Determining the hierarchical order by which the variables of sampling period, dust outbreaks occurrence, and sampling location, can shape the airborne bacterial communities in the Mediterranean basin

An NGS-based taxonomic analysis was carried out on airborne bacteria sampled at ground level in two periods (May and September) and two opposite locations on the North-South axis of the Sardinia Island. Located 20 in a central position of the Mediterranean basin, Sardinia constitutes a suitable outpost to reveal possible immigration of bacterial taxa during transcontinental particle discharge between Africa and Europe. With the aim of verifying relative effects of dust outbreaks, sampling period and sampling site, on the airborne bacterial community composition, we compared air collected during dust-carrying meteorological events to that coming from wind regimes not associated to long-distance particle lifting. Results indicated that: (a) higher microbial 25 diversity and richness (118 vs 65 orders) and increased community evenness were observed in the campaign carried out in September in comparison to the one in May, irrespective of the place of collection and of the presence or absence of dust outbreaks. (b) During the period of standard wind regimes without transcontinental outbreaks a synchronous concerted turnover of bacterial communities across distant locations of the same island, accompanied as mentioned by a parallel rise in bacterial diversity and community evenness appears to have 30 occurred. (c) changes in wind provenance could transiently change community composition in the locality placed on the coast facing the incoming wind, but not in the one located at the opposite side of the island; for this reason the community changes brought from dust outbreaks of African origin are observed only in the sampling station exposed to south; (d) the same winds, once proceeding over land appear to uplift bacteria belonging to a common core already present over the region, which dilute or replace those that were associated with the air coming from 35 the sea or conveyed by the dust particulate, explaining the two prior points. (e) the hierarchy of the variables tested

immigration of bacterial taxa during transcontinental particle discharge between Africa and Europe. With the aim of verifying relative effects of dust outbreaks, sampling period and sampling site, on the airborne bacterial community composition, we compared air collected during dust-carrying meteorological events to that coming from wind regimes not associated to long-distance particle lifting. Results indicated that: (a) higher microbial 25 diversity and richness (118 vs 65 orders) and increased community evenness were observed in the campaign carried out in September in comparison to the one in May, irrespective of the place of collection and of the presence or absence of dust outbreaks. (b) During the period of standard wind regimes without transcontinental outbreaks a synchronous concerted turnover of bacterial communities across distant locations of the same island, accompanied as mentioned by a parallel rise in bacterial diversity and community evenness appears to have 30 occurred. (c) changes in wind provenance could transiently change community composition in the locality placed on the coast facing the incoming wind, but not in the one located at the opposite side of the island; for this reason the community changes brought from dust outbreaks of African origin are observed only in the sampling station exposed to south; (d) the same winds, once proceeding over land appear to uplift bacteria belonging to a common core already present over the region, which dilute or replace those that were associated with the air coming from 35 the sea or conveyed by the dust particulate, explaining the two prior points. (e) the hierarchy of the variables tested 2 in determining bacterial assemblages composition results: sampling period >> ongoing meteorological events > sampling location within the island.

40
With a total volume evaluated of 4.5x10 18 m 3 , terrestrial lower atmosphere represents the most extended potential biome, followed by water, 1.3x10 18 m 3 (Gleick, 1993), and by soil with 6.2x10 16 m 3 (estimated on the basis of the deeper subsurface living bacteria currently described Szewzyk et al., 1993). Concerning atmosphere, microbial cells and propagules, embody a particularly suitable conformation to take advantage of air utilization as an 45 environment for survival and dispersal. Their movement can be favored by a natural mobile reservoir of physical solid carriers represented by the air-dispersed particulate matter. Such particles range between 0.2 and 100 um in size (Bernstein et al. 2004;Ryder et al., 2018) and average loads of 1-100 ug m -3 , Van Dingenen et al. 2004). It has been estimated that more than 5000 Tg of sea salt (Tegen et al. 1997) and  Tg of soil particles, passively uplifting and transporting live cells are released every year in the atmosphere giving 50 rise to a widely heterogeneous material conveyed from different sources (Guang et al. 2009;Mc Tainsh 1989, Knippertz et al. 2009).
The tropical African and Asiatic belts, represent two amongst the major airlift dust sources (Prospero et al. 2002, Schepansky et al. 2007. Several studies underline that this phenomenon strongly contributes to a cosmopolitan microbial distribution (Favet et al. 2013, Griffin 2008, Yang et. al. 2008, Wainwright et al. 2003, Smith et al. 55 2010. Moreover, the correlation between specific bacterial clades and particle size (Polymenakou et al. 2008) opened new hypotheses on differential dispersal of taxa in relation to the dust features. High amount of bacterial 'newcomers' have been pointed out in air samples collected in occasions of foreign dust outbreaks (Maki et al. 2014, Rosselli et al., 2015. Immigrant microorganisms classification (Sànchez de la Campa et al. 2013) and their 60 effects on an autochthonous ecosystem have also been reported (Peter et al. 2014, Shine et al. 2000. Evidences of a correlation between aerosol-related biodiversity and seasons (Gandolfi et al. 2015) underlines the natural complexity related to this process, suggesting that effects may vary also depending on climatic periodicity. The impact of desert dust in transiently changing the airborne microbiota in cities has been described (Mazar et al., 2016). In other works, the origin of dust was compared and storms from different origins exhibited distinct 3 movements has a Northern, Atlantic source in response to the pressure generated by the Azores high (Littmann, 2000). Southern winds from Africa, prone to carry desert sand, and potentially microbes, can be determined by specific climate conditions (Kostopoulou andJones 2007, Benkhalifa et al, 2019). It has been estimated that, as a consequence, 80-120 Tg of dust per year are transported across the Mediterranean towards Europe (d'Almeida 80 1986; Dulac et al. 1996), reaching the higher troposphere layers (Alpert et al. 2004) and spilling over up to the far-Northern sides of the continent (Franzèn et al. 1991).
In order to track the biodiversity of these airways, the Italian island of Sardinia was chosen as ideal observatory point to collect airborne bacteria moving inside and outside Europe. Located in the middle of the Mediterranean Sea, this landmass is separated from Italy, France, Spain and Africa coastal baselines by distances of 120, 150, 85 230, and 100 nautical miles (NM) respectively (Fig. 1). Its geographical position facilitates the displacement of western high-and low-pressure air masses coming from Gibraltar and becoming the first and the last frontier for microbes entering or leaving Europe, respectively. In a prior study (Rosselli et al. 2015), we described a core microbiome in the bacteria cast upon the Sardinia island under different wind regimes through analyses of DNA from deposited particles. The analysis compared the trans-Mediterranean airflow with that of winds from Europe, 90 and pinpointed a number of taxa which have records in clinical infections. In that investigation the sampling dates were all concentrated in a single period of six days (in February) and some variations of the airborne biota were observed in response to the opposite wind directions. However, the most remarkable evidence was a prevailing constancy of the microbial composition in spite of the changing winds provenances. In the present study instead we analyzed a series of events featuring a starting dust outbreak, a 109 days-long period devoid of dust-carrying winds, and a second dust outbreak. The analyses were performed in two oppositely located stations: Cagliari, on the South-East side of Sardinia, facing the African side, 100 and Sassari in the North-West, i.e. farthest from the dust-carrying winds. The sampled particulate was analyzed by NGS sequencing of the amplified 16S rRNA genes. The main goal of the project was to verify in which hierarchical order the different variables of (a) sampling period, (b) occurrence of dust-carrying outbreaks, and (c) sampling location, could act in determining airborne bacterial communities composition.

Meteorological monitoring
Surveillance of the weather trends and conditions to anticipate dust outbreaks from Africa towards Sardinia and 110 winds of interest was performed by routine checking of the MODIS satellite data and Meteosat imagery combined with the SKIRON forecasting model (Nickovic et al. 2001). Europe daily synoptic conditions were analyzed on the weather charts available from the www.eurometeo.com and www.metoffice.gov.uk websites. The origin and the trajectory of the dust carried by winds towards Italy were inferred by the NOAA HYSPLIT model (Hybrid Single Particle Lagrangian Integrated Trajectory Model) (Draxler et al. 2014;Rolph 2014).

115
Monitoring was aimed at predicting two distinct conditions: i) North-African high-pressure nuclei favoring Southern winds suitable to carry and deposit dust over Sardinia (dust-enriched events); and ii) North-European high-pressure nuclei, determining northern winds referred to as 'Controls' (dust-negative events).
In addition, PM10 concentration (particulate matter with a diameter of less than 10 µm) and meteorological data registered by the ARPAS (Regional Environmental Protection Agency of Sardinia) monitoring stations were taken 120 into consideration in relation to the arrival of African air masses.
Information about wind direction and intensity (every 10 minutes), temperature and humidity (once per hour) were downloaded by the ISPRA website (http://www.mareografico.it/) and two sampling stations located in Cagliari (39.21°N,8.40°E). Data covered a 7 months time-lapse, from March to September 2014, in order to obtain a nearly annual view to focus within the main weather instability 125 period.

Sampling
The experimental design involved: 2 sampling sites at the opposite corners of the Sardinia island (Sassari vs. 130 Cagliari), 2 sampling periods (May vs. September) 2 meteorological conditions (absence vs. presence of a dust outbreak). In each of these situations, two replicate samples were taken and processed independently throughout the DNA extraction step to be pooled before sequencing. Samples were collected on Teflon filters (Sartorius Stedim Biotech) by using a Skypost Tecora apparatus (compliant to the European legislation 96/62/gmeCE) processing 39 liters of air per minute. To constitute 'a sample' a continuous 24h-long air intake through the filters 135 was performed.
In the case of the dust outbreaks the 24h sampling was further divided in two periods, by considering independently the first 12 hours and the second 12 hours. The number of resulting samples was therefore 12; namely the module of three conditions: (a) Control; (b) Dust h 1-12, (c) Dust h 12-24; multiplied by 2 sampling periods and by the 2 sampling places, resulting in 3 x 2 x 2= 12 samples.

140
As technical note, the scope and meaning of 'controls' here was that of samples that could allow to individuate the shift between one condition and its adjacent one. In our cases, catching the sudden change of wind regime by sampling immediately before or after a dust outbreak. Therefore, the controls were thus defined as single 24h time points preceding or following the key dust events.
For each sample, date and atmospheric conditions are reported and fully described in the Results chapter and 145 Supplementary Materials.

DNA extraction and Sequencing
DNA was extracted using the E.Z.N.A.® Soil DNA Kit (Omega Bio-Tek Inc.) as described by the manufacturer.

150
Quality and quantity of the extracted nucleic acid were measured using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific Inc.).
Amplification of the 16S-rRNA genes for sequencing was performed using the universal primers 27F-1492R (AGAGTTTGATYMTGGCTCAG and TACGGYTACCTTGTTACGACTT, respectively). PCR was carried out using Platinum® Taq High Fidelity DNA Polymerase (Life Technologies) in a PTC-200 Thermal Cycler (MJ 155 Research Inc.) set as follows: 95°C for 5 min, (95°C for 0.5 min, 51°C for 0.5 min, 72°C for 2 min for 30 cycles), 72°C for 10 min and 4°C on hold. The amplification of the No Template Control (NTC) was negative. Next generation sequencing was carried out at the facilities of the Porto Conte Ricerche Srl (Alghero, Italy). Briefly, amplicons were quality-checked on an agarose gel and purified using the Agencourt® Ampure® XP PCR Purification Kit. One ng of DNA was processed using the Nextera XT DNA Sample Preparation Kit (Illumina 160 Inc.) and sequenced using the HiScanSQ (Illumina Inc.) with 93bp x 2 paired-end reads. Sequences were submitted to the European Nucleotide Archive(ENA) inside the "Dust Metagenome" BioProject with the accession numbers ERX836645-56.

165
Reads were cleaned on the basis of quality and fragments of Nextera adapters removed by Trimmomatic (Bolger et al. 2014) (Zakrzewski et al., 2017). Prior to the analyses, the relative abundances of OTUs were equalized by applying the total sum of squares scaling (TSS) normalization followed by square root transformation. 185 3. RESULTS

Meteorological events
To capture the air microbiota of Sardinia and to put in evidence taxa, which could be associated to specific events, weather forecasts and other data on air circulation were regularly browsed to select suitable dates for the sampling.

190
Such events were intended as winds in northbound direction prone to carry dust from shores across the Mediterranean as opposed to calm air or slow flows from the opposite quadrant. This allowed to integrate boundary conditions and environmental variables to assess possible correlations between these and microbial community fluctuations (Fiamma, 2016).  were from North African origin. About a week earlier instead, September 13 th had featured low pressures on Italy while high pressures were recorded over the southern part of Morocco, Algeria and Mauritania. This picture was not permissive for any transport of air loads from Africa to Sardinia and the day was therefore considered as the 230 "clear day" reference of the period. Air representative of the dust outbreak condition was thence sampled from Sep.19 th through 20 th , while the corresponding control air was collected on September 13 th .

Bacterial community composition
A synoptic view of the results at phylum/class level is shown in Fig. 3. Details on orders are shown in Table 1.
Complete data are available in the Supplementary Datasheet S1 the dust outbreak. The data available from the two 12h sub-periods (h 0-12, and h 12-24) collected during the dust events were merged together in the corresponding pie-charts to be compared at equal sampling duration with the 250 24h control samples. The Sardinia island contour map in white (Control) or black (Dust) and the red dot, pointing at the sampling site, are shown. The squares providing a color legend to the taxa names follow a corresponding clockwise order of their possible presence on each pie-chart.

255
In terms of conserved taxa the core of those observed more regularly included classes as Gammaproteobacteria, Bacilli (Firmicutes phylum) and orders as Actinomycetales (Actinobacteria phylum); for each of these their maxima were seen in the May samples where those reached percentages above 90%, while their minima appeared in the Sassari controls in September with values around 50%. .
The Actinomycetales order was found in all samples; in particular it was featured in the south-facing station 260 (Cagliari), and its numbers tended to double in relation to the dust events. in at least one of the two seasons are reported. These represent the 91.1% of the total sequences for the May sampling (on a total of 65 orders found) and 79.4% of the September sampling (on a total of 118 orders found).
Means ± standard deviation are shown.

290
The numerical effect of the different sampling period on bacterial communities is visible in Tab. 1, comparing in this case the mean relative abundances of the main orders in the two sampling months, grouped independently from site and meteorology events. Among the most evident phenomena confirmed by this single-variable 295 grouping, the September campaign shows the enrichment in the Actinomycetales order and in a number of others that were below detection in the May sampling. In parallel, the diminution of the formerly dominant Enterobacteriales and Pseudomonadales, and the substantial stability of the Bacilli across the compared times were observed.
Besides the comparisons that included all OTUs to put in evidence community variations, in parallel we exploited 300 an additional opportunity to detect possible dust-specific taxa. The rationale was to seek differential enrichment within the dust storm, by dissecting the process, during its progression, splitting its onset from its fully established stage. To this aim, we collected separately the filters of the first 12h of the event, and replaced them with new ones that collected air during the second lapse (hours 12 to 24). Thus the availability of two timeframes, both within the dust event, allowed to verify which OTUs would be incrementally enriched along with the progression 305 of the stormy condition.
This allowed to better refine the bacterial deposition dynamics during the outbreaks. From the visual and physical points of view, an increase in the inflow of air particulate was observed for the 12-24 h period, confirming the differential level of deposition occurring in the maturity stage of the meteorological phenomenon.
This within-outbreaks set up was essentially aiming at individuating OTUs that would display high variation in 310 relation to dust events in comparison to those who would not. The latter were considered to represent the common core of bacteria that were constantly present in samples, irrespective of the changing meteorological events. To apply this distinction, the criterion was to set a cutoff value with respect to the percent of variation occurring between the first 12 h of the collection time and the second half of it. The choice of this threshold was considered critical and, in order to ensure robust conclusions, we deemed necessary to require a considerable consistency of 315 variation. Pointing at this objective, only the taxa which displayed a mean variation higher than ½ of the corresponding standard deviation were taken into account. The resulting level of variation in the two sampling stations is reported in Tab. 2 and the corresponding number of orders is displayed in Tab  12 Table 3. Community richness at order level of taxa occurring during dust events and of those displaying variations higher than half the standard deviation between the first 12h and the second 12h sampling period (selected orders).
The percentage of orders selected upon this criterion over the total of the orders observed in samples collected during the dust events is indicated.

335
As regards the ecological indexes characterizing the communities, species diversity and evenness values were calculated, and results are shown in Tab. 4. The difference that can be appreciated is mainly relative to the series of September samples, in which all had higher values for each of the indexes when compared to the May ones.
Conversely, neither the presence of dust events nor the sampling location appeared to confer relevant differences 340 in this respect.
The patterns of conservation and diversity involving the bacterial communities analyzed were subsequently inspected by multivariate approaches. Principal Component Analysis yielded an output (Fig. 4) that confirms how a separation of communities can be viewed only when considering the temporal factor (Fig.4. A)  Results are shown in Fig. 5. Besides confirming the sampling period as the strongest driver of community change, the analysis further shows that the dust vs. control clustering is acting more efficiently than the Sassari vs, Cagliari sampling site comparison. This allows to draw a hierarchical ranking of the three variables in shaping the bacterial airborne communities, in which, noting also the different scale of the horizontal axis (Discriminant function 1) 375 adopted for the three graphs, the order results : Time >> Meteorology > Geography.
In order to determine which bacterial taxa were mostly accompanying/causing those changes in a statistically significant manner, and to rank their individual importance in this phenomenon, we run an analysis of the differentially featured taxa, testing both an ANOVA variance analysis and a non parametric Wilcoxon Rank test verification of the ranking. The two tools gave coherent scores and the results of the ANOVA output are shown 380 in Supplementary Table S1. A total of 76 taxa were found featuring p values < 0.05, from which, upon applying a stringent Bonferroni-adjusted p value correction, six of those stood above the significance cutoff, and all within minimal false discovery rate values (FDR < 0.005). All of them were cases which were highly reduced in September in comparison to May. The taxa included as the most effective in explaining the differences (p value = 0.000019), the order Oceanospirillales, known as marine oil spill-associated bacteria (Cao et al, 2013), followed 385 14 by known animal parasites as the Coxiellaceae family (Lory, 2014), marine extremophiles as the Thiohalorhabdales (Tian et al., 2017), and three species of Pseudomonas, including the pathogenic P. viridiflava (Hu et al., 1998), the decontamination-associated P. nitritireducens (Wang et al., 2012) and P. alcaligenes which is reported also a human pathogen (Suzuki et al, 2013  September.

425
Two technical aspects to be defined beforehand are the definitions of 'sample' and 'control' that apply to this type of studies. An atmospheric sampling is not to be regarded with the same conceptual metrics that would apply if one were to study water or soil environments, in whose cases, a sample could correspond to one liter or one gram taken respectively from those milieus. In air-filtering the operation is carried out continuously for days and one 430 sample, in our tests, accumulated the content of 56160 liters of continuously changing atmosphere, which takes into account the variations that occur during all those hours, inclusive of the day/night shifts. One sample is therefore not a 'point' but a built-in averaged replication protocol for the chosen window of events.
Second, as regards the definition of control in our study, it is not meant to represent a supposedly stable community that could apply to an hypothetical average dust-free local scenario. As several studies cited in the 435 introduction showed, even in the absence of dust outbreaks, the ambient state of the atmosphere is not stable either. Our goal therefore was not to compare a hypothetical status quo with an altered one. Since the airborne community composition does change daily even during periods that do not feature the dust carrying episodes, there is not a stable condition that could be considered as a durable control community. Even the evening and morning opposite breeze regimes that occur daily in coastal locations impart modifications in local airborne 440 communities. The meaning of the 'control' here was instead meant to catch 1) the first possible timeframe after the stopping of a northbound dusty wind outbreak (it occurred in May) or 2) the latest possible timeframe of a situation before the onset of a dust-carrying change of wind regime (it occurred in September). Thence, in the latter event the control is not intended as a situation of calm that could represent a period of unknown length, but rather the time-zero sample of the dust event itself. While for the former case in May, the control was 445 symmetrically defined as the earliest stage of quiet after the storm.
In the present study the filtered air particulate was analyzed in different seasons and under different wind regimes, using culture-independent DNA sequencing-based approaches targeting the species-diagnostic 16S-rRNA genes from the air-carried bacterial community and an Illumina next generation sequencing platform. Sites were selected also because of their opposite positions facing Africa (Cagliari) or continental Europe (Sassari). The whole 450 analysis was performed within a 7-month time lapse, March to September, chosen also as it offers higher probabilities of weather shifts favoring both northern-and southern-winds (Israelevich et al. 2012). This timeframe proved suitable to the scope as it was possible to exploit two episodes in which dust outbreaks carried by winds of African origin occurred and were preceded and followed by inversions of the air circulation offering control sampling periods with opposite features.

455
Commenting on the taxonomical abundance shifts observed between May and September and trying to interpret the rise of some phyla and the drop of others ( Fig.3; Tab. 1), a preliminary consideration needs to be recalled. The issue has to do with the distinction between actual population dynamics (ecologically-ruled) and mathematical effects of sampling from a 'ballot box' of objects, all of which compete for the constrained 100% format of results (probability-ruled). This caveat was put forward as early as statistics itself was born as a discipline (Pearson, 460 1897). There is in this respect a general problem in comparing communities at different time points (and even more with DNA-based methods with fixed total DNA amount processed), whose results are based on percent values (relative abundance of taxa). In that condition the multiplication of any determines obviously a reduction of the relative abundance of others, when those do not grow at equal or higher rate. Therefore, one given group could have been increasing, but its share in the sum could appear as if it had instead decreased, if a different group 465 17 has increased faster. This consideration, applies inevitably for all metagenomics/metabarcoding surveys, and should be kept in mind for all kinds of interpretations about increases and decreases, which could be either real or apparent (when driven by a stronger change of a different group). As consequence, comparing different sampling points through time is linked to this inevitable constraint: the compositional nature of the datasets binds all relative frequencies to each other (Gloor et. Al., 2017). Therefore, since, as mentioned, the sum of them is 470 bound to give always 100%, the decrease of a given species could be either apparent (driven by the increase of another), or real (due to its actual negative population dynamics). The problem is that the two causes can not be uncoupled by just comparing species frequencies at the two sampling periods. Moreover, as mentioned, even an actual increase of a given species could be masked by the parallel increase of another at a faster pace (or by its net immigration into the scene). For this reason, we consider with caution the possibility of looking at taxa 475 fluctuations as indicative of their actual ecological outcomes. Having clarified that we will therefore limit to comment only the major phenomena that stand out from the comparison. The largest taxa trade-off that is apparent when comparing the two periods is the decline of Proteobacteria and the parallel rise of Actinobacteria. Trying to frame this within possibly seasonal parameters we can consider that the latter are typically relying on profuse spore formation from colonial growth, while the former are non-sporeforming bacteria either motile via flagella 480 or gliding/swarming mechanisms. The basic life forms of the two groups predict therefore that proteobacteria would be more suited by wet seasons and vice versa for the soil-dwelling Actinobacteria. This phylum has been reported by other authors to reach its peaks in fall (Glöckner, et al., 2000). Being also a group of major litter decomposers their rise along with the end of the plants' vegetative season can be seen as compliant with their landscape and ecosystem cycles.

485
A further ecological aspect that can be commented is the higher diversity of the communities during the September sampling in comparison to May, independently from the dust events and from the sampling station location.
This phenomenon, besides the ecological values differences (Tab. 2, Tab. 4, Fig.S9, and Tab. S1) can be also appreciated visually, by comparing the left and the right pie charts in Fig.3, (featuring community composition at order-rank level, and the corresponding cluster analysis based on their relative percentages), and noticing the more 490 complex color-coded pattern of the latter sampling, showing also a consistent similarity of most color sectors presence and proportions. It is not possible from these single-year data to deduce whether such increase could be part of a recurring seasonal phenomenon causing, cyclically, higher species richness after summer periods, or if what we observe could be part of a different pattern of stochastic variability.
Nevertheless, the overall partitions of systematic groups observed in a given sampling time, irrespective of dust 495 outbreaks or sampling corner of Sardinia, share much more similarity within the samples of that period than with any of those collected in the other season. It appears that in general, air collected during dust discharge from a Saharan wind can account for less variation over its reference control sampling than the choice of sampling that site four months apart.
In our prior work (Rosselli et al., 2015) we had studied community composition in the same Sardinian stations in 500 a short period of winter (in late February) during and after a single dust-carrying event. In that study the main feature evidenced was the existence of a conserved core microbiome, encompassing 86-95 % of the taxa, to which the incoming dust would cause some detectable diversity variation but on a rather limited proportional scale. Such minor effect of the dust-lifting storms observed in winter is in fact confirmed in the present work in which the time of the year factor appears as the variable of major order in shaping community structure and richness. In that 505 18 prior analysis of ours the existence of a common core microbiome of the investigated area was one of the suggested evidences. That concept was stemming from the analysis run in February, therefore towards the end of a winter period throughout which Europe experiences its minima in terms of temperature-driven air turbulence events and as consequence receives more limited influxes of air travelling from seas to land. In the present analysis, we observe that, in spite of the major changes brought about by the temporal factor, the two sampling 510 stations at opposite corners of the 270 km-long island shared the closest level of community composition when they were compared at the same time (see CA Ctrl vs. SS Ctrl in Fig.3 Fig.4, Fig. S9, and Fig. 6). Moreover, this similarity was maintained in May even though the two control stations were compared after the dust outbreak.
Foremost, those two distant sites achieved the maximum of community overlap in September, when controls were compared right before the next outbreak, after a 109 days-long period without such events. During that time the 515 air microbiome of the whole area appears to have changed profoundly, but in a concerted fashion, leading to a high uniformity across the island. These data confirm the view of the prevalence of a core microbiome, as emerged in our 2015 report and add the evidence that such extended core community undergoes also a temporally related concerted turnover. Whether or not this could be also a seasonal (regularly recurrent) phenomenon, will have to be demonstrated by further research on this subject.

520
Literature reports have in this sense pointed out differences in airborne microbial composition between seasons; peaks of fungi causing invasive infections in humans were signaled in spring whereas higher proportions of allergenic fungi were observed in fall (Yamamoto et al. 2012).
Consistent with the present data, a higher diversity of both fungal and bacterial airborne cells in late summer and early fall has been observed in United States-based surveys (Bowers et al. 2012¸ Bowers et al., 2013. North-West, with stable intensities. Likewise in the Sassari area (Fig. S13), although some fluctuations in the strength of the westbound winds can be seen, the dominant air motion throughout the period remained the one heading South. In essence these data allow to rule out that the change in community patterns could be due to major 535 air-driven events of taxa immigration from other insular or continental sources.
In addition to the wind orientation and force, data from the two stations regarding temperature and humidity of the same winds can be analyzed ( Supplementary Fig. S11, Fig. S12, Fig. S14, Fig. S15). Humidity values from May to September winds tend to be rather similar, whereas air temperatures increase in line with the summer progression. These data do not account by themselves for events of species enrichment either.

540
Another aspect that can be verified is to compare the two periods in terms of PM10 particulate concentration; these are reported in Supplementary Fig. S1.C (May) and The observed data reveal that, while dust-associated winds can account for some specific limited ingression of taxa, a far more noticeable pattern appears consisting in a successional rise of taxa diversity. It is not yet possible to establish whether this occurrence could be linked to late summer in relation to the climatic conditions of the season. The second part of the summer, especially in the Mediterranean regions, is characterized by prolonged drought alternated to irregular thunderstorms. The income of a thunderstorm is accompanied by convective 550 instability of the atmosphere and this phenomenon has been already pointed out as conducive to the emission and transport of fungal spores plumes (Burch and Levetin, 2002). A possible explanation for a richer pattern of airborne microbes after several weeks of prevailingly dry climate can be sought in the acknowledged fact that those seasonal conditions enhance the daytime height of the planetary boundary layer over Europe and continental US (Seidel et al. 2012), and that the ensuing low pressures foster the turbulence near ground and the overall 555 convection, resulting in a frequent uplift of particles from land surfaces. In addition, it could also be postulated that the dryer and warmer summer conditions can eventually lead to partial cell dehydration in microbes lying at soil or vegetation surface, resulting in lighter cell weights more prone to be advantageously lifted by the local low layers air turbulence.
A further factor that can be hypothesized to have played a role in reducing the diversity of airborne community 560 samples in May, comes from the analysis of the differentially featured taxa between the spring and the fall samplings (Tab. S1) where the strongest statistically significant differences were six taxa that resulted highly enriched in the former period and that, as cited above, included marine bacteria associated to oil spill-related oleovory phenotypes, extremophyles, and potential pathogens. These occurrences can be interpreted as possible clues for a transient event of water pollution around the sampled areas that could have impacted also on the overall 565 airlifted microbial diversity.
In addition to the above, a series of considerations can be drawn upon inspecting the pairwise community difference analysis, whose similarity values are shown in Fig. 6. It also needs to be recalled that, in order to examine the effect of a dust-free period, in May the control (May 27 th ) was sampled after the dust event (May 21 st -22 nd ), while in September the control (Sep 13 th ) was taken before the new dust outbreak (Sep 19 th -20 th ).

570
Therefore, the summer, within which communities could undergo dust-independent changes, is in fact framed between the two control points, chosen as representative of dust-free atmosphere following a wind direction reversal. Within those months there were no dust-carrying wind outbreaks from the African land. This enabled also to verify whether a relatively long period without dust intrusions could have allowed an overall homogenization of the bacterial airborne communities over the island Sardinia. Regarding Fig. 6 the between-seasons comparisons, while the green ones (higher similarities) are all distributed in the two sameperiod comparisons. The first consideration that stems from the global view of these data is once again that the most distant communities are those compared from different seasons (Fig. 6, upper section). It is worth noticing 580 in this respect that no particular difference appears when comparing communities between those collected from the same site (right panel in the upper section) or in the cross-comparison between the two different places.
Moreover, in these samplings from different seasons, the effects of the dust events in comparison to calm air with dust-free wind regimes, is not apparent, being diluted in the major time-related divergence of the communities.

20
When inspecting cases of the same season, the situation in May is representing a comparison picturing the 585 recovery after the dust event, as the control follows the outbreak. One evident aspect in that is how the juxtapositions within the same place, feature the most similar cases (darkest shades of green) with the notable exceptions linked to the dust outbreak in Sassari, which is the North-facing station (notice the two yellow-shaded values). On the contrary, such divergence does not appear at all in the South-facing Cagliari site. One interpretation of this interesting difference is that in the May control, when the air flux reversed after the dust 590 event, the wind blowing from the northern quadrant, was conveying in Sassari air masses that came straight from the sea; while on the opposite corner (Cagliari) instead, the same air had passed over the whole Sardinia. Thus, the Northern collection station received sea-sweeping air, bringing 'fresh' taxa, i.e. not belonging to the Sardinian land-related common bacterial core of the period, while the southern station of Cagliari received instead landsweeping air that had travelled all the way over the island latitudinal extension, and that therefore would have 595 become mixed with the island-related core of biota. Thus, the sea-related entries would bring little contribution to the southern site communities after more than 100 miles of travelling and being diluted through the terra firma atmosphere. This would explain why, in May, shifting from dust outbreak to control in the Southern location, did not bring community divergence as it did in the Northern one.
The imprint of the common Sardinian core on homogenizing communities when the dust preceded the control, is 600 also testified by the left panel (same month of May but different places) resulting in all green shades of medium value, showing that in such situation there was little difference also between different places.
An independent and indirect confirm of this interpretation is given by the situation in September. In that case, the African dust-carrying event was set to be taken after the control; this originated a reversed situation in comparison to the one observed in May; this time the place where the dust-outbreak did not bring particular change was the 605 Northern site, Sassari, as the northbound wind from African origin had supposedly already discharged its load while passing over the land of the island from which, at the same time it would have lifted a vast portion of landrelated common biota. Vice versa, in Cagliari, appreciable changes occurred in relation to the dust arrival, which support the view of air blown over the sea plus dust, as the elements causing changes due to the new kinds of bacteria that hit this side at the frontal south-facing port of entry of the island. The left panel of the section (same 610 season, September, different place) further confirms this as: (a) the strongest drivers of community divergence (yellow to orange colors) are flagged by the two comparisons between the Sassari control and the Cagliari dust situations, and the second of those, in the 12-24 hours window of the dust event is progressively more divergent then that recorded during the first 12 h (0.434 vs. 0.523 similarity value). Moreover, the comparison between the two sites in the September control before dust, shows a rather high similarity (0.705), that is the highest among 615 the September comparisons of different sites, which confirms that, before the dust outbreak, when both localities had experienced a long period devoid such phenomena, the two places had achieved a high degree of uniformity in spite of their distance. In that status, both communities were also profoundly different from their composition in May. A period of over 100 days without intrusions of dust-carrying northbound winds, appears to have accompanied an appreciable concerted change of the air-associated bacteria upon the Sardinian territory.

620
Essentially it appears that when airborne dust has to cross longitudinally the entire large island, it reaches the Northern sampling site (Sassari) less charged with community-changing potential, and/or, that it must have lifted bacteria from of the Sardinian common core, thus causing little variation upon their discharge over a station on the same island. On the contrary, when landing on the south-facing outpost of Cagliari, coming straight form 21 Africa and, until that moment, having travelled over the sea only with air-lifted transcontinental dust, those air 625 masses delivered in the south outpost of Cagliari an appreciably novel community. The geographic position of the sampling sites in relation to the wind origin appears therefore to play a major role in the patterns outcome.
This supports the view that, in case of dust outbreaks, Cagliari, in the south, is at the forefront of changes that are substantially attenuated before they reach Sassari; and vice versa, in case of reversed winds. A distance of >100 miles appears sufficient to absorb and buffer wind-borne taxa immigration in quantitative terms, from either side.

630
In conclusion, the central goal of this study was to assess which variables (sampling period, dust outbreak vs. calm atmosphere, and north-facing vs. south-facing collection site) would be most effective in determining airborne community divergence or homogenization. Data are supportive of time-related turnover phenomena, involving a pattern of diffuse contemporary colonization over large portions of land, whose effect in shaping and homogenizing communities is stronger than the one conferred by occasional transcontinental discharges.

635
These clues entail novel aspects for our better understanding of microbial transport and spread across territories, of the epidemiological patterns for clinically relevant taxa, and can foster the predictive modeling of overall environmental microbiology dynamics.

640
The supplement related to this article is available online at: https……

645
Sequences were submitted to the European Nucleotide Archive(ENA) inside the "Dust Metagenome" BioProject with the accession numbers ERX836645-56.