Diversity, abundance and activity of ammonia-oxidizing microorganisms in fine particulate matter

Increasing ammonia emissions could exacerbate air pollution caused by fine particulate matter (PM2.5). Therefore, it is of great importance to investigate ammonia oxidation in PM2.5. This study investigated the diversity, abundance and activity of ammonia oxidizing archaea (AOA), ammonia oxidizing bacteria (AOB) and complete ammonia oxidizers (Comammox) in PM2.5 collected in Beijing-Tianjin-Hebei megalopolis, China. Nitrosopumilus subcluster 5.2 was the most dominant AOA. Nitrosospira multiformis and Nitrosomonas aestuarii were the most dominant AOB. Comammox were present in the atmosphere, as revealed by the occurrence of Candidatus Nitrospira inopinata in PM2.5. The average cell numbers of AOA, AOB and Ca. N. inopinata were 2.82 × 104, 4.65 × 103 and 1.15 × 103 cell m−3 air, respectively. The average maximum nitrification rate of PM2.5 was 0.14 μg (NH4+-N) [m3 air·h]−1. AOA might account for most of the ammonia oxidation, followed by Comammox, while AOB were responsible for a small part of ammonia oxidation. Statistical analyses showed that Nitrososphaera subcluster 4.1 was positively correlated with organic carbon concentration, and Nitrosomonas eutropha showed positive correlation with ammonia concentration. Overall, this study expanded our knowledge concerning AOA, AOB and Comammox in PM2.5 and pointed towards an important role of AOA and Comammox in ammonia oxidation in PM2.5.

To investigate the presence of AOA and AOB in PM 2.5 , the specific primer sets targeting AOA amoA gene and AOB 16S rRNA gene were applied for PCR amplification. For all the six samples, desired single bands of PCR products were observed, which were subsequently purified and cloned. Sequencing results further confirmed that the genes were AOA amoA gene and AOB 16S rRNA gene, suggesting the presence of AOA and AOB in PM 2.5 in BTH.
A total of 157 AOA amoA gene sequences were retrieved, and five unique operational taxonomic units (OTUs) were observed at 97% sequence similarity. As shown in Fig. 2a, within each individual clone library, 1-5 OTUs occurred. OTU3 was omnipresent, occurring at all the six PM 2.5 samples (Fig. 2a). The diversity of AOA in PM 2.5 was low based on the diversity indexes (see Supplementary Table S1). Phylogenetic analyses showed that the five OTUs fell into Nitrosopumilus cluster and Nitrososphaera cluster (Fig. 2a). Specifically, the dominant and      OTUs were recovered at 97% sequence similarity, which were higher than those of AOA amoA gene. The diversity indexes further suggest higher diversity of AOB than AOA (see Supplementary Table S1). As depicted in Fig. 2b, there were 2-6 OTUs in each sample. OTU2 was shared by all the samples. OTU4 occurred at five cities and the other OTUs only appeared in one to three samples. Phylogenetic analysis revealed the co-occurrence of Nitrsomonas and Nitrosospira in PM 2.5 . Nine OTUs (65 sequences, 59.63%) were affiliated to Nitrsomonas cluster, including Nitrosomonas aestuarii (17.43%), Nitrosomonas nitrosa (15.60%), Nitrosomonas eutropha (14.68%) and Nitrosomonas oligotropha (11.93%). Two OTUs (44 sequences, 40.37%) fell into Nitrosospira cluster. Especially, the predominant OTU2 was affiliated to Nitrosospira multiformis, accounting for 39.45% of total AOB 16S rRNA sequences. These results suggest that N. multiformis and N. aestuarii were the most dominant AOB in PM 2.5 in BTH.
Abundance of AOA and AOB in PM 2.5 . The qPCR results of AOA and AOB in PM 2.5 are shown in Fig. 3a.
The abundance of AOA ranged from 1.89 × 10 3 ± 3.10 × 10 2 (BD) to 1.14 × 10 5 ± 5.31 × 10 3 cell m −3 air (LF), averaged 2.82 × 10 4 ± 1.94 × 10 3 cell m −3 air. The abundance of AOA was higher in BJ and LF then in the other four cities. The abundance of AOB of PM 2.5 in BTH was in the same order of magnitude, and the average abundance of them was 4.65 × 10 3 ± 4.20 × 10 2 cell m −3 air. The abundance of AOA was higher than that of AOB in BJ, LF and TS with the ratio of AOA to AOB ranging from 1.88 to 22.22 (Fig. 3b), while the abundance of AOA was lower than AOB in the other three cities with the ratio ranging from 0.40 to 0.68.
Moreover, as shown in Fig. 3a, concentrations of PM 2.5 varied greatly (35.42 μ g m −3 in BDH to 194.44 μ g m −3 in TS) at six cities in BTH, which were higher in TJ, TS and BD than BJ, LF and BDH. However, the abundances of AOA, AOB, archaea and bacteria were relatively higher in the latter three cities, suggesting that higher concentrations of PM 2.5 did not mean higher abundance of microorganisms. Cities TS and BD are the industrial urban sites, the emissions of toxic substances into atmosphere by industrial processes might be much higher than those in BJ, LF and BDH, which might be harmful for the survival of microorganisms, resulting in the relative low abundances.
Presence and abundance of Ca. N. inopinata in PM 2.5 . The specific primer set targeting Ca. N. inopinata amoA gene was applied to investigate its presence and abundance in PM 2.5 . The expected single bands of PCR products were observed for each PM 2.5 sample. The PCR products were purified, cloned and sequenced. In total, 63 sequences were obtained, which were further aligned with MEGA 5.0 software, and compared with  Correlations between environmental factors and community and abundance of AOMs. Spearmans' rank correlation coefficients (SRCCs) were calculated to investigate the relationships between environmental factors (Table 1 and Supplementary Table S2 The results of principal components analysis (PCA) and redundancy analysis (RDA) for AOA and Beta-AOB are shown in Fig. 6. As shown in Fig. 6a, the principal component 1 (PC1) and PC2 explained 81.09% and 18.91% of the variance in overall community structure, respectively. Cities BDH, LF and TJ were located close to each other, indicating the similar occurrence of AOA species. However, AOA genera in BJ, BD and TS might be different from others. RDA analysis was carried out to further explore relationships between four factors selected by Monte Carlo permutation tests and the dominant AOA genera (Fig. 6a). The results showed that Nitrososphaera subcluster 4.1 had positive correlations with Na + , PM 2.5 , organic carbon (OC) and element carbon (EC). However, Nitrosopumilus subcluster 5.2 was negatively related with these environmental factors. OC showed significant positive correlation with Nitrososphaera subcluster 4.1.
For Beta-AOB species, PC1and PC2 explained 72.41% and 21.14% of the variance in overall community structure, respectively (Fig. 6b). Three groups of six PM 2.5 samples could be plotted off: the first group contained LF, TJ and TS, the second group contained BJ and BD, and there was only one sample (BDH) in the third group. The selected environmental factors were further used for RDA analysis (Fig. 6b). The results suggest that PM 2.5 , NH 3 , T and NH 4 + showed positive correlations with N. eutropha. NH 4 + was also positively correlated with N. oligotropha. HNO 3 showed significant positive correlation with N. aestuarii and N. nitrosa. RH was positively correlated with Nitrosospria.

Discussion
AOA or AOB are present in particulate matter with different aerodynamic diameters, e.g., PM 2.5 and PM 10 in Beijing 18 , coarse particulate (> 3 μ m) in Germany 17 , TSP in an urban area of Northern Italy 19 , and coarse particular (> 3 μ m) in a metropolitan subway system of New York 20 . This study provided fundamental information regarding AOA, AOB and Comammox in PM 2.5 in BTH. The newly discovered Comammox are widely distributed in a variety of environments, including natural and man-made ecosystems 6,13,14,22 . In this study, the occurrence of Ca. N. inopinata, a Comammox enrichment culture, in PM 2.5 expanded our knowledge of nitrification. The diversity of AOA and AOB was scarce in PM 2.5 , which might be due to the extremely low level of nutrient in the atmosphere 23 . In China, the total ammonia emission was 16.55 Tg for 2005 and keeps an increase trend 24 . Ammonia plays a significant role in the neutralization of acid species to form SIA and PM 2.5 pollution 4 . Ammonia is also the energy source for AOA and AOB, and plays an important role in the niche separations of different species of them 16 . In this study, Nitrosopumilus subcluster 5.2, was found to be the dominant AOA species (96.82%) in PM 2.5 , and Nitrososphaera cluster only accounted for a small percentage (3.18%). Nitrosopumilus maritimus, the cultivated representative of Nitrosopumilus, possesses high affinity to ammonia with low half-saturation constant (K m = 0.133 μ mol l −1 ) 16 . In this study, the low ammonia concentrations in the atmosphere (0.363-0.947 mg m −3 ) might be a reason for the dominance of Nitrosopumilus. Also, Nitrosopumilus is a marine AOA clade 16 , indicating that AOA observed in PM 2.5 might be similar to those in marine. Furthermore, the air movement might make Nitrosopumilus cluster prevail in a broader area, not only in BTH.
The previous studies suggest that Nitrosomonas and Nitrosospira were the main AOB in particulate matters [18][19][20] , which was in agreement with our study. In this study, 59.63% of AOB in PM 2.5 fell into Nitrosomonas cluster and 40.37% of them fell into Nitrosospira cluster. N. multiformis and N. aestuarii were the most dominant AOB in PM 2.5 in BTH. N. multiformis is a commonly used model organism for soil AOB, as it is a representative of Nitrosospira cluster 3, which is widespread in agricultural soils 25,26 . N. aestuarii, one of the marine AOB species, is retrieved from marine environments 27 . These results suggest that AOB observed in PM 2.5 might be similar to those in agricultural soils and marine. Moreover, N. nitrosa (15.60%), N. eutropha (14.68%) and N. oligotropha (11.93%) were also the main AOB in PM 2.5 . Members of N. nitrosa and N. oligotropha exhibit relatively low K m values 28 . Therefore, the low ammonia concentration in the atmosphere is in favor of their survival. N. nitrosa and N. eutropha are also the main AOB in eutrophic freshwater. Members of N. eutropha are common in eutrophic freshwater habitats 27 .
Occurrence of Ca. N. inopinata in PM 2.5 in BTH, suggesting that atmosphere (a substrate-limiting environment) maybe a common habitat for Comammox. Previous studies suggest that Comammox may have an advantage over AOA and AOB under substrate-limiting environments 6  enrich the biofilm samples collected from a pipe in a deep oil exploration well under low ammonia concentration for four years, after which most of the microorganisms within the culture are Comammox Nitrospira bacteria 13 . Overall, the low ammonia concentration in the atmosphere might be the main reason for occurrence and ecological niche distribution of AOMs in PM 2.5 .
Relationships between environmental factors and the communities of AOA and AOB were further investigated by SRCC, PCA and RDA. Some interesting relationships were observed. OC showed significant positive correlation with Nitrososphaera subcluster 4.1, suggesting that some compounds of OC could stimulate their growth. Kim et al. 29 has confirmed the activity of AOA strain DDS1 isolated from seawater can be enhanced by adding α -keto acids (e.g., pyruvate, oxaloacetate). These organic carbon substrates are not assimilated as a carbon source but act as chemical scavengers, suggesting that AOA broadly feature strict autotrophic nutrition 29 . RDA results also indicate that NH 3 showed positive correlation with N. eutropha cluster. Previous studies suggest that high ammonia concentration is in favor of N. eutropha cluster 16,30 , which was in accordance with our study.
The average abundance of AOA and AOB was 2.82 × 10 4 and 4.65 × 10 3 cell m −3 air, respectively. Comparisons of AOA and AOB cell numbers of PM 2.5 with different types of samples (soils, compost samples, activated sludge samples and sediment samples) from previous studies are summarized in Supplementary Table S3. AOA abundance in PM 2.5 was close to the quantities of AOA reported for soils collected from a large geographical scale across North to South China with different pH values 31 , activated sludge treating domestic wastewater 32 and sediments in the hyporheic zone of a eutrophic river in North China 33 . The abundance of AOB in PM 2.5 was comparable with the compost samples collected from the suburb of Changsha, China 34 and the activated sludge samples treating municipal wastewater 35 . The average abundance of AOA was one order of magnitude higher than that of AOB, which was consistent with most of the studies mentioned in Supplementary Table S3. The average abundance of Ca. N. inopinata was 1.15 × 10 3 cell m −3 air, which was in the same order of magnitude with AOB, but one to two orders of magnitude lower than AOA. However, the abundance of total Comammox in PM 2.5 is still unknown. Further investigations are needed to investigate the abundance of Comammox in PM 2.5 with suitable primer set.
Thus far, the actual contributions of AOA, AOB and Comammox to ammonia oxidation in PM 2.5 remain unknown. The maximum nitrification rate (NNR_max), in situ cell-specific ammonia oxidation activity (r in , fmol cell −1 h −1 ) for AOA and AOB, and the abundances of AOA and AOB were used to estimate their relative contributions to ammonia oxidation with formulas (3), (4) and (5) mentioned in methods. Ammonia assimilation of heterotrophic bacteria could remove 10-30% of the ammonia [36][37][38] . However, in this study, the ammonia assimilation of heterotrophic bacteria was not considered in the calculation of relative contributions of AOMs to nitrification because the ammonia assimilation was insignificant in low ammonia concentration environment 39 . On the one hand, in direct nutrient-limited competition, the ammonium turnover per unit biomass of Nitrosopumilus-like AOA would be at least 5 times higher than of oligotrophic heterotrophs 16 . On the other hand, the half-saturation constant (K m ) of Nitrosopumilus cluster for ammonia is much lower than the lowest K m of ammonia assimilation of heterotrophic bacteria 16 .
In the present study, the r in for AOA was set at 0.5 or 208 fmol cell −1 h −1 by referring to the following studies on in situ activity of AOA: 0.5 fmol cell −1 h −1 in freshwater sediment 40 Figure 4b shows the estimations of relative contributions of AOA, AOB and Comammox to ammonia oxidation based on different r in for AOA and AOB. If r in for AOA and AOB were set as 0.5 and 1 fmol cell −1 h −1 or 0.5 and 50 fmol cell −1 h −1 , the relative contributions of AOA and AOB to ammonia oxidation were extremely low, and Comammox accounted for almost 100% of ammonia oxidation for the three PM 2.5 samples. Since the ammonia concentrations in the nitrification potential test were low and AOA have lower K m and higher affinity to ammonia 16 , AOA contribution in the test may be underestimated. If r in values for AOA and AOB were set as 208 and 1 fmol cell −1 h −1 , for three PM 2.5 samples, AOA, AOB and Comammox were responsible for 69.83-93.10%, 0.04-0.05% and 6.85-30.14% of the ammonia oxidation, respectively. If r in values for AOA and AOB were set as 208 and 50 fmol cell −1 h −1 , AOA, AOB and Comammox were responsible for 69.83-93.10%, 1.76-2.35% and 4.55-28.41% of the ammonia oxidation, respectively. These two results suggest that AOA accounted for most of the ammonia oxidation, followed by Comammox, however, the contribution of AOB to ammonia oxidation was low, which might be related to the low ammonia concentration in PM 2.5 incubations and the higher abundance of AOA and Comammox in PM 2.5 . In previous studies, the same calculation methods were used to evaluate the relative contributions of AOA and AOB to the nitrification of activated sludge in full-scale WWTPs 41 and granular activated carbon used in a full-scale advanced drinking water treatment plant 39 . Their results reveal the significant contribution made by AOA to nitrification under low ammonia concentrations, which were consistent with our study. While, their results also suggest AOB play the dominant role of nitrification under higher ammonia concentration conditions 41 . In fact, a more accurate estimation of their contributions to nitrification should only depend on the active AOMs. RNA-based methods or DNA based stable-isotope probing (DNA-SIP) technique may be more effective and accurate to evaluate the contributions of AOMs to nitrification. Further investigation is still needed to validate the hypothesis that AOA played the predominant role in ammonia oxidation of PM 2.5 .
In conclusion, Nitrosopumilus subcluster 5. Comammox may be the major contributors to ammonia oxidation in PM 2.5 . However, further investigations regarding Comammox in PM 2.5 based on an appropriate primer set are still needed.

Sample collection, meteorological conditions and chemical analyses. A model KC-6120 com-
prehensive atmospheric sampler (Laoshan Electronic Instrument Factory, Qingdao, China) was used for the collection of PM 2.5 , NH 3 and HNO 3 samples. The glass fiber filters were pre-heated at 450 °C for 4 h to remove organic material and their weight were measured by a microbalance before PM 2.5 collection. PM 2.5 collection was carried out at a flow rate of 100 l min −1 for 24 h ( Table 1). The collections of NH 3 and HNO 3 were according to the national standard method of the People's Republic of China (GB/T 18204. 25 2000). During summer, the predominant wind direction of BTH is from southeast; therefore, as shown in Fig. 1, the first order of samples collection was from BJ to LF and then to TJ. After sampling from the three cities, samples were collected from the northeast to southwest axis of BTH (namely BDH-TS-BD).
The meteorological data, atmospheric pollutants and air pollution index (AQI) were recorded concurrently with air sampling (Table 1 and Supplementary Table S2). The carbonaceous species (OC and EC) and water-soluble inorganic ions of PM 2.5 were analyzed by a thermal/optical carbon aerosol analyzer (DRI Model 2001A, Desert Research Institute, USA) and ion chromatography (ICS-90, Dionex, USA), respectively (see Supplementary Table S2).
DNA extraction, PCR, cloning and sequencing. For PM 2.5 samples, 1/4 of the whole glass fiber was cut into pieces using sterilized handling instruments. Genomic DNA was extracted using a Fast-DNA ® SPIN Kit following the manufacturer's protocol (Qiagen, CA, USA).
Primer sets Arch-amoAF/Arch-amoAR 46 , CTO189f/CTO645r 47 , amoA-3F/amoB-4R 48 and Nino_amoA_19F/ Nino_amoA_252R 13 were used to amplify AOA amoA gene, Beta-AOB 16S rRNA gene, Gamma-AOB amoA gene and Ca. N. inopinata amoA gene fragments of PM 2.5 samples. The CTO189f was a mixture of CTO189fA/B and CTO189fc at a ratio of 2:1. For AOA amoA gene, the components of PCR mixture and protocols of PCR were followed by the study of Gao et al. 35 . For Beta-AOB 16S rRNA gene and Ca. N. inopinata amoA gene, the PCR protocols were the same as AOA amoA gene except the annealing temperature (58 °C and 60 °C for Beta-AOB and Ca. N. inopinata, respectively). Gradient PCR was applied to detect Gamma-AOB in PM 2.5 . However, the amplification of Gamma-AOB amoA gene was failed. PCR products for the other genes were purified, cloned and sequenced. For each sample, 15-30 white colonies for AOA amoA gene, Beta-AOB 16S rRNA gene and Ca. N. inopinata amoA gene were randomly picked for sequencing with ABI 3730 XL capillary sequencers (PE Applied Biosystems, Foster City, USA).
The AOA amoA gene, Beta-AOB 16S rRNA gene and Ca. N. inopinata amoA gene sequences have been deposited in the GenBank library under accession numbers from KM402456 to KM402612, KY008589 to KY008697 and KX273257 to KX273319, respectively. Phylogenetic analyses. The sequences were grouped into OTUs with a 97% sequence similarity using Mothur 1.28. Cytoscape 2.32 was applied for visualization of the shared OTUs between samples. MEGA 5.0 was used to construct a phylogenetic tree using the neighbor-joining (NJ) method with the Jukes-Cantor correction model. The NJ tree was calculated after bootstrapping with 1000 replicate trees.
Quantification of AOA, AOB, Comammox, bacteria and archaea of PM 2.5 . The abundance of AOA, AOB, Ca. N. inopinata, bacteria and archaea of PM 2.5 were quantified by the following primer sets: GenAOAF/GenAOAF 49 , amoA-1Fmod and GenAOBR 49 , Nino_amoA_19F/Nino_amoA_252R 13 , Uni1055F and 1392R 50 and 934f/1040r 51 on a Stratagene MX3005p thermocycler (Agilent Technologies, USA) in triplicate with a GoTaq ® qPCR Master Mix (Promega, USA). The components of qPCR mixture were the same as the previous study 35 . The qPCR conditions were also followed by this study except different annealing temperature: 56 °C, 58 °C, 60 °C, 53 °C and 59 °C for AOA, AOB, Ca. N. inopinata, bacteria and archaea. The standard curve was generated by using 10-fold serial dilutions of linearized plasmid extracted from the correct insert clones of each target gene. The amplification efficiencies of qPCR assays ranged from 90.6 to 106.0%, and R 2 value for each standard curves exceeded 0.996. Nitrification potential test. In order to investigate the nitrification potential of PM 2.5 , three PM 2.5 samples (PM 2.5 _1, PM 2.5 _2 and PM 2.5 _3) were collected in BJ for 24 h (8 h for each sampling) at a flow rate of 100 l min −1 , resulting a total of 48 m 3 air collected for each sample. A blank filter (as control) was put at the side of the sampler with each PM 2.5 sampling, therefore three blank samples (Blnak_1, Blank_2 and Blank_3) were obtained. After sampling, the PM 2.5 and blank filters were immediately wrapped with aluminum foil and taken back to the lab within five minutes.
Technologies for investigation of microorganisms in the atmosphere are still in immature stage of development, and the methods for evaluation of nitrification activity of AOMs in PM 2.5 have not been reported. The overall microbial community composition in PM 2.5 maybe similar as that in soil because soil is one of the main source for PM 2.5 52 . Soil suspension technique is a recommended method for assessing nitrification potential 53 . In this technique, the substrate and moisture limitations are eliminated, and the changes in AOMs are unlikely to occur after short-term incubation. Therefore, the nitrification rate measured approximates the maximum nitrification rate possible at the specific temperature of the incubation 53 . Moreover, this technique may be the easiest to interpret and most reproducible for all laboratory nitrification assays 53 .
In this study, maximum nitrification rate of PM 2.5 was investigated according to soil suspension technique with some modification, which may be defined as PM 2.5 suspension technique. Briefly, the PM 2.5 and blank filters were cut into pieces and put into 50 ml centrifuge tubes filled with sterilized and oxygenated phosphate-buffered saline (PBS) (g l −1 : NaCl, 8.0; KCl, 0.2; Na 2 HPO 4 , 1.44; KH 2 PO 4 , 0.24; pH, 7.4) followed by vortexing for 30 min and sonication for 2 h to generate suspensions of PM 2.5 and blank samples. The 50 ml suspensions of PM 2.5 and blank samples were incubated individually with 200 ml of inorganic medium in 500 ml Erlenmeyer flasks. The Erlenmeyer flasks were closed with plastic wrap and incubated at 30 °C under agitation at 100 rpm. Compositions of inorganic medium were as follows: 3 ml NH 4 Cl (10 μ g l −1 ), 3 ml NaHCO 3 (20 μ g l −1 ), 0.25 ml trace element and 193.75 ml PBS. The compositions of trace element were according to the previous study 54 . After 0, 2, 4, 6, 8, 10, 12, 14, 16 h of incubation, supernatant was collected and filtered through 0.2 μ m pore size polytetrafluoroethylene membranes. Concentrations of ammonia (NH 4 + -N), nitrite (NO 2 − -N) and nitrate (NO 3 − -N) were analyzed in triplicate in accordance with standard methods 55 .
The net concentrations of NH 4 + -N (NH 4 + -N_net, mg l −1 ) in PM 2.5 sample incubations were calculated using the following formula: Estimation of relative contributions of AOA, AOB and Comammox to ammonia oxidation in PM 2.5 . The estimation of relative contributions of AOA, AOB and Comammox to ammonia oxidation was carried out with some assumptions: (1) only AOA, AOB and Comammox determined were involved in the ammonia oxidation in PM 2.5 ; (2) the ammonia assimilation of heterotrophic bacteria was not considered; (3) there were 1 amoA gene copy per AOA and Comammox, and 2.5 amoA gene copies per AOB 13,15,21 ; (4) all AOA, AOB and Comammox were equally active enough to contribute to ammonia oxidation. Their relative contributions (RC) to ammonia oxidation of PM 2.5 were estimated using the following formulas according to previous studies 36 where RC_ AOA , RC_ AOB and RC_ Comammox represent the relative contribution of AOA, AOB and Comammox. Cell_ AOA and Cell_ AOB are the abundance of AOA and AOB (cells m −3 air), and r in _ AOA and r in _ AOB are the in situ cell-specific ammonia oxidation activity (r in , fmol cell −1 h −1 ) for AOA and AOB. Mr_ N is the relative molecular mass of nitrogen. NNR_max is the maximum nitrification rate (μ g (NH 4 + -N) [m 3 air·h] −1 ).
Statistical analysis. SRCC, PCA and RDA were applied to address the correlations between environmental factors and AOMs (AOA, Beta-AOB and Ca. N. inopinata). The Monte Carlo permutation test (999 replicates) was used to estimate the significance of the correlations. All of the statistical analyses were done using R software version 2.15.