Microb Ecol DOI 10.1007/s00248-013-0266-4 SHORT COMMENTARY A Unique Signal Distorts the Perception of Species Richness and Composition in High-Throughput Sequencing Surveys of Microbial Communities: a Case Study of Fungi in Indoor Dust Rachel I. Adams & Anthony S. Amend & John W. Taylor & Thomas D. Bruns Received: 24 January 2013 /Accepted: 8 July 2013 # The Author(s) 2013. This article is published with open access at Springerlink.com Abstract Sequence-based surveys of microorganisms in sequenced to an even depth. Next, we used in silico manip- varied environments have found extremely diverse assem- ulations of the observed data to confirm that a unique signa- blages. A standard practice in current high-throughput se- ture can be identified with HTS approaches when the source quence (HTS) approaches in microbial ecology is to se- is abundant, whether or not the taxon identity is distinct. quence the composition of many environmental samples at Lastly, aerobiology of indoor fungi is discussed. once by pooling amplicon libraries at a common concentra- tion before processing on one run of a sequencing platform. Biomass of the target taxa, however, is not typically deter- mined prior to HTS, and here, we show that when abun- Introduction dances of the samples differ to a large degree, this standard practice can lead to a perceived bias in community richness Use of high-throughput sequencing (HTS) has transformed and composition. Fungal signal in settled dust of five uni- the field of microbial ecology. Traditionally, sampling depth, versity teaching laboratory classrooms, one of which was defined both in terms of the number of samples and intensity used for a mycology course, was surveyed. The fungal of analysis within those samples, has been limited due to richness and composition in the dust of the nonmycology resources required to assess microbial composition. Howev- classrooms were remarkably similar to each other, while the er, with modern HTS technologies, a comparable effort can mycology classroom was dominated by abundantly sporu- result in hundreds of thousands and even millions of se- lating specimen fungi, particularly puffballs, and appeared to quence reads in a single run [1–3]. These tools have been have a lower overall richness based on rarefaction curves and directed toward understanding the presence and role of mi- richness estimators. The fungal biomass was three to five crobes in terrestrial [4] and marine [5] environments, includ- times higher in the mycology classroom than the other class- ing where the microbes associate with “macrobes” such as rooms, indicating that fungi added to the mycology class- plants [6], insects [7], and humans [8]. room swamped the background fungi present in indoor air. One of the advantages of HTS is that a high number of Thus, the high abundance of a few taxa can skew the per- different environmental samples can be processed simulta- ception of richness and composition when samples are neously through “multiplexing,” where individual samples are PCR-amplified with barcoded primers that are later used to computationally separate them. The accepted strategy is to Electronic supplementary material The online version of this article (doi:10.1007/s00248-013-0266-4) contains supplementary material, pool these amplicon libraries at equimolar concentrations which is available to authorized users. before sequencing together. Despite common concentra- tions, sequence counts are uneven across the samples, so in R. I. Adams (*) : J. W. Taylor : T. D. Bruns Department of Plant & Microbial Biology, University of California, the analysis stage, the sequence reads are rarefied to a com- Berkeley, CA 94720, USA mon number across all samples. The resulting assumption is e-mail: [email protected] that an even sampling depth, determined by both the con- centration of amplified DNA and subsequent rarefaction, A. S. Amend Department of Botany, University of Hawaii at Manoa, will give a representative glimpse of the true biological Honolulu, HI 96822, USA community within. R. I. Adams et al. One environment that is increasingly a target for microbial reported mold problems (Supplementary Fig. 1). The passive surveys is that of the built environment [9]. Not only a unique collection of settled dust on a sterile sampling surface fol- ecological habitat, the indoor environment is one that can have lows that of the “dustfall collector” developed by Würtz et al. important implications for human health [10], and identifying [14]. Here, we used a base of an empty 9-cm petri dish, a exposure often involves detecting a unique contaminant or an method previously used to sample bioaerosols in residences elevated level of a common source. Recent studies using [12], to collect dust for the duration of the fall teaching culture-independent techniques show that the outdoors is the semester, from late August 2011 through mid-December dominant origin for fungi detected indoors, and this pattern 2011. Samplers were placed on top of shelves at a height of occurs at both global and local scales and independent of 2.1 m and at a minimum of 1 m from a supply air outlet. The building function or human activity [11–13]. In a recent study volume of the classrooms ranged from 340 to 377 m3. Four of residences, no statistical differences in fungal assemblages pairs of samplers were distributed in each lab (Supplemen- could be found between rooms such as kitchens, bathrooms, tary Fig. 1). Among the classes held in classroom E was a bedrooms, and living rooms, even though they have obvious mycology course in which both instructors and students differences in uses, water sources, and potential substrates for brought local specimens into the room for study. Other fungi; instead, each room appeared to be a random subset of the rooms held laboratory courses on various topics related to outdoor fungal community [12]. This latter result may simply microbiology and plant biology but not mycology. All of the show again that outdoor sources dominate the process of fungal labs have identical custodial regimes and ventilation treat- community assembly. Alternatively, it may show a limitation of ments, in that fume hood air vents directly outside of the sampling dust with a HTS approach, in which the accumulated building and return vent air is circulated through the venti- contribution of extremely diverse outdoor sources (composed lation system (Supplementary Fig. 1). of hundreds to thousands of taxa) masks the few active resi- dents of individual rooms, a particular concern if trying to use DNA Extraction, Amplification, and Sequencing HTS methods to identify “sick buildings.” In preparation for the global study of settled dust [11]and Sample processing of dust followed the methods of Adams as part of a class study, we sampled vacuumed dust in the et al. [12]. Briefly, dust was first collected from the dish on a mycology teaching lab at University of California, Berkeley, sterile, cotton-tipped, wooden-handled swab by moistening analyzed it via pyrosequencing, and found an unexpectedly the swab in nucleic acid-free water, rubbing across the dish high representation of puffball taxa that were clearly derived for 5 s, twisting the dish 90°, and rubbing for another 5 s. from material previously used in teaching. This observation Genomic DNA was extracted from the swab tip using a provided us with a room with a unique fungal signature in chloroform–phenol procedure in conjunction with a soil airborne dust. In the current study, we followed up on this extraction kit. The extraction efficiency of this method was observation by passively sampling circulating dust over the not quantified. We targeted the internal transcribed spacer course of one semester in this lab and in four adjacent labs (ITS) region 1 of the nuclear ribosomal coding cistron [15], housed on the same floor of a single building. We then used using the ITS1F and ITS2 primers [16, 17], as these primers multiple in silico manipulates of the observed data to test the are predicted to amplify a diverse set of fungal lineages [18], sensitivity of the method to identify community differences the ITS1 region is taxonomically informative [19–21], and between rooms. Specifically, we reassigned puffball reads to shorter amplicon regions tend to increase the number of taxa that were already dominant in the background communi- reads per sequencing effort (compared to, for example, se- ty, and we lowered the read abundance of the abundant puff- quencing the entire ITS region, including ITS2). Genomic ball sequence to match that of common indoor fungi. Togeth- DNAs from the 40 environmental samples were PCR- er, this study allowed us to explore two questions as follows: amplified in triplicate at a 1:10 dilution, purified, quantified, (1) How sensitive are estimates of richness to differences in and pooled at equimolar concentrations for pyrosequencing abundance? (2) What is required for an assemblage to appear at the W.M. Keck Center for Comparative and Functional statistically different from other assemblages when the differ- Genomics of the University of Illinois at Urbana-Champaign ences are layered on top of a diverse background community? on 1/8th of a 454 FLX+ picotiter plate. Raw pyrosequences were submitted to the Sequence Read Archive of the Nation- al Center for Biotechnology Information under accession Materials and Methods number SRP018151. To assess the total fungal biomass in each of the samples, we employed quantitative PCR using Sample Collection the FF2/FR1 primers developed by Zhou et al. [22], which target a locus of nearly invariant length of the fungal 18S Samples were taken from five laboratory-type classrooms ribosomal genes. Standard curves were generated using di- located on the second story of a two-story building with no luted concentrations of Penicillium purpurogenum spores Biomass, Perceived Richness, and Microbial Surveys (see [12] for more details).