Hepatic steatosis relates to gastrointestinal microbiota changes in obese girls with polycystic ovary syndrome

Objective Hepatic steatosis (HS) is common in adolescents with obesity and polycystic ovary syndrome (PCOS). Gut microbiota are altered in adults with obesity, HS, and PCOS, which may worsen metabolic outcomes, but similar data is lacking in youth. Methods Thirty-four adolescents with PCOS and obesity underwent stool and fasting blood collection, oral glucose tolerance testing, and MRI for hepatic fat fraction (HFF). Fecal bacteria were profiled by high-throughput 16S rRNA gene sequencing. Results 50% had HS (N = 17, age 16.2±1.5 years, BMI 38±7 kg/m2, HFF 9.8[6.5, 20.7]%) and 50% did not (N = 17, age 15.8±2.2 years, BMI 35±4 kg/m2, HFF 3.8[2.6, 4.4]%). The groups showed no difference in bacterial α-diversity (richness p = 0.202; evenness p = 0.087; and diversity p = 0.069) or global difference in microbiota (β-diversity). Those with HS had lower % relative abundance (%RA) of Bacteroidetes (p = 0.013), Bacteroidaceae (p = 0.009), Porphyromonadaceae (p = 0.011), and Ruminococcaceae (p = 0.008), and higher Firmicutes:Bacteroidetes (F:B) ratio (47.8% vs. 4.3%, p = 0.018) and Streptococcaceae (p = 0.034). Bacterial taxa including phyla F:B ratio, Bacteroidetes, and family Bacteroidaceae, Ruminococcaceae and Porphyromonadaceae correlated with metabolic markers. Conclusions Obese adolescents with PCOS and HS have differences in composition of gut microbiota, which correlate with metabolic markers, suggesting a modifying role of gut microbiota in HS and PCOS.


Study approval
The study protocols were approved by the University of Colorado Institutional Review Board and the Children's Hospital of Colorado Scientific Advisory Review Committee. Informed written consent or assent was obtained from all participants as appropriate for age, and parental written consent from all participants <18 years of age.

Study protocol
Participants had a screening visit for consent, physical exam and laboratory measurements to confirm eligibility. They then underwent a 2 day study visit which included stool collection, DEXA and abdominal MRI, followed by a monitored 12-hour inpatient fast with morning fasting blood collection and then an oral sucrose tolerance test consisting of 75 grams of glucola and 25 grams of fructose, survey completion and another physical exam. Waist circumference, BMI (kg/m 2 ), and BMI percentile per Center for Disease Control and Prevention BMI growth charts [26] were obtained. Abdominal MRI was used to assess hepatic fat via the DIXON method of the entire liver, as previously described [27] HS was defined as HFF � 5.0%. Hepatic stiffness was assessed with MR elastography. Total body fat percentage was assessed by standard DEXA methods (Hologic, Waltham, MA).

Physical activity
A 3-day pediatric activity recall (3DPAR) questionnaire was completed with staff assistance from all participants to assess habitual physical activity [28].

Dietary intake
A diet interview by study staff was completed using the SEARCH food frequency questionnaire (FFQ) to assess macronutrient patterns. The FFQ is defined to incorporate and represent common food choices among ethnically and regionally diverse youth aged 10-19 years [29].

Laboratory measurements
Fasting glucose, sex hormone concentrations, inflammatory markers and lipid profiles were measured. Glucose was measured by a StatStrip hospital grade glucometer (Nova Biomedical, Waltham, MA). Serum insulin and adiponectin were analyzed with RIA (Millipore, Billerica, MA); FFA (Wako Chemicals, Inc., Richmond, VA) were assessed enzymatically. HbA1c was measured by DCCT-calibrated ion-exchange HPLC (Bio-Rad Laboratories, Hercules, Calif). Alanine aminotransferase (ALT) and aspartame aminotransferase (AST) were measured by multipoint rate with P-5-P method (Vitros 1 5600, Ortho Clinical Diagnostics, Raritan, NJ); total cholesterol, high density lipoprotein cholesterol (HDL-C), and triglyceride assays were performed enzymatically on a Hitachi 917 autoanalyzer (Boehringer Mannheim Diagnostics, Indianapolis, IN). Low density lipoprotein cholesterol (LDL-C) concentrations were calculated by the Friedewald equation; highly sensitive C-reactive protein (hs-CRP) was measured via immunoturbidimetric assay (Beckman Coulter, Brea, CA), C-peptide via chemiluminescent immunoassay (DiaSorin, Stillwater, MN), and estradiol and progesterone via chemiluminescent immunoassay (Beckman Coulter, Brea, CA). Total testosterone was measured by highpressure liquid chromatography/tandem mass spectrometry, free testosterone via equilibrium dialysis and sex hormone binding globulin (SHBG) via chemiluminescent immunoassay, all by Esoterix laboratories (Calbassas Hills, CA). Hepatic fat fraction (HFF) was measured by MRI as previously described [17] and hepatic stiffness was measured by MR elastography. DXA was used to measure percent body fat and lean mass as previously described [30].

Fecal collection and microbiome analysis
Stool samples were collected at home the day prior to blood sampling using stool collection tubes and frozen in the participant's freezer. Upon return to study staff, samples were stored at -80˚C until further processing. Bacterial profiles were determined by broad-range analysis of 16S rRNA genes following our previously described methods [24,31] . Four participants with HS and five without received a one-time dose of a glucose modulating medication following stool collection and were not included in the HOMA-IR and Matsuda index calculations.

Statistical analysis
Data analyses were performed using R version 3.5.2 and Sigmaplot version 13.0. Data were examined for normality. Differences between the groups were compared with students t-tests or Mann-Whitney U, as appropriate. For categorical data either Fisher's exact tests or Pearson chi-square tests were performed to test differences between groups. The %RA of each taxon was calculated as the number of 16S rRNA sequences of a given taxon divided by the total number of 16S rRNA sequences in a patient's sample. Differences in overall microbiome composition (β-diversity) between subsets were assessed by a non-parametric, permutation-based multivariate analysis of variance (PERMANOVA with 10,000 replicate re-samplings) using Morisita-Horn dissimilarities. Shannon diversity, Shannon evenness, and richness (Sobs) (measures of α-diversity) were calculated using rarefaction and compared across groups using linear models adjusting for batch effects [37]. Comparisons of %RA across groups were performed using Wilcoxon rank sum tests since batch and race/ethnicity effects were not significant in any of the individual phyla, family or genus comparisons. Spearman's correlations were used to evaluate the relationship between %RA and metabolic and hormonal variables. Bacterial taxa with %relative abundance > 1% were used for correlations and markers of insulin resistance, obesity, fatty liver disease were used as variables. Results were adjusted for age, race/ethnicity and protein intake. The correlations were adjusted for multiple testing and those with p-values �0.05 were reported.

Clinical characteristics
Thirty-four girls completed the baseline assessment and returned the stool sample and thus were included in final analyses. The group was equally split into those with or without HS (n = 17 for each). Participant demographic and physical characteristics and laboratory measurements are summarized in Table 1. The groups had similar age, age of menarche, and family histories of type 2 diabetes. There was however a significant difference in race/ethnicity across groups, with more Hispanic representation in the HS group. Physical characteristics including BMI, waist-to-hip ratio and blood pressure were similar across groups. Both groups reported a similar percentage of dietary fat, protein and carbohydrate intake, and habitual physical activity.
The groups had similar free and total testosterone, SHBG, estradiol and progesterone. Per study design, girls with HS had significantly higher HFF compared to those without HS, as well as ALT. There were no group differences in AST, hepatic stiffness or body composition. Girls with HS had higher fasting insulin and C-peptide, and HbA1c and were more insulin resistant (higher HOMA-IR and lower Matsuda index) than those without HS. There were no differences in fasting glucose, 2-hour glucose or insulin, triglycerides, total cholesterol, HDL-C, LDL-C or adiponectin between groups. Markers of inflammation including WBC, platelets, and hs-CRP were also similar between groups.

Dysbiosis in hepatic steatosis
Bacterial 16S rRNA gene profiling was completed for all samples; both groups had adequate depth of sequencing coverage (Good's coverage of >99.0% for all samples) indicating comparable and representative samples. Girls with HS had numerically but not statistically significantly measures of alpha diversity including lower bacterial richness (p = 0.202) and evenness (p = 0.087) and higher diversity (p = 0.069) (Fig 1) compared with those without HS. The βdiversity, reflecting overall gut microbial community composition, was similar between groups (R 2 = 0.036, p = 0.35). There were still no differences in α-and β-diversity measures after adjusting for group differences in race/ethnicity, age, and protein intake percentage. Actinobacteria and Firmicutes were the most predominant phyla in those with HS, and Firmicutes and Bacteriodetes were the most dominant phyla in girls without HS. At the phylum level, girls with HS had significantly lower percent relative abundance (%RA) of Bacteroidetes and higher Firmicutes:Bacteroidetes (F:B) ratio (47.8% vs. 4.3%, p = 0.018) than those without HS. At the family level, girls with HS had lower %RA of Bacteroidaceae (p = 0.009), Porphyromonadaceae (p = 0.011), Ruminococcaceae (p = 0.008), and higher Streptococcaceae (p = 0.034) than those without HS. Fig 2 depicts the comparison of %RA at the phyla, family, and genus levels.

Bacterial taxa are associated with metabolic markers
Several taxa correlated with hepatic steatosis and markers of the metabolic syndrome as shown in Table 2. A higher F:B ratio was correlated with more central adiposity (higher waist-to-hip ratio), higher ALT and HFF, and insulin resistance as assessed by two-hour OSTT insulin. A lower %RA of Bacteroidetes, Bacteroidaceae, Porphyromonadaceae, and Ruminococcaceae were correlated with HFF. A lower %RA of Ruminococcaceae was correlated with higher triglycerides.

Discussion
The gut microbiome is different in individuals with either PCOS or NAFLD, as compared to controls. We have demonstrated the novel finding that adolescents with PCOS, obesity and HS have an altered gastrointestinal microbiota compared to those with PCOS and obesity without HS. Significant differences were noted in the %RA of several phyla, families, and genera by HS  status, and these bacterial taxa were significantly correlated with multiple metabolic markers related to NAFLD and insulin resistance. While there were some non-significant group differences in α-diversity, the β-diversity, which indicates global microbial community alteration between groups, these did not differ by HS status. Thus, it appears that the addition of HS beyond obesity and PCOS status is associated with changes in specific microbiota, but not overall global changes in the microbiome. Our findings are consistent with results in adolescents and adult women with either PCOS or NAFLD, though there are limited data in adolescent girls with both PCOS and NAFLD. Zhu et al. demonstrated that ecological differences (α-diversity and β-diversity) in the gut microbiota among adolescent girls and boys (age 12-14 years) are related to health status, obesity, and NASH [38]. Another study by Chierico et al. found a significant difference in β-diversity when comparing normal weight healthy youth controls to those diagnosed with obesity without NAFLD, HS alone, and NASH. However, this study also found no difference in β-diversity between youth with obesity without NAFLD, HS alone, and NASH [39]. The combination of these findings and ours suggest that obesity may potentially play a role in influencing diversity measures, though we did not have normal weight group for comparison to confirm this.
Relative abundance of specific bacteria can vary with obesity and NAFLD status. For example, adolescents with NASH and obesity had predominantly Prevotella-rich microbiota [40], whereas non-obese, non-NASH groups were more frequently associated with Bacteroides rich enterotypes [38]. In contrast, we found that girls with PCOS, obesity and HS had lower %RA of the family Prevotellaceae, but had lower %RA of Bacteroides. Zhu et al. found a statistically significant decrease in Bacteroidetes and an increase in Firmicutes in a non-obese adolescent group when compared to adolescents with simple obesity and to those with NASH. We found a statistically significantly lower amount of Bacteroidetes in those with HS and no difference in Firmicutes between groups. The study by Chierico et al. in boys and girls (mean age, 10-12 years) comparing non-obese control youth to youth with obesity, NAFLD, and NASH alone, found that those with NAFLD had higher proportion of Actinobacteria and Proteobacteria compared to NASH, obese, and healthy control, and reduced Bacteroidetes and Firmicutes compared to youth with obesity [39]. This study also found that participants with NAFLD and NASH had gut microbiota signatures with an increase in %RA of Ruminococcus and Dorea, whereas we found lower %RA of Ruminoccocaeae in our HS cohort. The differences seen in our patient population compared to non-PCOS NAFLD patients could potentially reflect changes influenced by age, PCOS status, local dietary patterns and only including female sex participants. Future provocative interventional studies would be needed to confirm if these findings of association do indeed have mechanistic underpinnings.
The pathophysiological link between PCOS and NAFLD remains unclear [16,41]; however, insulin resistance and obesity are common critical components in both NAFLD and PCOS [8,13,42]. There is also evidence that insulin resistance and hyperandrogenism mediate the relationship between PCOS and NAFLD [43]. Studies have demonstrated that hyperandrogenic women with PCOS had higher liver fat compared with women with PCOS based on the Rotterdam criteria with normal androgens or with healthy controls [44]. We found significant  correlations between HFF and several bacterial families and phyla, suggesting a relationship between HFF and the gut microbiome, but not androgen concentrations. Another study demonstrated that women with PCOS and NAFLD had decreased hepatic LDL receptor expression, and hypothesized that hyperandrogenism may putwomen with PCOS at risk for development of dyslipidemia and NAFLD [45]. Although we did not find correlations between LDL-C and bacterial taxa, we had several taxa that correlated with characteristics of the metabolic syndrome and with a marker of insulin resistance, suggesting that the gut microbiota may relate to increased risk of T2D, NASH, and cardiovascular disease. Additionally, alcohol producing bacteria may contribute to the pathogenesis of NAFLD in PCOS. For example, Zhu et al. found that ethanol metabolism and Enterobacteriaceae have a functional relationship in contributing to the development of NASH [38]. Adolescent patients with NASH were also found to have upregulation of ethanol metabolism compared to controls [39]. In addition to Enterobacteriaceae, Bacteroides, Bifidobacterium and Clostridium are also alcohol producing bacteria [38]. Though we did not measure blood or breath alcohol concentration, we found higher %RA Bifidobacterium in our adolescents with HS, which suggests that bacterial taxa involved in ethanol production may contribute to endogenous ethanol production in NALFD in PCOS. Limitations to our study include the small sample size and lack of blood and breath alcohol tests, which may have provided further understanding of NASH pathogenesis in girls with PCOS. HOMA-IR, the Matsuda index and insulin values were used to estimate insulin sensitivity, rather than a gold-standard hyperinsulinemic euglycemic clamp and thus we were also not able to assess tissue specific of insulin sensitivity. It is also unknown if our participants had NASH or liver fibrosis as we did not perform liver biopsy, although MR elastography results do not indicate notable stiffness. The groups were not matched for ethnicity, with a greater proportion of Hispanics in the NAFLD group, consistent with a higher prevalence of NAFLD in those with Hispanic origin. We attempted to mitigate this group difference by adjusting analysis for race/ethnicity status, but it is possible that there is a reflection of underlying increased risk for NAFLD in this group. We can only comment on associations between measures, since our study design was not longitudinal, and thus our findings are hypotheses generating for future provocative studies on causation. There are several unique strengths to our study. Our groups were similar in terms of age, age of menarche, pubertal stage, BMI, diet and physical activity and PCOS markers. Further, liver fat of the entire liver was measured using MRI instead of using liver ultrasound or relying on only laboratory liver enzymes. Finally, we used the NIH criteria to define PCOS, which identifies a more metabolically at-risk population.

Conclusion
In girls with obesity and PCOS, the composition of the gut microbiota is different in those with HS compared to those without HS. In this cohort, HS was associated with alterations in the gut microbiota that are typically related to metabolically unhealthy obesity. Furthermore, in the overall cohort, certain taxa at the phylum and family level were correlated with insulin resistance, and the metabolic syndrome characteristics of central adiposity, and elevated triglycerides showing a relationship between the gut microbiota and increased risk of T2D, NASH, and cardiovascular disease. Our findings suggest that there is a relationship between the gut microbiome and metabolic disease in adolescents with HS and PCOS, but it remains unclear which components come first and whether the relationships are causative or just associations. Further work is warranted to better understand the pathogenesis of HS and PCOS, the role of the gut microbiota in adolescence and to potentially develop therapies in the future to help reduce risk of T2D, cardiovascular disease and liver disease.