Biodiversity assessment through DNA metabarcoding
Our analysis detected 160 Operational Taxonomic Units (OTUs) with 12,007,712 sequenced reads, 222,370 ± 41,954 (sd) reads per sample, for a total of 54 sequenced samples. The rarefaction curves showed good sequencing effort for the samples (Supplementary Figure S1) which were rarefied to the least count among samples corresponding to 135,443 reads. Twenty OTUs, (7.2% of the total), were assigned to taxa not relevant to our work (mainly to mosses and ferns during the periods October 2014–March 2015 and July–October 2015). From the remaining OTUs, 108 (88% of the reads) were taxonomically assigned to 32 families of vascular plants (68 identified taxa) (Table 2, Supplementary Table S1) and 32 OTUs (4.8% of the reads) remained unidentified either because of low sequence identity and/or query coverage percentage or the absence of any sequence classification result, even when compared to the complete ‘Nucleotide’ Genbank database. The results of the taxonomic assignment to vascular plants are presented in Supplementary Table S1. The OTU sequences were assigned to plant taxa with at least 95% identity and coverage, from which 70% of the OTUs had ≥ 98% sequence identity with the assigned taxa. The positive control of the DNA extraction, Corylus avellana pollen, was correctly identified after HTS. From the 19 negative controls included in the extraction plate, one negative control was selected for sequencing, the only one with sufficient amplicon concentration (2 ng μl−1). In this sample two OTUs were detected (263,649 reads), both assigned to Quercus spp. and contributing < 0.005% to the counts of the rest of the samples. The average amplicon concentration of the three pooled PCR products (before library preparation) was 8.7 ± 4.6 (sd) ng μl−1.
The sampling period July–October 2015 had a relatively lower number of OTUs and identified vascular plant taxa compared to the other periods (March–July 2015 and October 2014–March 2015). When grouped by habitat, the lowest number of OTUs and assigned taxa were found in Larch forests (Table 2). The alpha diversity of the habitats in each sampling period estimated as the number of OTUs per sample, is presented in Supplementary Figure S2. The interaction between sampling period and habitat category significantly influenced the alpha diversity (ANOVA, p < 0.01). Specifically, OTU richness decreased in July–October 2015 for all habitats, while in October 2014–March 2015 Larch forests and Alpine heath showed significantly lower plant biodiversity than Spruce, Beech and Lowland habitats.
Of the 68 identified taxa from the HTS dataset, 42 were woody (trees and shrubs) and 26 herbaceous, 10 of which were graminoids (Supplementary Table S1). Fifteen of these taxa were not present in the plant checklist of the park31 (Supplementary Table S1). There were 13 main pollen taxa contributing at least 0.5% to the total number of the sequence reads (Supplementary Table S1). These were Pinus spp. (36.8%), Larix decidua (14.4%), Cedrus spp. (12.4%), Picea spp. (11.6%), Corylus/Ostrya/Carpinus (5%), Alnus alnobetula (2.9%), Urtica dioica (2.8%), Abies spp. (1.5%), Juniperus communis (0.7%), Beta/Chenopodium/others (0.6%), Taxus baccata (0.6%), Festuca/Trisetum/Lolium (0.5%) and Cupressus sempervirens (0.5%). The relative abundance of the most represented taxa (with a relative contribution at least 0.5% to the sequenced reads (in each period and in total) is presented in Fig. 2. Pinaceae taxa (Pinus, Picea, Cedrus, Abies) were dominant in all three periods, Betulaceae (Corylus/Ostrya/Carpinus, Alnus, Betula), Cupressaceae (Juniperus, Cupressus) and Taxaceae (Taxus) were also abundant in the period October 2014–March 2015 as well as Amaranthaceae (Beta/Chenopodium/others), Betulaceae (Corylus/Ostrya/Carpinus, Alnus, Betula), and Urticaceae (Urtica/Parietaria) in the period July–October 2015. Pinus spp. were the most represented plant taxa in all habitats for the entire 1-year dataset. As regards the other taxa, each habitat showed a distinct plant biodiversity spectrum, while Picea spp. were mostly found in the samples of the Spruce forests (Fig. 2).
Beta diversity as calculated by Jaccard dissimilarity between samples is presented within and between habitats (Table S4, Fig. 3). Jaccard dissimilarity values were lower within habitats than between habitats for all three periods. For the within-habitat dissimilarity the period March–July 2015 had the highest values. Beta diversity was affected by the sampling period (PERMANOVA, p < 0.001, pseudo-F = 11.1, R2 = 0.30) and the habitat (PERMANOVA, p < 0.01, pseudo-F = 1.5, R2 = 0.39). According to the Generalized Linear Model predicting the Jaccard dissimilarity values as function of sampling period, habitat and their interaction (categorical variables), significant differences in beta diversity of different periods and habitats were detected, while also the interaction term of the latter variables had a significant effect on the variation of species composition (Supplementary Table S3). The effect of habitat (categorical variable) to Jaccard dissimilarity values is displayed in Fig. 3 showing higher dissimilarity between habitats than within habitats.
Comparisons between DNA metabarcoding and microscopic analysis
The period March–July 2015, which is the main pollen season for most of the vascular plants of the study, was selected for microscopic analysis. The number of identified taxa is presented in Supplementary Table S4 for each habitat. The mean number in the habitats was 23 ± 4 (sd) taxa, while in total we could identify 50 taxa. The lowest number of identified taxa was recorded in Beech and Spruce high habitats, while for the same period, HTS recorded the lowest numbers in Spruce high and Larch forests. The most represented ones were Pinus spp. (37.9%), Poaceae (20.9%) and Picea spp. (8%) and 19 taxa contributed > 0.5% to the counts (Fig. 2, Supplementary Table S1). Beta diversity, as estimated by Jaccard distance (Supplementary Figure S4), showed to be affected by the habitat type (PERMANOVA, p < 0.05, pseudo-F = 1.45, R2 = 0.37).
In total, DNA metabarcoding detected 68 plant taxa compared to 50 revealed by microscope (Supplementary Table S1). We identified a total of 39 families, 24 of which were identified by both methods (Fig. 4). Within these common families, DNA metabarcoding could distinguish 57 genera or groups of genera compared to 27 identified by the microscopic method (Fig. 4). In general, families not shown in the outputs of one of the two methods (eight and seven absent from the DNA metabarcoding and microscope outputs, respectively) were scarcely represented also by the other. Cyperaceae and Polygonaceae, although found with considerable abundance with conventional light microscopy analysis (0.4% and 0.6%, respectively), did not appear in the HTS results (Supplementary Table S1).
Log-scaled quantification results of pollen grains (microscope) and trnL sequence reads (metabarcoding) were compared between the two methods for the samples of March–July 2015 for each habitat within the altitudinal gradient (Fig. 5). For all habitats, there was significant correlation, i.e. p < 0.01 for Lowland (tau = 0.48) and Beech forest (tau = 0.54); p < 0.001 for Spruce low (tau = 0.59), Spruce high (tau = 0.67), Larch forest (tau = 0.67) and Alpine heath (tau = 0.8), which showed the highest correlation coefficient. Comparing the contribution of each family to the pollen spectra of all habitats, as derived by the two methods, we observe that the microscopic method systematically detected a higher frequency of Poaceae, Betulaceae and Oleaceae pollen than metabarcoding, whereas the latter showed a higher frequency of Pinaceae pollen (Fig. 2, Supplementary Table S1). For both DNA metabarcoding and pollen counts, Pinus was the most abundantly represented taxon (Supplementary Table S1). Notably, Fagus was not present in the HTS results and scarcely represented by the microscopic method. 11.7% of pollen grains were unidentified by the microscopic analysis due to unclear morphology, resulting mainly from the degradation of the pollen cell wall structure. 10.4% of sequence reads were unidentified (no significant match with reference sequences) by metabarcoding in the same period.
Procrustes analysis revealed that there was no significant accordance between distance matrices generated by HTS and microscopic analysis for the period March–July 2015, based on the permutational test (no statistical significance of the Procrustean fit, correlation 0.4, 999 permutations) (Supplementary Figure S4).
Spatio-temporal patterns of pollen taxa
For each sampling period, the results of the HTS data for the identified plant taxa are presented in Supplementary Table S1. The main HTS pollen taxa contributing with at least 0.5% in the total reads of each period are Larix decidua, Pinus spp., Corylus/Ostrya/Carpinus, Picea spp., Cedrus spp., Alnus alnobetula, Cupressaceae (Juniperus communis, Cupressus sempervirens, Juniperus sabina), Taxus baccata, Ulmus glabra and Fraxinus/Olea/Syringa for the period October 2014–March 2015; Pinus spp., Picea spp., Abies spp., Larix decidua and Festuca/Trisetum/Lolium for March–July 2015; and Cedrus spp., Pinus spp., Picea spp., Urticaceae (Urtica dioica and Parietaria judaica), Corylus/Ostrya/Carpinus, Alnus alnobetula, Larix decidua, Beta/Chenopodium/others and Poaceae (Agrostis capillaris, Festuca/Trisetum/Lolium, Deschampsia flexuosa) for July–October 2015. Significantly higher occurrence according to ANOVA was recorded for Cupressus sempervirens (p < 0.001), Juniperus communis (p < 0.01) and Larix decidua (p < 0.001) in the period October 2014–March 2015, for Pinus spp. (p < 0.001) and Abies spp. (p < 0.001) in the period March–July 2015 and for Urtica dioica (p < 0.001), Cedrus spp. (p < 0.01) and Beta/Chenopodium/others (p < 0.05) in the period July–October 2015.
For the 13 main plant taxa, detailed quantitative HTS and microscope data are presented for each habitat during the period March–July 2015 (Supplementary Table S2). Significant spatial patterns were revealed according to ANOVA for Larix decidua with the highest occurrence recorded in the Larch forest (p < 0.05), Alnus alnobetula with the highest occurrence in the Alpine heath (p < 0.05) and Festuca/Trisetum/Lolium with the highest occurrence in the Lowland and Alpine heath (p < 0.05) (Supplementary Table S2). Similarly, for the microscope data spatial patterns were statistically significant for Larix, with the highest occurrence recorded in the Larch forest (p < 0.001), for Alnus, with the highest occurrence in the Alpine heath (p < 0.05), and for Cedrus with the highest occurrence in the low Spruce forest (p < 0.05). Occurrence of Picea was higher in the Spruce and Larch forests and for Poaceae in the Lowland and Alpine habitats, but in both cases differences were not significant, (p < 0.05) (Supplementary Table S2).
According to metabarcoding results, high read numbers of the main plant taxa were recorded when high pollen counts were also detected in airborne pollen data from the aerobiological station of the Park (Supplementary Figure S3). Exceptions are Corylus/Ostrya/Carpinus, Alnus alnobetula and Pinus spp., for which eDNA showed that their pollen was at relatively high concentrations both during their pollen seasons and at different times. In contrast, for Beta/Chenopodium/others and Urtica dioica, plant DNA was detected only in the period July–October 2015, although their pollen season was longer, starting within the period March–July 2015 (Supplementary Figure S3).
The majority of the main pollen taxa (11 of 13 taxa) have representatives in at least one of the habitat types examined31, except for Cupressus sempervirens and Cedrus spp. (Supplementary Figure S3) that do not occur within the Park and are probably transported from surrounding areas; notably, Cedrus is found at all altitudes of all periods. Also Pinus spp., Larix decidua and Picea spp. were retrieved from all samples (Supplementary Table S1).
Source: Ecology - nature.com