metabolic abnormalities associated with 52 hyperandrogenism. Recent studies have reported that the gut microbiome is altered in women 53 with PCOS and rodent models of the disorder. However, it is unknown whether the gut 54 microbiome plays a causal role in the development and pathology of PCOS. Given its potential 55 role, we hypothesized that exposure to a healthy gut microbiome would protect against 56 development of PCOS. A co-housing study was performed using a letrozole-induced PCOS 57 mouse model that recapitulates many reproductive and metabolic characteristics of PCOS. Since 58 mice are coprophagic, co-housing results in repeated, non-invasive inoculation of gut microbes 59 in co-housed mice via the fecal-oral route. In contrast to letrozole-treated mice housed together, 60 letrozole-treated mice co-housed with placebo mice showed significant improvement in both 61 reproductive and metabolic PCOS phenotypes. Using 16S rRNA gene sequencing, we observed 62 that the gut microbial composition of letrozole-treated mice co-housed with placebo mice 63 differed from letrozole mice housed together. In addition, our analyses identified several 64 bacterial taxa including Coprobacillus, Dorea and Adlercreutzia associated with the improved 65 PCOS phenotype in letrozole-treated mice co-housed with placebo mice. These results indicate 66 that disruption of the gut microbiome may play a causal role in PCOS and that manipulation of 67 the gut microbiome may be a potential treatment option for PCOS. 68 69 Significance: 70 71 PCOS is a common cause of female infertility and ~80% of women with PCOS have metabolic 72 dysregulation that predisposes them to type 2 diabetes and cardiovascular disease. Since 73 treatment options for the metabolic symptoms of PCOS are limited, there is a need to develop 74 novel therapeutic options. The gut microbiome has emerged as an important player in human 75 health and has been shown to play a causal role in obesity. In this study, we found that exposure 76 to a healthy gut microbiome through co-housing protected against the development of 77 reproductive and metabolic dysregulation in a PCOS mouse model. These results suggest that 78 manipulation of the gut microbiome may be a potential treatment option for women with PCOS. 79 80 Introduction: 81 82 Polycystic ovary syndrome (PCOS) is a common endocrine disorder affecting approximately 83 10% of women worldwide (1). Diagnosis of PCOS, using the Rotterdam Consensus criteria 84 (2003), requires two of the following: hyperandrogenism, oligo- or amenorrhea and polycystic 85 ovaries. PCOS is the leading cause of anovulatory infertility in women, and women with PCOS 86 also have an increased likelihood of miscarriage and pregnancy complications. Though it is often 87 perceived as a reproductive disorder, PCOS is also a metabolic disorder. Women with PCOS 88 have an increased risk of developing obesity, type 2 diabetes, hypertension, and non-alcoholic 89 fatty liver disease (2-4). PCOS-related metabolic dysfunction is associated with 90 hyperandrogenism and occurs irrespective of body mass index (5, 6). While studies indicate that 91 androgen excess is an important contributor to metabolic dysregulation in women with PCOS, 92


https://doi.org/10.1101/472688

3

the mechanisms that lead to obesity and insulin resistance in PCOS are not well understood. 93 Although genetic and environmental factors undoubtedly influence the development and 94 pathology of PCOS (7-10), it is worth exploring whether gut microbes contribute to this disorder. 95 96 Studies over the past decade have shown that the gastrointestinal tract harbors a complex 97 microbial ecosystem (the gut microbiome) that is important for human health and disease (11, 98 12). Gut microbes offer many benefits to the host including protection against pathogens as well 99 as regulation of host immunity and the integrity of the intestinal barrier (13-15). The gut 100 microbiome is also involved in the production of short-chain fatty acids via fermentation of 101 dietary fibers, production of essential vitamins such as folic acid and B12 and modification of 102 bile acids, neurotransmitters and hormones (16, 17). Studies have also shown that changes in the 103 gut microbiome are associated with metabolic disorders such as obesity and type 2 diabetes (18, 104 19). Moreover, studies have reported that transplantation of stool from obese donors into germ-105 free mice results in an obese phenotype (20), suggesting that the gut microbiome may play a 106 causative role in metabolic dysregulation. These transplantation studies were complemented with 107 co-housing studies that took advantage of the fact that, since mice are coprophagic, co-housing 108 provides a means for repeated, non-invasive microbial inoculation. Co-housing germ-free mice 109 transplanted with stool from obese donors with germ-free mice transplanted with stool from lean 110 donors was shown to protect the mice transplanted with obese donor stool from developing 111 obesity (20-22). Altogether, these studies suggest that modulation of the gut microbiome may be 112 a potential treatment option for metabolic disorders. 113 114 With regards to PCOS, several recent studies reported that changes in the gut microbiome are 115 associated with PCOS (23-26). These studies detected lower alpha diversity and differences in 116 the relative abundances of specific Bacteroidetes and Firmicutes in women with PCOS compared 117 to control individuals (23-25). In particular, decreases in the relative abundance of Bacteroides 118 and genera from the Ruminococcaceae and S24-7 families were observed in several studies. In 119 addition, changes in the gut microbiome correlated with hyperandrogenism (25), suggesting that 120 testosterone may influence the composition of the gut microbiome in females. In addition to 121 studies in humans, several studies reported a significant association between the gut microbiome 122 and PCOS in rodent models (27, 28). Since the rodent models are diet independent, these studies 123 suggest that the mechanisms that result in an altered gut microbiome in PCOS are distinct from 124 diet-induced effects on the gut microbiome observed in obesity models. 125 126 We previously developed a PCOS mouse model that uses treatment with letrozole, a non-127 steroidal aromatase inhibitor, to increase testosterone levels and decrease estrogen levels by 128 inhibiting the conversion of testosterone to estrogen (29). Letrozole treatment of pubertal female 129 mice results in the recapitulation of many reproductive hallmarks of PCOS including 130 hyperandrogenism, acyclicity, polycystic ovaries, and elevated luteinizing hormone (LH) levels. 131 This model also exhibits metabolic dysregulation similar to the phenotype in women with PCOS 132 including weight gain, abdominal adiposity, increased fasting blood glucose (FBG) and insulin 133 levels, impaired glucose stimulated-insulin secretion, insulin resistance, and dyslipidemia (30). A 134 caveat is that low estrogen levels in this model do not reflect estrogen levels in women with 135 PCOS. Similar to women with PCOS, 16S rRNA gene sequencing showed that letrozole 136 treatment was associated with lower microbial richness of the gut microbiome, a shift in the 137 overall gut microbial composition, and changes in specific Bacteroidetes and Firmicutes. Since a 138


https://doi.org/10.1101/472688

4

recent study examining the effects of non-antibiotic drugs on the gut microbiome found that 139 letrozole did not alter growth of ~40 representative gut bacteria (31), this suggests that the 140 differences in gut microbial composition found in the PCOS mouse model are not due to a direct 141 effect of letrozole. Overall, these studies indicate that the gut microbiome of individuals with 142 PCOS differs significantly from healthy individuals and suggest that a microbial imbalance or 143 “dysbiosis” in the gut may contribute to the pathology of PCOS. 144 145 To begin to address whether the gut microbiome contributes to PCOS pathophysiology and if 146 manipulation of the gut microbiome can be used to treat PCOS, we used a co-housing paradigm 147 to determine whether exposure to a healthy gut microbiome protected against development of 148 PCOS metabolic and reproductive phenotypes. Since mice are coprophagic, gut microbes can be 149 readily transferred from one mouse to another through the fecal-oral route. In this study, pubertal 150 female mice were treated with placebo or letrozole and co-housed two per cage for 5 weeks. The 151 groups consisted of co-housed placebo mice, co-housed letrozole mice or a placebo mouse co-152 housed with a letrozole mouse. Overall, co-housing letrozole with placebo mice resulted in 153 substantial improvement in both PCOS metabolic and reproductive phenotypes compared to 154 letrozole mice co-housed together. Furthermore, 16S rRNA gene sequence analysis demonstrated 155 that co-housing letrozole with placebo mice resulted in changes in the beta diversity of the gut 156 microbiome and highlighted specific bacterial genera that may be candidates for pre- or probiotic 157 therapies. Our findings support the idea that there may be a causal link between the gut 158 microbiome and PCOS and that restoration/manipulation of the gut microbiome may be a 159 potential treatment option for PCOS. 160 161 Results and Discussion 162 163 Co-housing letrozole mice with placebo mice resulted in less weight gain and abdominal 164 adiposity 165 To investigate whether exposure to a healthy gut microbiome can protect against the 166 development of a PCOS metabolic and reproductive phenotype, we performed a co-housing 167 study. Female mice were implanted with a placebo or letrozole pellet at 4 weeks of age 168 (n=8/group). The mice were divided into 3 groups and housed two mice per cage resulting in 4 169 groups: co-housed placebo mice (P/P), co-housed letrozole mice (L/L) and co-housed placebo 170 with letrozole mice (P/L and L/P, respectively) (Fig. 1A). As shown in Fig. 1B, weight was 171 measured each week during the 5 weeks of treatment. Similarly to previously published studies 172 (27, 30), 2 weeks of letrozole treatment resulted in increased weight compared to placebo 173 treatment that was maintained for the rest of the study (Fig. 1B). Five weeks of letrozole 174 treatment also increased abdominal adiposity compared to placebo treatment (Fig. 1C). 175 Interestingly, P/L mice had similar weight gain and abdominal adiposity compared to P/P mice. 176 In contrast, L/P mice gained less weight and exhibited a trend towards less abdominal adiposity 177 compared to L/L mice (Fig. 1B-C). It is noteworthy that our previous study showed that letrozole 178 treatment induced weight gain and abdominal adiposity in female mice without a change in food 179 intake or total energy expenditure (30). Since studies have indicated that obesity may be 180 influenced by an increased capacity of the gut microbiome to harvest energy from dietary fiber 181 (21), it would be informative to test whether this occurs in the letrozole-treated gut microbiome 182 and whether exposure to a healthy gut microbiome ameliorated this effect. 183 184


https://doi.org/10.1101/472688

5

Co-housing letrozole mice with placebo mice resulted in reduced fasting blood glucose and 185 insulin levels and insulin sensitivity 186 As reported in previous studies (27, 29, 30), five weeks of letrozole treatment resulted in 187 increased FBG and insulin levels and insulin resistance (Fig. 1D-F). P/L mice had similar serum 188 glucose and insulin levels and insulin sensitivity to that of the P/P mice while L/P mice had 189 reduced FBG and insulin levels as well as less insulin resistance compared to L/L mice (Fig. 1D-190 F). Interestingly, the protective effect of co-housing letrozole mice with placebo mice on weight 191 gain only manifested after several weeks of treatment. While our results demonstrated that co-192 housing letrozole mice with placebo mice resulted in protection from changes in FBG and insulin 193 levels and insulin resistance by the end of the study, future studies will be required to ascertain 194 how much time co-housing takes to exert a protective effect on these metabolic factors and what 195 triggers these changes. 196 197 Co-housing letrozole mice with placebo mice resulted in estrous cyclicity 198 In addition to characterizing the effect of co-housing on the PCOS metabolic phenotype, we also 199 assessed the effect on the reproductive axis. As previously published (27, 29), letrozole treatment 200 resulted in hallmarks of PCOS including elevated testosterone and LH levels and acyclicity in 201 the L/L mice (Fig. 2A-C). P/L mice did not have changes in testosterone, LH or estrous cyclicity 202 compared to P/P mice (Fig. 2A-C). On the other hand, L/P mice had decreased testosterone and 203 LH levels compared to L/L mice (Fig. 2A-B). In addition, the L/P mice displayed a full estrous 204 cycle or changes in the morphology of vaginal epithelial cells compared to the constant diestrus 205 exhibited by L/L mice (Fig. 2C). 206 207 Co-housing letrozole mice with placebo mice protected ovarian function 208 Consistent with previous reports (27, 29), the ovaries of letrozole-treated mice (L/L) lacked 209 corpora lutea and displayed cystic follicles and hemorrhagic cysts while the ovaries of P/L mice 210 had a similar morphology to that of P/P mice (Fig. 3A). Interestingly, the ovaries of L/P mice 211 lacked cystic follicles and hemorrhagic cysts and contained corpora lutea (Fig. 3A). The 212 presence of estrous cycles and corpora lutea in the L/P mice suggests that the mice were able to 213 ovulate. Similar to previous reports (29), L/L mice showed a significant increase in both ovarian 214 weight and ovarian mRNA expression levels of follicle-stimulating hormone receptor (Fshr), 215 cytochrome P450 17A1 (Cyp17), and aromatase (Cyp19) compared to P/P mice (Fig. 3B-E). The 216 ovarian weight and mRNA expression levels in P/L mice mirrored those of the P/P mice (Fig. 217 3B-E). Compared to L/L mice, L/P mice showed a significant decrease in ovarian weight and 218 mRNA expression levels of Cyp17 while Fshr and Cyp19 mRNA expression levels were 219 comparable to L/L mice (Fig. 3B-E). Since Cyp17 gene expression is induced by both androgen 220 and insulin (32, 33), it is unsurprising that Cyp17 levels were normalized in L/P mice that had 221 reduced circulating levels of testosterone and insulin. On the other hand, it is not clear why Fshr 222 mRNA levels were increased in both L/L and L/P mice. With regards to aromatase expression, it 223 is notable that letrozole treatment resulted in an increase in mRNA levels that was not altered by 224 co-housing letrozole with placebo mice. This data indicates that suppression of the aromatase 225 enzyme with letrozole treatment in both L/L and L/P mice resulted in a compensatory increase in 226 Cyp19 mRNA and provides indirect evidence that the effect of co-housing did not occur because 227 of decreased letrozole activity in the L/P mice. 228 229 Gut microbial richness did not correlate with an improved PCOS phenotype 230


https://doi.org/10.1101/472688

6

The overall composition of the gut microbiome from samples collected before placebo and 231 letrozole treatment (time 0) was compared amongst the four co-housing groups (P/P, L/L, P/L 232 and L/P). No significant difference in alpha or beta diversity was observed among groups, 233 indicating that the gut microbiomes were similar prior to treatment (Supplemental Fig. 1 A-B). 234 Linear regression was used to examine the relationship between alpha diversity of the gut 235 microbiome (Faith’s PD) and time. There was a strong positive relationship between alpha 236 diversity and time in P/P mice (r = 0.23) but not L/L mice (r = 0.05) (Fig. 4A-B). To account for 237 the repeated measures in this longitudinal study, we also used a linear mixed-effect model to 238 examine the association between microbial diversity and time. This analysis confirmed that there 239 was a significant effect of time on alpha diversity in P/P mice (p = 0.003) but not L/L mice (p = 240 0.2) (Fig. 4A-B). These results agree with previous studies that reported lower alpha diversity of 241 the gut microbiome in women with PCOS and rodent PCOS models compared to controls (24, 242 25, 27, 28). We then investigated whether changes in alpha diversity correlated with improved 243 metabolic and reproductive phenotypes in L/P mice. In contrast to P/P mice, we did not observe 244 a significant effect of time on alpha diversity using linear regression or the linear mixed-effect 245 model in P/L (r = 0.009; p = 1) or L/P (r = 0.08; p = 0.71) mice (Fig. 4C-D). These results 246 indicate that the physiological differences between L/L and L/P mice were probably not due to 247 changes in alpha diversity per se but may reflect changes in specific gut microbes. 248 249 Composition of the gut microbiome was altered by co-housing 250 In addition to investigating changes in alpha diversity, we used weighted UniFrac distances to 251 compare the similarity of gut microbial composition (beta diversity) amongst the different 252 groups. While visualization of the UniFrac distances via Principal Coordinate Analysis (PCoA) 253 did not result in distinct clustering, a PERMANOVA test (ADONIS) detected a significant effect 254 of co-housing treatment on the microbial community structure (p=0.001) (Fig. 5A). This trend 255 was also observed using unweighted UniFrac (data not shown). Canonical Analysis of Principal 256 Coordinates (CAP) analysis was then used to analyze the microbial composition in response to 257 an a priori defined experimental variable (co-housing treatment). PERMANOVA demonstrated 258 a strong relationship between co-housing treatment and the overall composition of the gut 259 microbiome (p=0.001) (Fig. 5B), suggesting that co-housing resulted in a distinct gut microbial 260 community in L/P mice compared to L/L mice. To understand when the gut microbiome 261 diverged, we then compared the fecal samples from the different co-housing treatment groups at 262 each time point (Fig. 5C-G). By week 2, we observed a significant separation of the bacterial 263 communities in the co-housing treatment groups (p = 0.004) (Fig. 5D). This separation occurred 264 at the same time that we observed a protective effect of co-housing on weight gain in L/P mice 265 (Fig. 1B), suggesting that changes in the abundance of specific bacteria may result in protection 266 from developing the PCOS phenotype. It is notable that the CAP analysis indicated that co-267 housing also resulted in changes in the gut bacterial community of the P/L mice compared to P/P 268 mice. However, since these changes were not sufficient to alter the metabolic and reproductive 269 phenotypes of the host, these results suggest that the healthy gut microbiome was resistant to any 270 pathological influence from the feces of the letrozole mice. 271 272 Differentially abundant genera are associated with co-housed letrozole mice 273 Differential abundance of gut bacteria between placebo- and letrozole-treated mice was 274 determined using DESeq2. This approach used a negative binomial regression for modeling 275 count variables and is commonly used for overdispersed data, which is typical of microbiome 276


https://doi.org/10.1101/472688

7

data (31). DESeq2 identified five bacterial genera that were of higher relative abundance and 277 four bacterial genera that were of lower relative abundance in placebo compared to letrozole 278 mice (Fig. 6A). The gram-positive bacteria included Coprobacillus, Candidatus Arthromitus, 279 Roseburia, Dorea, Lactobacillus, Adlercreutzia, and Akkermansia while the gram-negative 280 bacteria included Christensenella and Turicibacter. 281 282 From the DESeq2 analysis, we then identified three bacterial genera that had shifts in relative 283 abundance corresponding with protection from developing the PCOS phenotype (Fig. 6B-D). For 284 instance, the relative abundance of Coprobacillus was higher in both P/P and L/P mice compared 285 to L/L mice. Interestingly, all three bacteria have been linked with host metabolism. 286 Coprobacillus was reported to be enriched in healthy individuals compared to obese individuals 287 and was proposed as a novel probiotic due to its association with a healthy gut microbiome (34, 288 35). Dorea was found to be more abundant in lean than obese individuals (36)and Dorea 289 longicatena was included as a component of the RePOOPulate stool substitute being developed 290 as a treatment for Clostridium difficile (37). Moreover, a higher abundance of Adlercreutzia was 291 previously correlated with obesity (38). Adlercreutzia is also known to convert soy isoflavonoids 292 to non-steroidal estrogens (equol) and was overrepresented in the guts of individuals who eat a 293 soy diet (39). 294 295 In summary, our study demonstrated that co-housing letrozole mice with placebo mice resulted 296 in protection from developing PCOS metabolic and reproductive phenotypes. In particular, 297 exposure to a healthy gut microbiome resulted in letrozole-treated mice with body weight, FBG 298 and insulin levels, and insulin resistance similar to placebo mice (Fig. 1). Surprisingly, co-299 housing letrozole-treated mice with healthy mice also resulted in letrozole-treated mice with 300 testosterone and LH levels, estrous cycling and ovarian morphology similar to placebo mice 301 (Fig. 2-3). This improvement in the PCOS phenotype correlated with changes in the relative 302 abundance of specific gut bacteria such as Coprobacillus, Dorea and Adlercreutzia, indicating 303 that manipulation of the gut microbiome towards a dysbiotic (letrozole treatment) or healthy 304 state (co-housing) may influence the degree of pathology. Additional studies should characterize 305 the effects of co-housing on the composition (strain level) and function of the gut microbiome in 306 the letrozole-induced PCOS mouse model using metagenomics and metabolomics. Since these 307 results imply that manipulating the composition of the gut microbiome may be a potential 308 treatment option for women with PCOS, future studies should also investigate whether 309 supplementation with prebiotics or novel probiotics such as Coprobacillus and Dorea can protect 310 against development of PCOS. 311 312 Materials and Methods 313 314 PCOS Mouse Model 315 C57BL/6NHsd female mice from Envigo were housed in a vivarium under specific pathogen-316 free conditions with an automatic 12h:12h light/dark cycle (light period: 06.00-18.00) and ad 317 libitum access to water and food (Teklad S-2335 Mouse Breeder Irradiated Diet, Envigo). To 318 establish the pubertal PCOS model, 4 week-old female mice were implanted subcutaneously 319 with a placebo or 3 mg letrozole pellet (3 mm diameter; 50 µg/day; Innovative Research of 320 America) for 5 weeks. For the co-housing paradigm, mice were housed two per cage: two 321 placebo mice, two letrozole mice or one placebo and one letrozole mouse. All animal procedures 322


https://doi.org/10.1101/472688

8

in this study were approved by the University of California, San Diego Institutional Animal Care 323 and Use Committee (protocol number S14011). 324 325 Analysis of Reproductive Phenotype 326 Mice were weighed weekly. Estrous cycle stage was determined from the predominant cell type 327 in vaginal epithelial smears obtained during weeks 4-5 of treatment as previously described (29). 328 At the end of the experiment, ovaries were dissected, weighed, fixed in 4% paraformaldehyde, 329 paraffin-embedded, sectioned at 10 μm and stained with hematoxylin and eosin (Zyagen). Serum 330 testosterone and LH levels were measured using a mouse ELISA (range 10-800 ng/dL) and a 331 radioimmunoassay (range 0.04-75 ng/mL), by the University of Virginia Center for Research in 332 Reproduction Ligand Assay and Analysis Core Facility. 333 334 Analysis of Metabolic Phenotype 335 After 5 weeks of treatment, mice were fasted for 5 hours and blood from the tail vein was 336 collected to measure fasting insulin levels. Blood glucose was measured using a handheld 337 glucometer (One Touch UltraMini, LifeScan, Inc) and an intraperitoneal (IP) insulin tolerance 338 test (ITT) was performed. Tail vein blood glucose was measured just before (time 0) an IP 339 injection of insulin (0.75 U/kg in sterile saline; Humulin R U-100, Eli Lilly) was given and at 15, 340 30, 45, 60, 90, and 120 minutes post injection. At the end of the experiment, the mice were 341 anesthetized with isoflurane, blood was collected from the posterior vena cava, and parametrial 342 fat pads were dissected and weighed. Serum insulin was measured using a mouse ELISA 343 (ALPO) by the University of California, Davis Mouse Metabolic Phenotyping Center. 344 Differences among groups were analyzed by one-way ANOVA or two-way repeated measures 345 ANOVA followed by post-hoc comparisons of individual time points. 346 347 Quantitative Real-Time Polymerase Chain Reaction of Ovarian Genes 348 Total RNA was isolated from ovaries using a RNeasy Mini Plus kit (Qiagen), which also 349 removes genomic DNA. Complementary DNA (cDNA) was made by reverse transcription of 350 total RNA using an iScript cDNA synthesis kit (Bio-Rad Laboratories). cDNA products were 351 detected using SYBR Green Supermix (Bio-Rad Laboratories) on a Bio-Rad CFX Connect 352 quantitative real-time polymerase chain reaction system (Bio-Rad Laboratories) using previously 353 described primers (29). Data were analyzed by the 2

2ΔΔCt method (40) by normalizing the gene of 354

interest to GAPDH. Data were represented as mean fold change compared with placebo ± the 355 standard error of the mean. 356 357 Fecal Sample Collection, DNA Isolation and 16S rRNA Gene Sequencing 358 Fecal samples were collected prior to treatment and once per week for 5 weeks. Fecal samples 359 were frozen immediately after collection and stored at -80°C. Bacterial DNA was extracted from 360 the samples using the DNeasy PowerSoil Kit (Qiagen) and stored at -80°C. The V4 361 hypervariable region of the 16S rRNA gene was PCR amplified with primers 515F and 806R 362 (41). The reverse primers contained unique 12 base pair Golay barcodes that were incorporated 363 into the PCR amplicons (41). Amplicon sequence libraries were prepared at The Scripps 364 Research Institute Next Generation Sequencing Core Facility where the libraries were sequenced 365 on an Illumina MiSeq as previously described (27). 366

16S rRNA Gene Sequence Analysis 367


https://doi.org/10.1101/472688

9

Raw sequences were imported into QIIME 2 (version 2018.4) using the q2-tools-import script 368 and sequences were demultiplexed using the q2-demux emp-single script. This resulted in 7.3 369 million sequences with an average of 36,000 sequences per sample. The 16S rRNA sequences 370 generated in this study were deposited into the European Nucleotide Archive (Study Accession 371 Number PRJEB29583). DADA2 software was used to obtain a set of Observed Sequence 372 Variants (SVs) (42). Based on the quality scores, the forward reads were truncated at position 373 240 using the q2-dada2-denoise script. Taxonomy was assigned using a pre-trained Naïve Bayes 374 classifier (Greengenes 13_8 99% operational taxonomic units) and the q2-feature-classifier 375 plugin (43). In total, 318 SVs were identified from 186 fecal pellets. The resulting SVs were then 376 aligned using MAFFT (44) and a phylogenetic tree was built using FastTree (45). Taxonomic 377 distributions of the samples were calculated using the q2-taxa-barplot script. Alpha and beta 378 diversity metrics were computed using the q2-diversity core-metrics script at a rarefied sampling 379 depth of 1250. Rarefaction resulted in the removal of one sample that had ~400 sequences. The 380 alpha diversity metric, Faith’s Phylogenetic Diversity (PD), was used to measure phylogenetic 381 biodiversity by calculating the total branch lengths on a phylogenetic tree of all members in a 382 community (46). UniFrac was used to compare the similarity (beta diversity) among the 383 microbial communities by calculating the shared PD between pairs of microbial communities 384 (47, 48). 385

Statistical Analysis 386 Statistical calculations were performed in the R Studio statistical package (version 0.99.893) 387 with the phyloseq (version 1.19.1) (49) and vegan package (version 2.5.2). Alpha diversity data 388 was tested for normality via the Shapiro-Wilk test. Variables that were not normally distributed 389 were ranked. Changes in alpha diversity over time were analyzed using simple linear regression 390 and Pearson’s rank correlation on ranked diversity measures. Linear mixed effects analysis of the 391 relationship between alpha diversity and time was done with the lme4 R package (version 392 1.1.18.1). P-values were obtained by likelihood ratio tests of the full model with the effect in 393 question against the model without the effect in question. Principal Coordinate Analysis (PCoA) 394 and CAP plots (50) were constructed using the phyloseq R package. PCoA plots were used to 395 represent the similarity of post-treatment (Weeks 1-5) fecal microbiome samples based on 396 multiple variables in the data set, while CAP was used to visualize the relationship of the fecal 397 microbiome with specific parameters. Permutational multivariate analysis of variance 398 (PERMANOVA) used post-treatment weighted UniFrac distance measures to assess bacterial 399 community compositional differences and its relationship to co-housing treatment group (999 400 permutations “vegan” package). DESeq2 (51) (version 1.14.1) in the microbiomeSeq package 401 (version 0.1 http://www.github.com/umerijaz/microbiomeSeq) was used to identify bacterial 402 genera that were differentially abundant between placebo and letrozole-treated mice. 403

List of Abbreviations 404 fasting blood glucose, FBG; insulin tolerance test, ITT; intraperitoneal, IP; sequence variant, SV; 405 permutational multivariate analysis of variance, PERMANOVA; phylogenetic diversity, PD; 406 luteinizing hormone, LH; canonical analysis of principal coordinates, CAP; principal coordinates 407 analysis PCoA; polycystic ovary syndrome, PCOS; Quantitative Insights Into Microbial 408 Ecology, QIIME 409 410 Acknowledgements: We thank members of the Kelley and Thackray labs for insightful 411 comments and suggestions. Hormone levels were measured by the University of Virginia Center 412


http://www.github.com/umerijaz/microbiomeSeq

https://doi.org/10.1101/472688

10

for Research in Reproduction Ligand Assay and Analysis Core Facility (P50 HD28934) and the 413 University of California, Davis Mouse Metabolic Phenotyping Core (U24 DK092993). This 414 work was funded by the National Institute of Child Health and Human Development through a 415 cooperative agreement as part of the National Centers for Translational Research in 416 Reproduction and Infertility (P50 HD012303). P.J.T. was also funded by the UCSD Microbial 417 Sciences Research Initiative Fellowship and the SDSU ARCS Foundation. L.S. and A.C. were 418 funded by a Doris A. Howell Foundation Research Scholarship for Women’s Health. 419 420 Authors’ contributions: V.G.T. and S.T.K. conceived and designed the study; P.J.T., B.S.H., 421 P.A., L.S., and A.C. collected samples, performed reproductive and metabolic assessments, 422 quantitative RT-PCR; P.J.T. performed DNA extractions and PCR amplifications; P.J.T., V.G.T. 423 and S.T.K. analyzed the data; P.J.T., S.T.K. and V.G.T. wrote the manuscript. 424 425 References 426 427 1. Fauser BC, et al. (2012) Consensus on women's health aspects of polycystic ovary 428

syndrome (PCOS): the Amsterdam ESHRE/ASRM-Sponsored 3rd PCOS Consensus 429 Workshop Group. Fertil Steril 97(1):28-38 e25. 430

2. Azziz R, et al. (2016) Polycystic ovary syndrome. Nat Rev Dis Primers 2:16057. 431 3. Churchill SJ, Wang ET, & Pisarska MD (2015) Metabolic consequences of polycystic 432

ovary syndrome. Minerva Ginecol 67(6):545-555. 433 4. Goodman NF, et al. (2015) American Association of Clinical Endocrinologists, 434

American College of Endocrinology, and Androgen Excess and Pcos Society Disease 435 State Clinical Review: Guide to the Best Practices in the Evaluation and Treatment of 436 Polycystic Ovary Syndrome -Part 2. Endocrine Practice 21(12):1415-1426. 437

5. Barber TM, Wass JAH, McCarthy MI, & Franks S (2007) Metabolic characteristics of 438 women with polycystic ovaries and oligo-amenorrhoea but normal androgen levels: 439 implications for the management of polycystic ovary syndrome. Clinical endocrinology 440 66(4):513-517. 441

6. Moghetti P, et al. (2013) Divergences in Insulin Resistance Between the Different 442 Phenotypes of the Polycystic Ovary Syndrome. The Journal of Clinical Endocrinology & 443 Metabolism 98(4):E628-E637. 444

7. Barber TM, Wass JA, McCarthy MI, & Franks S (2007) Metabolic characteristics of 445 women with polycystic ovaries and oligo-amenorrhoea but normal androgen levels: 446 implications for the management of polycystic ovary syndrome. Clin Endocrinol (Oxf) 447 66(4):513-517. 448

8. Legro RS, Driscoll D, Strauss JF, Fox J, & Dunaif A (1998) Evidence for a genetic basis 449 for hyperandrogenemia in polycystic ovary syndrome. Proc. Natl. Acad. Sci. 14956-450 14960. 451

9. Vink JM, Sadrzadeh S, Lambalk CB, & Boomsma DI (2006) Heritability of polycystic 452 ovary syndrome in a Dutch twin-family study. J Clin Endocrinol Metab 91(6):2100-453 2104. 454

10. Abbott DH & Bacha F (2013) Ontogeny of polycystic ovary syndrome and insulin 455 resistance in utero and early childhood. Fertil Steril 100(1):2-11. 456

11. Clemente JC, Ursell LK, Parfrey LW, & Knight R (2012) The impact of the gut 457 microbiota on human health: an integrative view. Cell 148(6):1258-1270. 458


https://doi.org/10.1101/472688

11

12. Walker AW & Lawley TD (2013) Therapeutic modulation of intestinal dysbiosis. 459 (Translated from eng) Pharmacol Res 69(1):75-86 (in eng). 460

13. Baumler AJ & Sperandio V (2016) Interactions between the microbiota and pathogenic 461 bacteria in the gut. Nature 535(7610):85-93. 462

14. Gensollen T, Iyer SS, Kasper DL, & Blumberg RS (2016) How colonization by 463 microbiota in early life shapes the immune system. Science 352(6285):539-544. 464

15. Natividad JM & Verdu EF (2013) Modulation of intestinal barrier by intestinal 465 microbiota: pathological and therapeutic implications. Pharmacol Res 69(1):42-51. 466

16. den Besten G, et al. (2013) The role of short-chain fatty acids in the interplay between 467 diet, gut microbiota, and host energy metabolism. J Lipid Res 54(9):2325-2340. 468

17. Ridlon JM, Kang DJ, Hylemon PB, & Bajaj JS (2014) Bile acids and the gut microbiome. 469 Curr Opin Gastroenterol 30(3):332-338. 470

18. Ley RE, Turnbaugh PJ, Klein S, & Gordon JI (2006) Microbial ecology: human gut 471 microbes associated with obesity. Nature 444(7122):1022-1023. 472

19. Qin J, et al. (2012) A metagenome-wide association study of gut microbiota in type 2 473 diabetes. Nature 490(7418):55-60. 474

20. Turnbaugh PJ, et al. (2009) A core gut microbiome in obese and lean twins. Nature 475 457(7228):480-484. 476

21. Turnbaugh PJ, et al. (2006) An obesity-associated gut microbiome with increased 477 capacity for energy harvest. Nature 444(7122):1027-1031. 478

22. Ridaura VK, et al. (2013) Gut microbiota from twins discordant for obesity modulate 479 metabolism in mice. Science 341(6150):1241214. 480

23. Liu R, et al. (2017) Dysbiosis of Gut Microbiota Associated with Clinical Parameters in 481 Polycystic Ovary Syndrome. Frontiers in Microbiology 8. 482

24. Lindheim L, et al. (2017) Alterations in Gut Microbiome Composition and Barrier 483 Function Are Associated with Reproductive and Metabolic Defects in Women with 484 Polycystic Ovary Syndrome (PCOS): A Pilot Study. PLoS One 12(1):e0168390. 485

25. Torres PJ, et al. (2018) Gut Microbial Diversity in Women with Polycystic Ovary 486 Syndrome Correlates with Hyperandrogenism. J Clin Endocrinol Metab. 487

26. Insenser M, et al. (2018) Gut Microbiota and the Polycystic Ovary Syndrome: Influence 488 of Sex, Sex Hormones, and Obesity. J Clin Endocrinol Metab 103(7):2552-2562. 489

27. Kelley ST, Skarra DV, Rivera AJ, & Thackray VG (2016) The Gut Microbiome Is 490 Altered in a Letrozole-Induced Mouse Model of Polycystic Ovary Syndrome. PLoS One 491 11(1):e0146509. 492

28. Guo YJ, et al. (2016) Association between Polycystic Ovary Syndrome and Gut 493 Microbiota. Plos One 11(4). 494

29. Kauffman AS, et al. (2015) A novel letrozole model recapitulates both the reproductive 495 and metabolic phenotypes of Polycystic Ovary Syndrome in female mice. Biol Reprod 496 93(3):69. 497

30. Skarra DV, Hernandez-Carretero A, Rivera AJ, Anvar AR, & Thackray VG (2017) 498 Hyperandrogenemia Induced by Letrozole Treatment of Pubertal Female Mice Results in 499 Hyperinsulinemia Prior to Weight Gain and Insulin Resistance. Endocrinology 500 158(9):2988-3003. 501

31. Maier L, et al. (2018) Extensive impact of non-antibiotic drugs on human gut bacteria. 502 Nature 555(7698):623-628. 503


https://doi.org/10.1101/472688

12

32. Kumar A, Magoffin D, Munir I, & Azziz R (2009) Effect of insulin and testosterone on 504 androgen production and transcription of SULT2A1 in the NCI-H295R adrenocortical 505 cell line. Fertil Steril 92(2):793-797. 506

33. Zhang G & Veldhuis JD (2004) Insulin drives transcriptional activity of the CYP17 gene 507 in primary cultures of swine theca cells. Biol Reprod 70(6):1600-1605. 508

34. Hou YP, et al. (2017) Human Gut Microbiota Associated with Obesity in Chinese 509 Children and Adolescents. Biomed Res Int 2017:7585989. 510

35. Tap J, et al. (2009) Towards the human intestinal microbiota phylogenetic core. Environ 511 Microbiol 11(10):2574-2584. 512

36. Aguirre M, Bussolo de Souza C, & Venema K (2016) The Gut Microbiota from Lean and 513 Obese Subjects Contribute Differently to the Fermentation of Arabinogalactan and Inulin. 514 PLoS One 11(7):e0159236. 515

37. Petrof EO, et al. (2013) Stool substitute transplant therapy for the eradication of 516 Clostridium difficile infection: 'RePOOPulating' the gut. Microbiome 1(1):3. 517

38. Yun Y, et al. (2017) Comparative analysis of gut microbiota associated with body mass 518 index in a large Korean cohort. BMC Microbiol 17(1):151. 519

39. Rafii F (2015) The role of colonic bacteria in the metabolism of the natural isoflavone 520 daidzin to equol. Metabolites 5(1):56-73. 521

40. Livak KJ & Schmittgen TD (2001) Analysis of relative gene expression data using real-522 time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods 25(4):402-408. 523

41. Caporaso JG, et al. (2012) Ultra-high-throughput microbial community analysis on the 524 Illumina HiSeq and MiSeq platforms. ISME J 6(8):1621-1624. 525

42. Callahan BJ, et al. (2016) DADA2: High-resolution sample inference from Illumina 526 amplicon data. Nat Methods 13(7):581-583. 527

43. Bokulich NA, et al. (2018) Optimizing taxonomic classification of marker-gene amplicon 528 sequences with QIIME 2's q2-feature-classifier plugin. Microbiome 6(1):90. 529

44. Katoh K & Standley DM (2013) MAFFT multiple sequence alignment software version 530 7: improvements in performance and usability. Mol Biol Evol 30(4):772-780. 531

45. Price MN, Dehal PS, & Arkin AP (2009) FastTree: computing large minimum evolution 532 trees with profiles instead of a distance matrix. Mol Biol Evol 26(7):1641-1650. 533

46. Faith DP (1992) Conservation evaluation and phylogenetic diversity. Biological 534 Conservation 61(1):1-10. 535

47. Lozupone C & Knight R (2005) UniFrac: a new phylogenetic method for comparing 536 microbial communities. Appl Environ Microbiol 71(12):8228-8235. 537

48. Lozupone C, Lladser ME, Knights D, Stombaugh J, & Knight R (2011) UniFrac: an 538 effective distance metric for microbial community comparison. ISME J 5(2):169-172. 539

49. McMurdie PJ & Holmes S (2012) Phyloseq: a bioconductor package for handling and 540 analysis of high-throughput phylogenetic sequence data. Pac Symp Biocomput:235-246. 541

50. Anderson MJ & Willis TJ (2003) Canonical analysis of principal coordinates: A useful 542 method of constrained ordination for ecology. Ecology 84(2):511-525 . 543

51. Love MI, Huber W, & Anders S (2014) Moderated estimation of fold change and 544 dispersion for RNA-seq data with DESeq2. Genome Biol 15(12):550 . 545

546 547 548 549


https://doi.org/10.1101/472688

13

Figure Legends 550 551 Fig 1. Co-housing letrozole mice with placebo mice protected against development of the 552 PCOS metabolic phenotype. Design of co-housing study over the course of a 5-week 553 experiment included 3 groups of 4-week-old female mice housed 2 per cage: co-housed placebo 554 mice (P/P), co-housed letrozole mice (L/L), and co-housed placebo with letrozole mice (P/L and 555 L/P, respectively) (A). Letrozole treatment (L/L) resulted in metabolic dysfunction compared to 556 placebo (P/P) including increased weight, abdominal adiposity, fasting blood glucose (FBG) and 557 insulin levels and insulin resistance (B-F). Compared to L/L mice, L/P mice showed a decrease 558 in body weight, a decrease in abdominal adiposity, a decrease in FBG and insulin levels, and 559 restored insulin sensitivity (B-F). 560 561 Fig. 2. Letrozole mice co-housed with placebo mice did not become hyperandrogenemic or 562 acyclic. The co-housing study included 3 groups of female mice housed 2 per cage: co-housed 563 placebo mice (P/P), co-housed letrozole mice (L/L), and co-housed placebo with letrozole mice 564 (P/L and L/P, respectively). Letrozole treatment (L/L) resulted in increased testosterone and 565 luteinizing hormone (LH) levels compared to placebo (P/P) (A-B). L/P mice displayed a 566 decrease in testosterone and LH (A-B) as well as a restoration of estrous cyclicity compared L/L 567 mice stuck in diestrus (C). Stages of the estrous cycle are indicated as diestrus (D), metestrus 568 (M), estrus (E) and proestrus (P). 569 570 Fig. 3. Co-housing letrozole mice with placebo mice improved the ovarian phenotype. The 571 co-housing study included 3 groups of female mice housed 2 per cage: co-housed placebo mice 572 (P/P), co-housed letrozole mice (L/L), and co-housed placebo with letrozole mice (P/L and L/P, 573 respectively). Letrozole treatment (L/L) resulted in a lack of corpora lutea, cyst-like follicles and 574 hemorrhagic cysts in the ovaries compared to ovaries from placebo mice (P/P) (A). Unlike L/L 575 mice, L/P mice lacked polycystic ovaries and their ovaries contained corpora lutea (CL) which is 576 evidence of ovulation (A). Letrozole (L/L) also resulted in increased ovarian weight and 577 increased mRNA expression of several ovarian genes important in ovarian follicular 578 development and steroidogenesis (B-E). Ovarian weight was lower in L/P mice compared to L/L 579 mice (B). Follicle-stimulating hormone receptor (Fshr) and aromatase (Cyp19) mRNA levels 580 were similar between L/L and L/P mice while cytochrome P450 17A1 (Cyp17) was decreased in 581 L/P mice compared to L/L mice. 582 583 Fig 4. Co-housing letrozole mice with placebo mice did not restore alpha diversity of the 584 gut microbiome. Alpha diversity as approximated by Faith’s phylogenetic diversity (PD) ranked 585 estimate was graphed over time for co-housed placebo mice (P/P), co-housed letrozole mice 586 (L/L), and co-housed placebo with letrozole mice (P/L and L/P, respectively). Results of linear 587 regression model (LM) and P value are in the box inset, while the gray shaded area indicates the 588 95% confidence interval for the line of best fit. P values for the linear mixed effects model 589 (LME) were obtained by the likelihood ratio test of the full model with the effect in question 590 (time) against the model without the effect in question and are in the box inset. 591 592 Fig 5. Co-housing letrozole mice with placebo mice influenced the overall composition of 593 the gut bacterial community over time. Unconstrained principal coordinates analysis (PCoA) 594 of weighted UniFrac distances demonstrated changes in the microbial composition (beta 595


https://doi.org/10.1101/472688

14

diversity) amongst samples collected post-treatment (A). Permutational analysis of variance of 596 the weighted UniFrac distances indicated that co-housing had a strong influence on the gut 597 microbial community (p=0.001). Constrained analysis of principal coordinates (CAP) of 598 weighted UniFrac distances further illustrated the relationship between beta diversity and post-599 treatment with a significant effect of constraining the data based on co-housing treatment group 600 (p=0.001) (B). Samples from the different co-housing groups were then compared at each time 601 point (C-G). Permutational analysis of variance of the weighted UniFrac distances was done for 602 each time point. 603 604 Fig. 6. Specific bacterial genera were associated with improvement of the PCOS phenotype 605 during co-housing. 606 The co-housing study included 3 groups of female mice housed 2 per cage: co-housed placebo 607 mice (P/P), co-housed letrozole mice (L/L), and co-housed placebo with letrozole mice (P/L and 608 L/P, respectively). Results from the DESeq2 differential abundance analysis were expressed as 609 log2 fold change for the comparison of L/L mice and P/P mice (A). Positive log2 fold changes 610 represent bacterial genera increased in P/P mice relative to L/L mice while negative changes 611 represent bacterial general increased in L/L mice relative to P/P mice. * p<0.05, ** p<0.01, *** 612 p<0.001. Of the bacteria identified with the DESeq2 analysis, three bacterial genera had changes 613 in relative abundance associated with co-housing letrozole mice with placebo mice (B-D). 614 615 616 617 618


https://doi.org/10.1101/472688

L/LP/P

P/LL/P

4 9876516

18

20

22

24

26

28

Bod

yW

eigh

t(g)

Age (Weeks)

B

Fig. 1

5

25201510

0

30

Par

amet

rialF

atP

ad/

Bod

yW

eigh

t

C4035

P/P L/L P/L L/P

b

aba

b

40

14012010080

20

160

0

a

b

aa

200180

60

P/P L/L P/L L/P

FBG

(mg/

dL)

D

L/LP/P

P/LL/P

0 1209060453015Time (Minutes)

*

100

80

60

40

20

0

*

%B

asal

Glu

cose

*

insulin tolerance testF

200

1000800600400

0

1200

Insu

lin(p

g/m

L)

E16001400

P/P L/L P/L L/P

a

b

ca

A

a

b

aa

Letrozole:LetrozolePlacebo:Placebo

Co-Housing Study Design

Placebo:Letrozole

P P

P L

L L


https://doi.org/10.1101/472688

40

140120100

80

200

60

P/P L/L P/L L/P

Test

oste

rone

(ng/

mL)

A

a

b

aa

P/P L/L P/L L/P

0.5

2.52.01.51.0

0.0

3.0

4.03.5

LH(n

g/m

L)

B

a

b

ca

Fig. 2

E

D

P

M

Est

rous

Cyc

leS

tage

E

D

P

M

E

D

P

M

E

D

P

M

Time (Day)41 76532

L/P

C

Time (Day)41 76532

P/P L/L

L/P

Est

rous

Cyc

leS

tage

E

D

P

M

P/L

E

D

P

M

Time (Day)41 76532

L/P


https://doi.org/10.1101/472688

L/P

CL

CL

L/L

L/P

CLCL

CL

CLL/P

CL CL

CL

CL

P/PA

CFs

hrm

RN

A(fo

ldch

ange

)

P/P L/L P/L L/P0

6

5

4

3

2

1

bb

aa

P/P L/L P/L L/P

2

10

8

6

4

0

12

Ova

ryw

eigh

t(m

g)

B

a

b

aa

P/P L/L P/L L/P

2

10864

0

1214

Cyp

19m

RN

A(fo

ldch

ange

)

E

b

b

aa

D

P/P L/L P/L L/P

10

50

40

30

20

0Cyp

17m

RN

A(fo

ldch

ange

)

c

b

aca

Fig. 3

CL

CLCL

P/L


https://doi.org/10.1101/472688

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

A

0 1 2 3 4 5Time Post-Treatment (Weeks)

Divers

ity (F

aith’s

PD)

0

10

20

30

40

P/P

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�


L/L

Divers

ity (F

aith’s

PD)

0

10

20

30

40

B

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�


C

Divers

ity (F

aith’s

PD)

0

10

20

30

40

P/L

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�

�


D

Divers

ity (F

aith’s

PD)

0

10

20

30

40

L/P

Fig. 4

LM: r = 0.23, p = 0.0003

LME: p = 0.003

LM: r = 0.05, p = 0.07

LME: p = 0.20

LM: r = 0.009, p = 0.23

LME: p = 1

LM: r = 0.08, p = 0.03

LME: p = 0.71


https://doi.org/10.1101/472688

●

●

●

●

●

●

●●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●●

●●

●

●

● ●

●

●

●

●

●

●

●

●●

●

●

●

●

● ●●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●

●●

●

●●

●

●

●●

●

●

●●●

●

●

●

●

●

●

●●●

●

●

●

●

●

●

●●

●

●

●

●●

●

CAP - Percent variation explained 1.3%

-3 -2 -1 0 1 2 3

-3 -2 -1 0 1 2 3-3 -2 -1 0 1 2 3

CAP2

- Perc

ent v

ariati

on ex

plaine

d 0.8%

-2

-1

0

1

2

3

-3

CAP - CAP1 vs CAP2 (weighted)Week 1



L/LP/L

P/P

L/LP/L

P/P

L/LP/L

P/P

ANOVAp=0.42

ANOVAp=0.004

ANOVAp=0.01

●

●

●

● ●

●

●●

●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●● ●

●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●●

● ●

●

●

●

●

●

●

●

●

●●

●

●

●●

●

●

●●

●

●

●

●

●

●

-3 -2 -1 0 1 2 3

-3 -2 -1 0 1 2 3

L/LP/L

P/P

L/LP/L

P/P




CAP2

- Perc

ent v

ariati

on ex

plaine

d 0.8%

-2

-1

0

1

2

3

-3

ANOVAp=0.01

ANOVAp=0.13

C D

E F

G

Fig 5.

P/PL/LP/LL/P

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●●

●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

● ●

●

●

●

●

●

●●

●

●

●

●

●

●

●

● ●●

●

●●

●

●

●

●

●

●

● ●

●

●

●

●

●

●

●

●

●

●●

●●

●

●

●

●

●

●

● ●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

● ●

●

●

●

●

●

●

●

●

●

● ●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

● ●

●

●

●

●

●

●

●●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●●

●

●

●

●

●

●

●

●

●

●●●

●

●

●

●

●

●

● ●●

●

●●●

●

●

●

●

●

●

●

●

ANOVAp=0.001

A BPCOA - PC1 vs PC2 (weighted)

Post-Treatment (Week 1-5)CAP - CAP1 vs CAP2 (weighted)

Post-Treatment (Week 1-5)

CAP - Percent variation explained 1.3%PC1 - Percent variation explained 6.1%

PC2 -

Perce

nt va

rition

expla

ined 4

%

CAP2

- Perc

ent v

ariati

on ex

plaine

d 0.8%

-0.3 -0.2 -0.1 0.0 0.1 0.2 0.3-0.3

-0.2

-0.1

0.0

0.2

0.1

-3 -2 -1 0 1 2 3

-2

-1

0

1

2

L/L P/L

P/P

ADONISp=0.001

●

CAP2

- Perc

ent v

ariati

on ex

plaine

d 0.8%

-2

-1

0

1

2

3

-3



https://doi.org/10.1101/472688

1.7

1.5

1.5

1.3

-1.2

-1.3

-2.3

-3.5Akkermansia***

Turicibacter***

Adlercreutzia**

Lactobacillus**

Dorea*

Roseburia*

Christensenella*

Candidatus Arthromitus***

Coprobacillus***

-2 0 2

2.9

Coprobacillus

Dorea

Adlercreutzia

Log2 Fold Change

Placebo vs. Letrozole (Weeks 2-5)

P/P L/L P/L L/P

01x( ecnadnubA4-

)

0

2

4

6

8

01x( ecnadnubA3-)

0.00.51.01.52.02.5

01x( ecnadnubA4-)

0.00.40.81.21.62.0

P/P L/L P/L L/P

P/P L/L P/L L/P

Fig.6

A B

C

D


https://doi.org/10.1101/472688

Scott T. Kelley, [email protected] Classification: Major- · 169 (n=8/group). The mice were divided...

Documents

Transcript of Scott T. Kelley, [email protected] Classification: Major- · 169 (n=8/group). The mice were divided...