Microbial Lineages in Sarcoidosis. A Metagenomic Analysis Tailored for Low-Microbial Content Samples

Abstract

Rationale: The etiology of sarcoidosis is unknown, but microbial agents are suspected as triggers.

Objectives: We sought to identify bacterial, fungal, or viral lineages in specimens from patients with sarcoidosis enriched relative to control subjects using metagenomic DNA sequencing. Because DNA from environmental contamination contributes disproportionately to samples with low authentic microbial content, we developed improved methods for filtering environmental contamination.

Methods: We analyzed specimens from subjects with sarcoidosis (n = 93), control subjects without sarcoidosis (n = 72), and various environmental controls (n = 150). Sarcoidosis specimens consisted of two independent sets of formalin-fixed, paraffin-embedded lymph node biopsies, BAL, Kveim reagent, and fresh granulomatous spleen from a patient with sarcoidosis. All specimens were analyzed by bacterial 16S and fungal internal transcribed spacer ribosomal RNA gene sequencing. In addition, BAL was analyzed by shotgun sequencing of fractions enriched for viral particles, and Kveim and spleen were subjected to whole-genome shotgun sequencing.

Measurements and main results: In one tissue set, fungi in the Cladosporiaceae family were enriched in sarcoidosis compared with nonsarcoidosis tissues; in the other tissue set, we detected enrichment of several bacterial lineages in sarcoidosis but not Cladosporiaceae. BAL showed limited enrichment of Aspergillus fungi. Several microbial lineages were detected in Kveim and spleen, including Cladosporium. No microbial lineage was enriched in more than one sample type after correction for multiple comparisons.

Conclusions: Metagenomic sequencing revealed enrichment of microbes in single types of sarcoidosis samples but limited concordance across sample types. Statistical analysis accounting for environmental contamination was essential to avoiding false positives.

Keywords: bacterial 16S ribosomal RNA; fungal internal transcribed spacer ribosomal RNA; metagenomic; microbiome; sarcoidosis.

Figure 1.

Dominant bacterial and fungal orders in lymph node tissue samples and matched controls (set A). The major bacterial (A) and fungal (B) orders identified by 16S and internal transcribed spacer ribosomal RNA gene sequencing, respectively, are shown as proportions of total reads. Less-common lineages are aggregated under “Other.” For each pair, the solid circle indicates the formalin-fixed, paraffin-embedded lymph node sample, and the open circle indicates a slice of blank paraffin cut from the same block to serve as an environmental control. Empty (white) bar charts indicate that the sample was either not available or had no detectable lineages. The difference in Cladosporiaceae reads between a sample and its environmental control are shown in C. Solid circles represent samples with more Cladosporiaceae reads in the sample than the matched environmental control, and open circles represent samples in which the number of Cladosporiaceae reads were not greater than in the environment control. The abundances are shown as reads to more accurately reflect the input to the test, which used raw read counts as input. Normalization between differing sequencing depths was accounted for by modeling library size as a random effect for each sample (see Methods).

Figure 2.

Bacterial and fungal lineages in lymph node tissue samples (set B). The major bacterial (A) and fungal (B) orders identified by 16S ribosomal RNA and internal transcribed spacer (ITS) gene sequencing are shown as proportions of total reads. Less-common lineages are aggregated under “Other.” Seventeen samples failed to amplify any usable ITS sequences in B and are omitted. Blank paraffin controls matched to each tissue specimen were not available for these samples.

Figure 3.

Bacterial, fungal, and viral lineages in BAL (sample set C). The major bacterial (A) and fungal (B) lineages identified by 16S ribosomal RNA (rRNA) and internal transcribed spacer (ITS) rRNA sequencing and viral (C) lineages identified by shotgun sequencing of all nucleic acids in virus particle preparations are shown as proportions of the total reads. Data are shown at the order level for A and B and the family level for C. Less-common lineages are aggregated under “Other.” For each pair, the solid circle indicates the BAL fluid, and the open circle represents the prewash fluid for that scope. Empty (white) bar charts indicate that the sample was either not collected or had no detectable lineages. Three sample pairs failed to amplify any ITS sequences and are omitted from B.

Figure 4.

Microbial lineages in the Kveim reagent (set D) and a granulomatous sarcoid-involved spleen (set E). The major lineages in sample sets D (Kveim) and E (sarcoidosis spleen) shown by sequencing. (A) Results from 16S sequencing, (B) internal transcribed spacer sequencing, and (C) results from whole-genome shotgun sequencing (WGS), after filtering, as described in the online supplement. Less-common lineages are aggregated under “Other,” including fungal detections in C.