Airborne pollen patterns along an altitudinal gradient of the Italian Alps: combination of classical pollen identification methods and next generation sequencing on environmental DNA

With the combination of classical pollen identification methods and next generation sequencing on environmental DNA, airborne pollen patterns were investigated in six habitats along an altitudinal gradient of Pale di San Martino-Paneveggio National Park (Italian Alps). Pollen was identified from environmental samples that were collected with gravimetric Tauber traps. Results of both methods were evaluated to investigate how similarly they can capture pollen spatio-temporal patterns and how they can be used for different applications, such as for diversity assessment and monitoring studies. For the taxonomic identification of pollen DNA, a reference database of DNA sequences was first constructed. ΤrnL sequences were downloaded from Genbank for most of the anemophilous taxa present in the study area (Trentino, northern Italy) and they were processed and stored in a local database. For plant species without available sequences, plant material was collected, the trnL gene was targeted and sequenced and the new sequences were integrated in the final database, in total 1188 sequences corresponding to 403 species of 198 genera and 46 families; from these 44 were new sequences, corresponding to 26 species. Preliminary experiments for metabarcoding analysis οf airborne pollen were designed in order to develop appropriate lab procedures. The experimental design included sample collection from volumetric and gravimetric traps at the Aerobiological Monitoring Centre of Fondazione Edmund Mach, which were utilized both for microscope and molecular analysis. Experiments were performed for the steps of sample processing to retrieve pollen pellets and of DNA extraction of optimal yield. The protocols were applied across a complexity of samples from single-species pollen to environmental multiple- species samples. It was found that a number of factors like the mechanic disruption of the pollen cell walls or the extraction kit used influence significantly the DNA yield. On the basis of the results taken, the optimal combination was selected (Nucleomag kit and disruption with steel beads). Also, it was found that the DNA yield decreases after chemical treatment of the sample, which suggests that better washing of the pollen pellet should be applied instead. A small fragment (about 150 base pairs) of the chloroplast trnL gene was amplified and sequenced (with cloning and classic Sanger sequencing) using universal primers for plants. For the taxonomic identification we used a custom trnL database as reference. The results showed that 75% and 37% of the taxa identified by microscope were revealed by metabarcoding for the volumetric and the gravimetric samplings, respectively. The whole procedure allowed the identification of pollen from environmental samples. The taxonomic information showed higher resolution via metabarcoding, while the amount and quality of information could increase with the application of next generation sequencing. Having defined the optimal methodology, we could proceed to the analysis of the samples from the park (54 in total). Sequencing was performed this time with the Illumina next generation sequencing platform. The sequencing data resulted in a total of 11,137,178 sequences, clustered in 140 Operational Taxonomic Units which were assigned taxonomically to 32 families, 55 genera (or group of genera). According to the plant growth form, 37 taxa represent woody and 25 herbaceous plants. Thirteen of these species (21%) are not present in the plant checklist of the park; Cedrus and Cupressus sempervirens pollen, in particular, had a considerable contribution of >1%. There are 13 main pollen taxa contributing at least 0.5% to the total number of the sequence reads. These are Pinus (36.8%), Larix decidua (14.5%), Cedrus (12.4%), Picea (11.6%), Abies (5.4%), Corylus/Ostrya/ Carpinus (5%), Alnus viridis (2.9%), Urtica dioica (2.8%), Juniperus communis (0.7%), Taxus baccata (0.6%), Chenopodium album (0.6%), Festuca/Trisetum/Lolium (0.5%) and Cupressus sempervirens (0.5%). When we used concurrently metabarcoding and classical microscopic analysis of the pollen trapped, almost all non-rare families were commonly identified by the two methods, but the molecular method could discern more genera. Nevertheless, Cyperaceae and Polygonaceae, although with considerable abundance in the microscopic dataset, did not feature in the metabarcoding results. Compared to the total pollen recorded, Poaceae Betulaceae, Corylaceae and Oleaceae were found to contribute less with the metabarcoding method than with the microscopic one, and Pinaceae more. For the main pollen season, Pinus is the most abundantly represented taxon in the pollen spectrum after both methods and similarly at high concentrations in the aerobiological data (Lanzoni sampler). Nevertheless, its contribution to the park’s vegetation is lower than that of the dominant Picea (85%), the pollen of which consists only 12% of the annual sum. Regarding the biodiversity assessment, metabarcoding could discern the sampling periods. It detected March-July 2015 as the period with the highest number of taxa (alpha diversity), and revealed significant changes in diversity (beta diversity) among sampling periods. These results matched the features of the pollen season, as defined by aerobiological studies running in parallel. Spatial patterns could not be clearly defined; nevertheless, results of metabarcoding were in accordance to the ones obtained with the microscopic method. Optimized molecular protocols can increase our potential for time-efficient analysis of pollen datasets. Providing high resolution taxonomic results, the molecular method that we applied can be used for biodiversity assessments and floral surveys or for monitoring vegetation changes, particularly those expressed in species composition rather than in species abundance. On the basis of our results and previous reports, we can argue that the metabarcoding and the microscopic methods have each their weak and strong points and they should be applied in a complementary way, at least until the quantitative and qualitative issues associated with metabarcoding are adequately addressed.