The homes (n = 14) utilized for this study were selected from homes recruited for the “HEPA Intervention Study for Asthma” conducted in Cincinnati, OH. The HEPA intervention study is evaluating the impact of HEPA filtration on the air quality and the asthma of children. The samples referenced in this paper were collected during the placebo treatment of the study. Assessment of conditions in the home such as age and type of building, number of occupants, type of ventilation and indications of moisture damage, were recorded by the research team. Nine of the 14 homes in this study were single family homes, and the remaining were apartments or condominiums. The homes were built between 1865 and 2011. Eight homes were heated with a gas furnace, and six were heated with electric furnaces and all but two utilized air conditioning. Seven of the 14 homes had some type of moisture damage observations (moldy odor, visible mold, or visible moisture damage). Home specific observations can be seen in Table S1.† This study required and received approval from the Institutional Review Board of the University of Cincinnati.

In the laboratory, two Dutch EDCs (Albert Heijn, Zaandam, Netherlands) (each 20.2 × 26.5 cm) were placed in two aluminum trays (each 26.4 × 32.4 cm) cleaned with 70% ethanol.¹⁵ The trays were then placed into a polypropylene bag until arrival at the home. Also in the laboratory, a cardboard box (62 cm × 28 cm) was covered with an aluminum foil sheet and then a Swiffer extra-large cloth (Procter & Gamble, Cincinnati, OH) (45.2 × 25.4 cm). The box was closed and was not opened until arrival in the home. Different types of EDCs have been used in previous studies,^13,16 and we compared these EDCs to ensure that they would provide similar results. The trays were placed in the bedroom at a location where they were least likely to be disturbed,e.g., top of a bookshelf or desk. These EDCs were left to collect the settling dust for four weeks. At the end of the four weeks, the cloths were recovered by folding the sides on top of each other and then the folded EDCs were placed in a sealable plastic bag (Ziplock®, SC Johnson, Racine, WI). The samples were placed in a cooler (4 °C) and returned to the laboratory where they were stored at −20 °C until analyzed.

During the last two days of the four-week period, airborne inhalable particles were collected onto 25 mm diameter, 1 μm pore-size polytetrafluoroethylene (PTFE) filter (Merck Millipore, Billerica, MA) using a Button™ sampler (SKC, Inc., Eighty-four, PA) at 4 l min⁻¹ for 48 hours. Filters collected with the Button™ samplers were placed into a glass-bead filled tube and stored at −20 °C before analysis.

After the sedimentation and air samples were collected, a floor dust sample was obtained in the child’s bedroom by vacuuming 1 m² of a rug for 4 min m⁻² (2 min horizontally and 2 min vertically). For a non-carpeted floor (wood, linoleum, or tile), the sample was collected from the entire room at a rate of 1 min m⁻², as previously described.¹⁷ The vacuuming was performed using a Filter Queen Majestic® (HMI Industries Inc., Seven Hills, Ohio) vacuum cleaner with a high-efficiency particle (HEPA) filter trap (Midwest Filtration, Cincinnati, OH).¹⁸ These traps were recovered and placed in a sealable plastic Ziploc® bag, placed in a cooler (4 °C) and returned to the laboratory where the samples were stored at −20 °C until analyzed.

Immediately after the collection of the vacuum dust sample, a wipe dust sample was collected using a regular-sized dry, Swiffer cloth (26.5 × 20.3 cm) in the child’s bedroom where dust collects, e.g., door frames, bookshelves and window sills. Once the cloth was gray with dust, it was placed in a Ziploc® plastic bag, placed in a cooler (4 °C) and returned to the laboratory where the samples were stored at −20 °C until analyzed.

Temperature and humidity were recorded (HOBO Humidity Data Logger, Onset, Borne, MA) for the entire one-month duration of the EDC sampling.

A minimum of three blanks of each method, e.g., vacuum filter traps, Swiffer wipe cloths, sedimentation Swiffer EDCs, and sedimentation Dutch EDCs, were collected and placed in a Ziploc® plastic bag and stored at −20 °C before analysis. Blank Button™ filters were also placed in glass bead tubes and stored at −20 °C before analysis. Control outdoor air samples were collected using Button™ samplers outside each home during the same time as the indoor air samples to assess outdoor species.

Dust was recovered from the surface wipe sample cloths and passive collection EDCs using a laboratory paddle blender (Seward 80 Stomacher® Lab Blender, Seward Limited, Worthing, UK).¹³ Each surface wipe and EDC was placed into a sterile Stomacher® roll-bag. Two extractions of 30 ml sterile water plus 0.05% Tween 20 were used. The roll-bags were placed in the Stomacher® for 10 min per extraction. The two extracts (total 60 ml) were pelleted by centrifugation (6000 × g, 15 min, 4 °C) and the pellets dried at 30 °C.

Five mg of sieved (300 μm pore size) vacuum dust, 5 mg of surface wipe pellet, 5 mg of sedimentation EDC pellet, or the entire air sampling PTFE filter was added to a 2 ml extraction tube containing 0.3 g of glass beads, as previously described.¹³ Each sample (vacuum, wipe, Dutch, Swiffer and Button) was spiked with 1 × 10⁶ conidia of Geotrichum candidum at the time of extraction as an internal reference to ensure that the extraction and purification were performed correctly.¹⁹ A bead beater (Biospec Products, Bartlesville, OK) was used to shake each extraction tube at 5000 rpm for one minute to release the DNA from the cells. The DNA was then purified using the DNA-EZ extraction kit (GeneRite, Monmouth Junction, NJ), following the manufacturer’s instructions.

The fungal species that form the ERMI^20–22 were quantified with qPCR assays described earlier.²³ The standard qPCR assay contained 1 μl of a mixture of forward and reverse primers at 25 μM each, 12.5 μl of “Universal Master Mix” (Applied Biosystems Inc., Foster City, CA), 2.5 μl of 2 mg ml⁻¹ fraction V bovine serum albumin (Sigma Chemical, St. Louis, MO), 2.5 μl of a 400 nM TaqMan probe (Applied Biosystems Inc., Foster City, CA) and 2.5 μl of DNA free water (Cepheid, Sunnyvale, CA). Five μl of the DNA extract from the sample and this mix were combined. Reactions were performed with thermal cycling conditions consisting of 2 min at 50 °C, 10 minutes at 95 °C, followed by 40 cycles of 15 seconds at 95 °C for template denaturation and 1 minute at 60 °C for probe and primer annealing and primer extension. The ERMI value for each sample was then calculated, as described previously.^20,24

The ERMI metric classifies the 36 indicator mold species into two groups (Table S2†). Group 1 includes the 26 species indicating water damage, and Group 2 includes 10 species which come primarily from outdoors and are commonly found in homes, even without water damage, across the United States.²⁰ The ERMI calculation mathematically converts the results from the concentrations (cells per mg dust) of each of 36 molds into a single number as shown in eqn (1). ERMI=∑i=126log10(s1i)−∑j=110log10(s2j)

Since air samples are in different units (cells per m³) than the dust samples (cells per mg), the term “ERMI-like” was used instead of ERMI for air samples.²⁵

Statistical analysis in R (version 3.1.1) was used to compare the methods in terms of ERMI/ERMI-like values and the total cell concentrations (sum of the cell concentrations of 36 species; cells per mg for Swiffer EDC, Dutch EDC, vacuum and wipe; cells per m³ for Button™ air samples), as well as concentrations of Group 1 species and Group 2 species (sum of the cell concentrations of species in Group 1 and Group 2, respectively). Kruskal–Wallis test by ranks was performed to determine if there was a difference in cell concentrations and ERMI/ERMI-like values between the methods. Spearman’s correlation was performed to determine if there was a correlation between the results obtained with the tested methods. When comparing the species-specific data, the concentration results were normalized to account for the differing units between the air and dust samples. Each sample was normalized to a percentage so that the concentration of each species was divided by the total cell concentration of the 36 species in that sample and multiplied by 100. A multidimensional scaling (MDS) ordination plot was created from normalized sample data to visually demonstrate the relationships within fungal species across the various methods.²⁶ Multidimensional scaling is a method for visualizing proximities or similarities of individual points in the multidimensional data. The idea of MDS is to place each multidimensional data point into two dimensional space in such a way that the distances between points are preserved as well as possible. The corresponding two dimensional points are then easily visualized using two dimensional scatterplots, as opposed to multidimensional plots that are hard to interpret.

Principal component analysis (PCA) was also performed to represent the normalized species data.²⁷ While principle component analysis (PCA) is also a multivariate data analysis method similar to MDS, the main idea of PCA is to reduce the dimensionality of the data by considering a smaller number of important variables (usually two or three), called principle components. In PCA, these principal components are constructed as linear combinations of the original variables in the data in such a way that (1) they are orthogonal or uncorrelated to each other, and (2) they account for as much of the variability in the observed data as possible. These lower dimensional principal components are then visualized in two or three dimensional plots. Finally, also utilizing the normalized data, multivariate analysis of variance (MANOVA) followed by analysis of variance (ANOVA) was performed to determine if there were differences in the species specific data obtained with the five sampling methods. If the multivariate test, MANOVA, found the species within the test were significantly different, ANOVA was performed to determine which species were causing this difference. P-Values less than 0.05 were considered significant for all tests with the exception of the ANOVA. Bonferroni adjustment for multiple comparisons was performed on the ANOVA p-values to determine significance.

The median of the total cell concentrations in the dust samples ranged from 3135 to 8427 cells per mg, with both passive collections yielding the highest concentrations. The median for the Button™ air samples was 44 cells per m³ (Fig. 1). There were no significant differences found in the total cell concentrations (cells per mg) collected with any of the four dust sampling methods. However, there was a significant difference in the total concentration of fungal cells collected in the air samples and four dust sampling methods (Fig. 1, right y-axis). The median ERMI values for the dust collection methods ranged from −0.02 to 1.3, and the median ERMI-like value for the Button™ air sample was 0.6 (Fig. 2). There was no significant difference between the ERMI values calculated for each of the four dust sampling methods or the ERMI-like values calculated from the results of the air samples.

The total cell concentrations obtained with three of the dust sampling methods, Swiffer EDC, Dutch EDC and wipe values, were significantly correlated (range of r = 0.64–0.79, p < 0.05). The strongest correlation for the total cell concentrations was found between the sedimentation Dutch EDCs and wipe sample results (r = 0.79, p < 0.01) (Table 1). The total cell concentrations obtained with vacuum and wipe methods had a significant correlation (r = 0.67, p < 0.05), but a significant correlation was not found between the concentrations of the vacuum and the passive dust collection methods (Dutch EDC or Swiffer EDC).

The ERMI values obtained with all four dust collection methods were significantly correlated (range of r = 0.60–0.93, p < 0.01, Table 1). When evaluating the total cell concentrations of just the 26 Group 1 species, the dust samples all correlated significantly (range of r = 0.68–0.86, p < 0.01, Table 1). When only the 10 Group 2 species were compared, the total concentrations in vacuumed-floor samples were significantly correlated with the wipe samples (r = 0.55, p < 0.05) and the two newly-settled dust samples (Dutch and Swiffer EDC) and wipe samples were also significantly correlated (range of r = 0.62–0.70, p < 0.05). However, Group 2 concentrations were not correlated between the vacuumed-floor dust samples and the newly-settled dust samples (Table 1).

None of the correlations between air and dust results were significant. However, a borderline significant correlation was demonstrated between the ERMI-like Button™ air sampling results and the ERMI Swiffer EDC results (r = 0.52, p = 0.06, Table 1).

Fig. S1† shows each normalized sample taken within all five tested methods plotted separately on a MDS ordination plot. This plot reflects the species similarities between samples and allows for the visualization of the relationships between the different sampling methods by clustering. As expected from the analysis described in Table 1, the dust sample results clustered together regardless of sampling method. However, the air sampling results clustered loosely on the right side of the plot (Fig. S1†). Principal component analysis orthogonal plot shows the same trends as the MDS ordination plot with the dust collection methods and air samples clustering separately (Fig. S2†).

Across all methods, the 10 most abundant species accounted for 90% of the 36 species sampled based on the normalized data. Within each method, the most abundant 10 species accounted for 94% in Swiffer and Dutch results, 92% in vacuum results, 88% in wipe results and 83% in Button™ air sampling results. The concentrations of each of the 10 species for each method can be seen in Table S3.† The normalized data was averaged by method and the percentage of each of the 10 species within each method can be seen in Fig. S2.†

MANOVA followed by ANOVA was performed on the normalized data to study differences in species-specific results between the methods (Table 2). MANOVA analysis showed that there was no significant difference when including the results of all of the 36 species from the dust collection methods. However, the results for the dust sampling and air sampling methods combined were borderline significantly different (p = 0.08, Table 2). Further ANOVA analysis of the results of the 36 species from the dust sampling methods and the air samples revealed Cladosporium cladosporioides Type 1 and Epicoccum nigrum percentages were significantly different between the methods. MANOVA analysis of the 26 Group 1 species yielded no difference between the dust methods with or without air sampling. However, MANOVA analysis showed that the composition of the 10 Group 2 species was significantly different between the sampling methods. Further ANOVA analysis within the dust collection methods revealed that Cladosporium cladosporioides Type 1 was the only species in Group 2 with a significant difference. Further ANOVA analysis within the air and dust sample collection methods revealed Cladosporium cladosporioides Type 1 and Epicoccum nigrum were significantly different between the methods.

The two most abundant and statistically different species were Cladosporium cladosporioides Type 1 and Epicoccum nigrum. Air samples consisted of 27% C. cladosporioides and 13% E. nigrum, and dust samples consisted of 6% to 17% C. cladosporioidesand 45% to 64% E. nigrum (Fig. S3†). For the dust sampling methods, the medians of C. cladosporioides concentrations ranged from 290 to 490 cells per mg and the median in air samples was 4 cells per m³. For Epicoccum nigrum, the median concentrations in the dust samples ranged from 1100 to 2450 cells per mg, and the median value in air samples was 8 cells per m³. Cladosporium cladosporioides and E. nigrum are represented in boxplots for each of the methods in Fig. 3.

The concentrations in the blank Button™ samples ranged 1–25 cells per m³, which comprised ≤1.5% of the averaged Button™ samples (1660 cells per m³). The concentrations in the vacuum blanks ranged from 0–51 cells per mg, which consisted of ≤0.3% of the averaged vacuum samples (14 619 cells per mg), and the respective values of the wipe blanks ranged from 2–9 cells per mg, which consisted of ≤0.1% of the averaged wipe samples (8542 cells per mg). The concentrations Swiffer blanks ranged from 10–284 cells per mg and consisted of ≤2% of the averaged Swiffer samples (13 121 cells per mg), and the values in Dutch blanks ranged from 2–27 cells per mg, which constituted ≤0.2% of the averaged Dutch samples (13 266 cells per mg). All blanks were under 51 cells per mg with the exception of one Swiffer EDC cloth (284 cells per mg) where Epicoccum nigrum and Aspergillus versicolor were the primary contributors to the increased cell count. Outdoor air samples consisted of 33% C. cladosporioidesand 31% E. nigrum. The outdoor Button™ validated the species associated with outdoor sources and showed consistency with the indoor air samples (Fig. S3†). The temperature in the homes collected for the entire month of passive Swiffer and Dutch sample collection ranged from 16 °C to 36 °C with an average of 23 °C and a relative humidity that ranged from 15% to 98% with an average of 45%. Sampling occurred in the span of one year and all seasons were included. Home specific temperature and humidity data can be seen in Table S4.†