To our knowledge, this is the first time that the effect of moisture and moisture damage remediation on indoor fungal assemblages has been studied using a whole community approach and source tracking. It is also the first study to compare fungal community composition using a large selection of species-specific qPCR assays and clone library sequencing in combination with culture. We found increased fungal diversity in one of the studied buildings with moisture damage, while in the second damaged building, high numbers of Penicillium were present. In neither building did we find a concomitant increase in culturable fungal concentrations or fungal biomass in surface dust. A majority of the fungal species isolated from contaminated building materials was not prevalent in the pre-remediation dust samples collected from those buildings. Methodological comparison indicated that cultivation in combination with a large qPCR panel, failed to detect a majority of the fungi in indoor samples; however, the most abundant species appeared to be detected by all methods. Clone library sequencing, to the extent used here, was found to be less sensitive than qPCR for detecting individual species.
Fungal diversity in dust samples
Cloning and sequencing studies revealed an average of 54 observed and 146 estimated species-level phylotypes (OTUs) per sample. This level of diversity is similar to that observed previously using molecular methods in floor dust and indoor air filter samples [21–23] and higher than that detected in outdoor air filter samples [27, 28].
The dominant genera we observed in dust and material samples were in agreement with previous studies using cultivation [29–32]; Aureobasidium, Cladosporium and Penicillium were the most prevalent genera in dust according to molecular and culture-independent methods. These and other common indoor mold genera, including Aspergillus, Botrytis, Epicoccum, Eurotium, Fusarium, Mucor, Rhizopus, Trichoderma, Ulocladium, Wallemia and Phoma/Sphaeropsidales-group fungi accounted for 95-96% of total CFUs and qPCR CE counts and approximately 40% of clones in nucITS libraries. The remaining 60% of nucITS clones, however, accounted for almost 90% of the total diversity in the sequence material, showing that a vast diversity of indoor fungi remain uncharacterized by cultivation or targeted molecular methods. While the proportion of individual sequence types representing the uncultivable diversity was low in the material, it must be remembered that the clone library sequencing method does not accurately reflect the original proportions of species in the community and both under- and over-estimating bias may occur . Our results from individual qPCR assays indeed showed that the species occurring as singletons in nucITS libraries were in many cases abundant taxa, commonly between 104-105 CE g-1 of dust. According to previous data from Finland and the US, the median qPCR assayed concentrations of many common indoor fungi, e.g. Aspergillus spp., Epicoccum
nigrum, the Eurotium
amstelodami group, Penicillium spp. and Trichoderma
viride are between 104 and 105 CE g-1 of floor dust [18, 34]. No such data are available for settled dust collected from elevated surfaces, but the fungal concentrations in the latter sample type can be expected to be similar or lower than those in floor dusts [22, 35].
Based on the number of described fungal species  and estimates on total global fungal biodiversity  nearly 90% of fungal biodiversity may as yet be unidentified. A large proportion of unidentifiable phylotypes was observed in our sequence material also. In total, 42% of OTUs could only be identified to the class or phylum level, or remained of unknown affiliation. This is comparable to previous studies reporting 16-62% unidentified fungal OTUs from diverse environments [27, 38, 39]. While artefactual sequence motifs, resulting from polymerase errors and chimera or heteroduplex formation are known to occur in clone libraries [33, 40], we are confident that the number of such sequences was low in our material because of our prior efforts to optimize PCR conditions . 36 unknown OTUs occurred in several samples in the present material or matched with unknown environmental phylotypes from previous studies. At least, these 36 sequences most probably represent natural phylotypes, because the formation of a unique artefactual PCR product from diverse template pools independently more than once would be highly unlikely. Interestingly, about one fifth of the unknown OTUs were found in indoor samples collected from the same geographic region in our previous study . A novel phylotype related to skin-associated lipophilic yeast genus Malassezia (with 79% sequence similarity to M. sympodiales) detected previously  was prevalent in the present material. Moreover, several clusters of unknown filamentous ascomycetes were found. Some were affiliated with common indoor taxa capable of growing on indoor materials. This suggests that it is possible that building materials may also harbour yet to be identified fungal species.
Besides unknown ascomycetes, Basidiomycetes and yeasts accounted for a substantial part of the unculturable majority of nucITS sequence diversity. These are common in culture-based studies as well, but cannot be routinely identified by morphology [41–43]. While the significance of building-associated filamentous basidiomycetes relates to their wood-decomposing ability rather than human health effects, these fungi may have an indicator value in building investigations, and thus may be important targets for designing molecular diagnostic tools . In the present study, certain building-associated basidiomycetes including Serpula
lacrymans (the causative agent of timber dry rot), Antrodia
versicolor and Gloeophyllum
sepiarium [45, 46], were found, mostly from the water-damaged, wood-framed Index-1 building. These species may have had an intramural source also in the present study. However, this connection could not be verified by examination of the building materials.
Several opportunistically pathogenic taxa  were also identified, including Candida
zeylanoides, Cryptococcus albidus, Exophiala xenobiotica, Mucor spp. and Trichosporon mucoides.
In addition to a wide diversity of fungi, we also found DNA signatures of an impressively diverse array of plants including cultivated crops (fruits, vegetable crops and tobacco), deciduous trees, grasses, mosses and weeds. The amplification of plant DNA was likely due to a lack of specificity in our forward PCR primer . Despite the fact that the inclusion of plant targets was not our intent, their recovery further confirms the biological complexity of dust, and indicates that DNA-based methods may be useful for the detection of dust-borne plant particles. Like fungal particles, those originating from plants may also have allergenic potential, and obviously persist in indoor dust, long past the respective pollen season.
The representativeness of different dust sample types has been discussed in the context of airborne exposure analysis; for example, the presence of heavy, non-resuspending particulate material in floor dusts, as well as potential microbial proliferation in dusts collected from locations with elevated relative humidity have been suspected to bias dustborne measurements [48–50]. A comparison of our above-floor surface samples with floor dust samples collected earlier during the cold season from the same geographic region  indicated differences in fungal community composition. Especially, lower frequencies of basidiomycetous yeasts (mainly Malassezia and Cryptococcus) and rusts were found in dusts collected from elevated surfaces. This difference was also reflected in the differential ratios of Ascomycetes and Basidiomycetes (NAsc:NBas) between the two sample types; the average NAsc:NBas ratio was 3.03 for the elevated surface dust, but lower (0.95) for floor dust. The differences may relate to the aerodynamic properties of different fungal particles; while the spores of the mentioned genera are not distinguishingly large, they are commonly carried along with larger particles (i.e. Malassezia cells on human skin scales and Cryptococcus cells on plant debris), which makes them more prone to deposit on floor surfaces. In contrast, many ascomycetous particles are small, air-dispersed microconidia that stay airborne for long periods, resuspend efficiently and deposit on elevated surfaces. This finding fits into the hypothesis of differential size distribution of particles in the two sample types, and suggests that the small particle fraction may be better represented in elevated surface samples.
Local fungal amplification may have a significant biasing effect on fungal measurements of the dust samples [48, 49]. Our findings suggest that microbial proliferation in settled dust itself had not been extensive in the studied conditions. This was supported by the high molecular diversity coupled with the low dominance of individual OTUs, a strong contribution of species unable to proliferate in indoor habitats and a generally low proportion of Aspergillus, Eurotium and Penicillium (genera known to proliferate efficiently in dust in elevated humidity; ). This dust type seems to act as a sink for fungal propagules arising from various sources, as previously suggested by Scott et al. . These observations may yet hold for temperate regions only; differential observations were made by Amend et al.  from dust samples collected from the tropics with higher relative humidity; there Aspergillus, Eurotium and Wallemia were prevalent, and the overall molecular diversity was lower. The observations by Amend et al.  from temperate regions were similar to ours.
Fungal diversity in building material samples
The spectrum of fungi in building material samples was very different from that observed in dust: Practically all phylotypes were affiliated with filamentous ascomycetes and only a few with basidiomycetes, all of which were yeast-like species. The number of phylotypes observed in material samples was low compared to dust samples. This may have been partly caused by technical problems in the clone library construction; it may also reflect the profound differences of these substrata. While dust acts as a repository of particles, wet building materials support a limited set of taxa, probably as a function of restrictive nutritional characteristics of the substrata and interference competition. The phylogenetic spectrum of fungi observed by sequencing was similar to that observed by cultivation; both methods showed a predominance of taxa affiliated with Dothideomycetes, Eurotiomycetes and Leotiomycetes.
The analyzed building material samples were collected from two moisture-damaged buildings of different construction types. The community composition differed in the two buildings: The Index-1 building was dominated by filamentous xerophilic soil fungi, whereas plant and wood-associated species favouring higher water activity, including yeasts, predominated in samples from the Index-2 building. While others have reported associations between fungal genera and building material types , such separation was not obvious here. Instead, we hypothesize that the predominance of different fungal ecotypes was linked to the sampled building locations: Soil-associated xerophiles tended to dominate the water-damaged ground-level and below-grade sites sampled from the Index-1 building, while phylloplane fungi dominated in roof constructions sampled from the Index-2 building. However, these observations were made from a very limited number of samples, and thus need further testing with larger sample numbers.
Nearly all clones and isolates from building materials could be identified to species level by their nucITS sequences. Most of the fungi detected had been isolated from building materials before [41, 51, 52]. In addition, we identified several species that have not previously been reported as contaminants of building materials (e.g. Penicillium
adeliensis). Moreover, clones and isolates without close sequence relatives in DNA databases were also found. This confirms that the present, largely cultivation-based view of building-associated fungal diversity is incomplete and should be studied in detail using cultivation-independent methods. Advanced isolation techniques using minimal selectivity , as well as novel massively parallel sequencing applications, may offer feasible alternatives to further elucidate this unexplored biodiversity from large numbers of samples.
Effect of moisture damage and remediation on fungal assemblages in dust
We found higher molecular diversity and ERMI scores in dusts collected from damaged buildings than their matched references. In contrast, elevated total concentrations of fungal biomass, total cell counts of common indoor molds or culturable fungi were not seen. Visible water damage and mold growth on surfaces is often associated with elevated concentrations of fungi in dust , but low levels in dust are not uncommon when the growth is located inside the building envelope , as was the case in the present study.
The increased diversities in index buildings were associated with fungal classes that include building inhabiting decomposers (Agaricomycetes) and saprotrophic molds (Dothideomycetes and Eurotiomycetes); elevated ERMI scores suggested an increase in water-associated fungi in index buildings. Despite this, few of the fungi detected from the water-damaged building materials were actually found in the corresponding dust samples, even using the combination of qPCR (a sensitive technique) and clone library sequencing (a non-selective technique). This may indicate that the transfer of DNA containing cell material from the site of growth to the room space was not remarkable compared to other fungal sources. On the other hand, the low number of shared taxa between materials and dust may have been a consequence of undersampling of materials from contaminated building sites and/or the failure to construct clone libraries from individual material samples. We used 69 different qPCR assays to study the fungi in dust, but this selection covered less than one third of the 45 phylotypes found in materials. Thus, it is possible that a larger proportion of the observed fungi in dust was attributable to building material sources than could be verified here.
Remedial and cleaning efforts were associated with a decrease in the diversity of dustborne fungi in one of the buildings. This, as well as the disappearance of certain material-associated species, supports the assumption that remediation was effective in the removal of the fungal burden contributed by indoor mold growth sources. In the second location, clear indications of an intervention effect on the diversity were not seen. Due to a delay in remediation schedules the interval between completion of the remediation and post-remediation sampling was short, which may explain the increase in the abundance of material-associated fungi in post-remediation dust; despite efforts to prevent the spread of contamination, fungal particles aerosolized during remediation may have spread, not being sufficiently removed by post-remedial cleaning. In addition, there was an unexpected diversification in the reference building's microbial profile, which undermined the case-control comparison. The diversification may have been caused by an increase in the transfer of fungal material from outdoors. This hypothesis is supported by the appearance of many probably outdoor-related phylotypes in the clone libraries. Yet the diversification included many species that may proliferate indoors, and thus the occurrence of water damage in the reference building cannot be ruled out. In Location-2, the considerable distance between the index and reference buildings also challenged the comparison. These findings highlight the strong variation in indoor mycobiota within and between buildings, the uniqueness of individual buildings' microbial profiles and the complexity of potential sources. For these reasons, the choice and matching of reference building for each study building is crucial. In general, our findings are only suggestive due to the deep normal variation between buildings and the small building number, and should be further examined with larger data sets.
Comparison of methods
Of all methods tested, clone library analysis provided the most thorough inventory of fungal diversity in settled dust. Nevertheless, a comparison of the sequencing results with qPCR results (a technique with higher analytical sensitivity) showed that many species present in the samples were not represented by the libraries. The species only detected by qPCR tended to be those of lower qPCR cell counts, whereas highly abundant species were much better represented in the clone libraries.
Taking into account the semiquantitative nature of clone library results and the presently deficient species-level information of potential building-associated fungi, the usefulness of clone library sequencing for assessment of building sources remains uncertain. This uncertainty also arises from the universal nature of the technique, i.e. its sensitivity in detecting background diversity acting as a dampening factor on the ability to detect shifts in indicator species. Novel highly parallel sequencing techniques like 454 pyrosequencing overcome the limitations of sensitivity, but the quantitative representativeness remains a problem . In the present study, despite its selectivity, plate cultivation was partly successful in reflecting increased fungal diversity and/or detecting major indicator fungi arising from building material sources in settled dust samples. This was not, however, consistent across all samples, as the masking effect of certain species occurring in very high concentrations was considerable.
ERMI is an index derived from a set of qPCR assays used to describe the indoor fungal burden . Here, the ERMI values were below 5, i.e. relatively low compared to US homes. Vesper et al. reported ERMI values greater than 5 for the highest quartile of randomly selected US homes, whereas over 75% of homes with asthmatic children were above this value . However, no similar data are available in Finland. In the present study, the ERMI index was observed to reflect the overall level of diversity. In our sample material, the group 1 members A. pullulans and Eurotium spp. occurred in significant concentrations in all studied dust samples and in similar concentrations in the index and reference buildings. This suggests that the placement of these species in the indicator group may not be appropriate.