Study design and patient population:
A total of 36 fecal samples were analyzed in this work. Among them, 24 had been obtained from virologically suppressed PWH (> 55 years old) from our frailty cross-sectional study 10, including 7 frail, 9 pre-frail, and 8 robust ones, matched for age and nadir CD4. The samples had been stored frozen (–80 ºC) since then. In the frame of such study, the population was screened for the prevalence of frailty and tested for physical function, depression, nutritional status, and associated factors 10. In the present study, 12 non-HIV healthy people with a similar age distribution were also included as controls.
Data collection
Sociodemographic data, comorbidities (ie, self-reported and physician-diagnosed chronic conditions), medications (ie, polypharmacy was defined as taking ≥ 6 medications), and variables related to HIV infection (ie, risk practice for HIV infection, the baseline and current immunovirologic status, and the stage of HIV infection at diagnosis) were recorded. Depression status was evaluated using the Short Geriatric Depression Scale (ie, S-GDS or Yesavage test) 19.Physical function was assessed by quantifying using the Short Physical Performance Battery (SPPB) 20.The recorded laboratory data included HIV-related data (ie, HIV-RNA, nadir and current CD4 count, and CD4/CD8 rate).
Frailty
Frailty was assessed according to Fried’s frailty phenotype, defined by 5 functional criteria 21,namely shrinking (unintentional weight loss of ≥ 4.5 kg or ≥ 5% of body weight during the previous year), weakness (grip strength adjusted for gender and BMI), poor endurance and energy (self-reported exhaustion identified by 2 questions from the Center for Epidemiologic Studies Depression scale), slowness (based on the time to walk 4 meters, adjusting for gender and standing height), and low physical activity level (< 383 kcal/week in men and < 270 kcal/week in women using the Minnesota Leisure Time Activity Questionnaire). Patients were considered frail when they met at least 3 of the 5 criteria, pre-frail when they met 1 or 2 criteria, and robust when they met no criteria.
Metataxonomic analysis
A dual-barcoded 2-step PCR reaction was conducted to amplify a fragment of the V3–V4 hypervariable region of the bacterial 16S ribosomal RNA (rRNA) gene. Equimolar concentrations of the universal primers S-D-Bact-0341-b-S-17 (ACACTGACGACATGGTTCTACACCTACGGGNGGCWGCAG) and S-D-Bact-0785-a-A-21 (TACGGTAGCAGAGACTTGGTCTGACTACHVGGGTATCTAATCC) were used. Barcodes used for Illumina sequencing were appended to 3′ and 5′ terminal ends of the PCR amplicons to allow for the separation of forward and reverse sequences. A bioanalyzer (2100 Bioanalyzer, Agilent) was used to determine the concentration of each sample. Barcoded PCR products from all samples were pooled at approximately equimolar DNA concentrations and run on a preparative agarose gel. The correctly sized band was excised and purified using a QIAEX II Gel Extraction Kit (Qiagen) and then quantified with PicoGreen (BMG Labtech, Jena, Germany). Finally, 1 aliquot of pooled, purified, and barcoded DNA amplicons was sequenced using the Illumina MiSeq pair-end protocol (Illumina Inc., San Diego, CA, USA) at the facilities of the Scientific Park of Madrid (Spain). The sequences analyzed for this study are available in the BioSample database of the National Center for Biotechnology Information.
The amplified fragments and results were taxonomically analyzed using the Illumina™ software according to the manufacturer’s guidelines and pipelines (version 2.6.2.3 San Diego, CA, USA). The resulting high-quality reads were assembled and classified taxonomically into operational taxonomic units (OTUs) by comparison with the Illumina™ software according to the manufacturer’s guidelines and pipelines (version 2.6.2.3) using a Bayesian classification method and a level of similarity of at least 97%.
The concentration of DNA in the 3 blank preparations was approximately 0.01 ng/µL. The decontam R package cite was used to identify, visualize, and remove contaminating DNA based on the DNA concentration in each sample.
Statistical analysis
We used descriptive statistics to examine participant characteristics, which were expressed as frequency (percent) for categorical variables, mean (± SD) for normally distributed continuous variables, or median (interquartile range) for continuous variables with a skewed distribution.
We compared continuous variables using the t-test for independent variables. Then, we used the Wilcoxon and Mann–Whitney test for variables with 2 factors with a non-normal distribution or when the group size was small, and the Kruskal-Wallis test was used with variables with 3 or more factors and non-normal distribution. We assessed the association between qualitative variables using the chi-square test or the Fisher exact test when the groups were very small. In addition, we used linear regression to assess the differences in biological age between frail and robust patients, and we considered differences to be significant when p ≤ 0.05. We used SPSS statistical package (version 20.0).
For bacteriome, analysis quantitative data were expressed as the median and interquartile range (IQR). We assessed differences between groups using Kruskal–Wallis tests and pairwise Wilcoxon rank sum tests to calculate comparisons between groups. Also, we made Bonferroni corrections to control for multiple comparisons. We generated a table of amplicon sequence variants’ OTU counts per sample and normalized the bacterial taxa abundances to the total number of sequences in each sample. Then, we studied alpha diversity using the Shannon diversity index with the R vegan package (version 2.5.6; Oksanen, 2007). We used principal coordinates analysis (PCoA) to evaluate beta diversity and to plot patterns of bacterial community diversity through a distance matrix containing a dissimilarity value for each pairwise sample comparison. We performed quantitative (relative abundance) and qualitative (presence/absence) analyses using the Bray–Curtis index and binary Jaccard index, respectively. Then, we performed analysis of variance of the distance matrices using the “nonparametric manova test” (PERMANOVA) adonis with 999 permutations, as implemented in the R vegan package, to reveal statistical significance. For multilevel pairwise adonis comparisons, we used the Holm–Bonferroni method for p-value correction using the “pairwiseAdonis” R package (version 0.0.1). We performed the linear discriminant analysis (LDA) effect size (LEfSe) algorithm to predict those taxa that violate the null hypothesis of no difference between the control and PWH groups of patients. We performed this analysis with the online interface Galaxy 22.