Turning “Sarkoid” into “Dropsy”. A Valiant, Next-Generation Attempt

My search for micro-organisms has produced no result.

—Cæsar Boeck, “Multiple Benign Sarkoid of the Skin”

Though the word has vanished from rounds, “dropsy” was considered a common and well-characterized diagnosis throughout most of the history of medicine. Vividly described in the 14th century as “a watery disease inflating the body” (1), dropsy has since lost its status as a “diagnosis” and is now a mere “finding” (diffuse edema of the abdomen and extremities) in need of a workup. We now know that diffuse edema can reflect various internal organ dysfunctions (of the liver, kidney, heart, lungs, or blood) caused by innumerable underlying etiologies (e.g., viral hepatitis, diabetic nephropathy, ischemic cardiomyopathy, emphysema with cor pulmonale, and malnutrition).

For the past 140 years, we’ve been trying to turn “sarcoidosis” into a similarly archaic-sounding amalgam of pathologies. We’ve long suspected that the disease’s granulomatous inflammation, like the edema of dropsy, is merely a final common manifestation of yet-unidentified etiologies. In the first report of the disease, Jonathan Hutchinson speculated that it could be infectious (“a tuberculous affection”) or autoimmune (“of the lupus family”) (2). We’ve since broadened the hunt to environmental, genetic, and primary immunologic causes. As Freiman noted in 1948, “at some time or other practically every agent capable of producing a granulomatous reaction . . . has been suggested as a possible cause of sarcoidosis” (3). Yet, with few exceptions (e.g., berylliosis and common variable immunodeficiency), the disease’s instigating triggers remain unknown.

Occult microbial exposure has perennially led the list of suspected etiologies. For the first half of the 20th century, investigators debated whether the disease is a manifestation of tuberculosis (3). Since then, various other microbes have been implicated but never convicted, including various bacteria (most commonly Propionibacterium spp. and nontuberculous mycobacteria [4]) and viruses (e.g., herpesviruses [5]). Each technological advance in microbial detection (e.g., in situ hybridization and PCR) has prompted renewed interest in finding the disease’s elusive microbial trigger.

In this issue of the Journal, Clarke and colleagues (pp. 225–234) share what is surely the most ambitious and comprehensive survey to date seeking a microbial cause of sarcoidosis (6). This time around, the technological advance prompting a new look at this old question is the recent revolution in next-generation sequencing. The authors summoned a variety of sequencing strategies (16S ribosomal RNA gene sequencing for bacteria, internal transcribed spacer ribosomal RNA gene sequencing for fungi, and shotgun metagenomics for viruses) to cast the broadest possible net. The breadth of their sequencing approach was matched by the variety of tissues they studied, which included BAL fluid, two cohorts of surgically excised lymph nodes, fresh splenic tissue, and the Kveim reagent (homogenized splenic tissue historically used for diagnostic provocation testing). The authors’ molecular approach was more sophisticated and their specimen set more comprehensive than those used in any prior effort.

Thus, if ever there was a study equipped to find the disease’s missing microbial culprit, this was it. Yet the authors found no smoking gun. No single microbe—bacterial, fungal, or viral—was consistently enriched across tissue specimen types in patients with sarcoidosis. They did identify some provocative candidates within individual specimen types (Cladosporia among lymph nodes and Corynebacteria in BAL fluid), but if their goal was to uncover a subtle, smoldering pathogen analogous to Tropheryma whippelii in Whipple disease or Helicobacter pylori in peptic ulcer disease, the search came up empty.

Yet I see cause for celebration in these negative findings. With them, the authors have provided us with a valuable demonstration of how to handle the crucial challenge of DNA contamination. As molecular tools of microbial identification have increased in sensitivity, they have also shown increased vulnerability to contamination. Microbial DNA can be detected—and sequenced—in virtually everything used to study the lung microbiome, including prelavage saline, sterile water, extraction reagents, and even the blank wells of the sequencers. This contaminating DNA, as well as the “batch effect” introduced by clustered specimen processing, is a rampant source of type I errors (false positives) in low-biomass microbiome studies (7). Although much of the field has been regrettably slow to address this issue, Clarke and colleagues got it right: they sequenced copious negative control specimens representing a variety of potential sources of contamination, and asked whether each significant finding could be attributed to contamination and false clustering. And indeed, they demonstrated that specific sarcoid-enriched taxa (Aspergillus and Penicillium spp.) were almost certainly artifacts of environmental contamination.

Sequencing contamination is a central methodological challenge in the lung microbiome field, and we can’t say we weren’t warned. The purported discoveries of the “placental microbiome” (microbial DNA detected in placental tissue) and “dinosaur DNA” (the premise of Jurassic Park) both received enormous publicity when they were first reported (8, 9), but both have subsequently been attributed to false signals from DNA contamination (10, 11). If we wish to avoid a similar fate, we must hold ourselves to the same high standard of rigor as the authors did in their study.

Is this the final word on the microbial origins of sarcoidosis? Surely not. Even the most “universal” of sequencing strategies has blind spots, biases, and limits of detection (12), and I can think of several that are pertinent here. The study’s lymph node specimens were fixed in formalin, which degrades DNA (13). Mycobacteria are notoriously stubborn against DNA extraction and are underrepresented in community sequencing. Reference databases for viral and fungal sequences are less complete than their bacterial equivalents. Archaea (an entire domain of life, ubiquitous in nature but not yet implicated in human disease) are poorly amplified by our 16S primers (14). Illumina sequencers are prone to “index switching,” in which sequences are misattributed across specimens (another cause, aside from contamination, of false signal in low-biomass specimens) (15). Finally, several sarcoid-suspect taxa, including Propionibacteria and atypical mycobacteria, are environmental microbes and thus may legitimately appear in both environmental controls and patient tissues. Any of these factors may have contributed to type II errors (false negatives), and will likely one day be used to justify another microbial survey.

At the turn of the 20th century, in the first account of sarcoidosis as a systemic disease, Cæsar Boeck wistfully admitted, “my search for micro-organisms has produced no result” (16). More than a century later, despite our arsenal of molecular tools, we must make the same concession. Our struggle to turn sarkoid into dropsy continues.