All samples were collected during the Integrated Ocean Drilling Program Expedition 336 (16 September−16 November 2011) aboard the R/V JOIDES Resolution. Surface water samples used in this study were collected when on station at the North Pond area on the western flank of the Mid-Atlantic Ridge at 22°45′ N and 46°05′ W (Figure 1) during the period of 5 October−5 November 2011. Surface water grabs were collected daily between 1,045 and 1,200 h (local time) from a port/bow location (under the bridge) at a maximum depth of ~0.25 m. For each sample, two 1-ml subsamples were centrifuged at ~13,000 × g for 20 min; after centrifugation, the supernatant fluid was removed and discarded using a micropipette. Boström et al. (2004) reported ~99% recovery using this approach to concentrate bacteria from seawater. Resulting pellets were stored at −80°C for Vibrio qPCR analyses, which we used to estimate population changes at the genus level. Two 10 ml subsamples were fixed in formalin and used for bacteria and virus like particle direct counts (Noble and Fuhrman, 1998).

DNA was extracted from the pelleted seawater samples using the DNeasy Blood and Tissue Kit (Qiagen, Valencia, CA) according to the manufacturer's protocol for Gram-positive bacteria using AE buffer as the elution medium. Two duplicate elutions of 25 μl each were combined for a final eluent volume of 50 μl. Extracts were stored at −80°C until use.

PCR standards were prepared from genomic DNA extracted from V. alginolyticus (American Type Culture Collection strain 33839). Briefly, a culture was grown overnight in LBS (Miller Broth [Fisher, BP1426] supplemented with 2% NaCl and 0.05 M Tris Buffer) then pelleted by centrifugation for 10 min at 8,000 × g. Supernantant fluid was discarded and the genomic DNA (gDNA) in the final pellet was extracted using a DNeasy Blood and Tissue Kit (Qiagen, Valencia, CA) according to the manufacturer's protocol for Gram-negative bacteria. A PCR product was generated from the gDNA using Vibrio group specific primers targeting a variable region of the 16S rRNA gene, 567F, 5′GGCGTAAAGCGCATGCAGGT-3′ and 680R, 5′-GAAATTCTACCCCCCTCTACAG-3, (the PCR cycling conditions were described by Thompson et al. and outlined below; Thompson et al., 2004). The PCR amplicon was cleaned (QIAquick PCR purification kit; Qiagen) and checked by agarose gel electrophoresis to verify it was a single amplicon of the correct size (113 bp). The amplicon was inserted into a PCR-4 vector and cloned into E. coli using a TA-TOPO kit (Life Technologies Grand Isle, NY). The plasmid was extracted (QIAquick Spin Miniprep kit; Qiagen) and the cloned region sequenced to verify the correct insert. The plasmid was linearized with NotI (Roche, Indianapolis, IN) after cleanup (QIAquick PCR purification kit; Qiagen) and was quantified using a Qubit Fluorometer (Thermo Fisher Scientific, Grand Island, NY). The linearized standard was serially diluted in AE and run with each qPCR assay.

Total Vibrio concentrations were quantified using a SYBR green qPCR method utilizing the Vibrio specific primers 567F and 680R (described above) (Thompson et al., 2004); this methods has previously been found to track very closely to culturable Vibrio counts in tropical marine water (Westrich et al., 2016). Briefly, DNA template (5 μl) was added to 10 μl SYBR SELECT PCR Master Mix (Applied Biosystems, Foster City, CA), 4.76 μl of PCR-grade water (Thermo Fisher Scientific), and primers were added to a final concentration of 0.16 μM in a total reaction volume of 20 μl. All reactions were run on a StepOne real-time PCR system (Life Technologies, Grand Isle, NY) with the following cycling conditions: 2 min at 50°C for UDG activation and 95°C for 2 min to activate AmpliTaq polymerase and UP, followed by 40 cycles of 95°C for 3 s for denaturation and 60°C for 30 s for annealing and extension. Each run was followed by a dissociation step (95°C for 15 s and 60°C for 1 min and 95°C for 15 s) to determine a melt curve for analysis of specificity. Each sample was analyzed in triplicate (technical replication) and at least two replicate positive controls and two replicate negative (no template) controls were included with every run. PCR inhibition tests were conducted by the addition of 5 × 10³ gene copies of linearized standard to each sample. If inhibition was detected, the sample was diluted 1:10 and amplification was repeated. Samples that continued to show inhibition were excluded from the analysis. Cycle threshold (C_T) values for all qPCR assays were compared to a standard curve representing 10¹–10⁶ gene copies per reaction volume. The standard curve linear regression (R² = 0.994) resulted in an equation of y = −3.3x + 40.4 with an efficiency of 100.4%. Linearized standards could be detected to a level of 2 copies μl⁻¹, corresponding to a C_T value of 37. Samples were considered not detectable if the mean C_T of technical replicates for each sample was above this value. Vibrio concentrations in seawater were expressed as gene copies ml⁻¹. To give biological context and because copy numbers of 16S rRNA are highly variable across the genus Vibrio, the range of Vibrio genome equivalents (GE) was also determined, by dividing the sample copy number by the currently reported Vibrio 16S rRNA gene copy numbers found in the Ribosomal RNA Database (rrnDB). The rrnDB is an online, curated resource for current ribosomal operon (rrn) copy number information for bacteria (Stoddard et al., 2015). The database currently contains 122 genomes from the genus Vibrio, with a range of 6–14 16S rRNA gene copy numbers per species and a mean of 10 copies per genome. GE ml⁻¹ are reported for the low and high ends of this range, in other words, providing a range of possible cell numbers (genomes) based on 14 copies per genome (to give lower end) and 6 copies per genome (to give higher end).

Total bacteria and virus-like particles (VLP) were enumerated following the protocol first described by Noble and Fuhrman (1997) and modified by Noble and Fuhrman (1998) and Griffin et al. (2001). In short, 10 ml aliquots of 1% formalin fixed surface water (collected daily at ~1,200 h local time) were filtered through Whatman Anodisc 0.02 μm pore-sized, 25 mm diameter membranes, in duplicate. Material was stained by placing the membrane sample-side up on top of a drop of diluted SYBR Gold nucleic-acid stain (Molecular Probes, Eugene, OR) (97.5 μl of 0.02 μm filtered water +2.5 μl of a 1:10 dilution of SYBR Gold) and incubated at room temperature in the dark for 10–15 min. Twenty-seven microliters of Prolong Antifade (Molecular Probes, Eugene, OR) was placed on a coverslip, and the coverslip placed over the filter on a glass slide. Fifteen fields per slide were counted at 1,000 × magnification under oil immersion using a Carl Zeiss, Inc. (Jena, Germany) Axioskop 40 epifluorescent microscope. For negative controls, 10 mL of autoclaved 0.02 μm filter-sterilized deionized water was filtered and stained as described above.

The Navy Aerosol Analysis and Prediction System (NAAPS) surface-level dust and aerosol optical depth (AOD3), a vertical integral of total dust (unitless), was downloaded for the IODP sampling location (22°N and 46°W) over the full sampling time period (http://www.nrlmry.navy.mil/aerosol/; Lynch et al., 2016). The daily analyses (based on 4 6-h intervals) were averaged for each day providing one AOD value per day. This predictive model was used to estimate dust arrival into the larger geographic area at a 1° × 1° spatial resolution for comparative analysis purposes.

To monitor dust arrival and load specifically at the sampling station, a ParticleScan Pro (IQAir, La Mirada, CA) was used shipboard to determine aerosol particle counts daily at 10 min intervals. Measurements were taken on the roof of the bridge of the ship, ~10 meters above sea level. A daily average was calculated.

High-volume membrane filtration samples were collected at 24 h intervals for daily composite aerosol analyses (Haley et al., 2014). The unit flow rate was 0.57 m³ min⁻¹. Collection began at ~0700 h each day using a Staplex TF1A filtration unit (Staplex, Brooklyn, NY) and Whatman TFAGF41 glass fiber filters. Each filter was weighed daily and stored in a labeled zip-top bag at −80° C. Filters were visually inspected for color and other features. In particular, daily filters with an orange tint, indicative of a heavy Saharan dust event (Figure 2) or those with a black to gray tint, indicative of diesel or other smoke contaminant from the ship exhaust, were noted. Local dust arrival was confirmed for specific dates using a conservative approach, which required the following observations to coincide: (1) increased daily aerosol particle counts, (2) orange-tinted filters, and (3) regional NAAPS modeling analysis indicating dust in the region. Any dates with black tinted filters were excluded from analysis because the soot obscured the ability to observe any orange coloration attributed to dust.

Aerosol particulates, bacterial and viral direct counts, and Vibrio gene copy counts were compared for the time series with and without lag factors, using the Pearson correlation analysis, calculating a two-tailed P-value for each analysis. Samples with non-detectable Vibrio levels were assigned a value of one-half of the detection limit for statistical analyses (detection limit was 100 gene copies ml⁻¹). Aerosol particle counts were compared between confirmed dust days (identified both by the presence of orange or red tint on high volume filters and corroborated with NAAPS modeled data) and clear days (no dust and no soot; identified by clear high volume filters and corroborated with NAAPS modeled data) using an unpaired t-test. For all tests, significance was declared when P ≤ 0.05. All tests were conducted in GraphPad Prism 5.0.

Aerosol particle counts were used as a proxy for the presence of Saharan dust particulate matter in the immediate vicinity of sampling when NAAPS modeled data indicated Saharan aerosols in the region. Daily counts during the entire period of sampling averaged 1.89 × 10⁷ particles l⁻¹, with a range of 1.82 × 10⁶ particles l⁻¹ on 5 October to 4.88 × 10⁷ particles l⁻¹ on 22 October. The potential increase of particle counts could also be elevated on slack wind days when ship exhaust was not blown away from the ship and therefore could confound determination of dust arrival. A visual inspection of concurrently collected HV filters was made to determine those days when particle counts were compromised with possible exhaust contamination. Days with notable black particulates and little to no wind (indicating a high likelihood of originating from diesel exhaust from the ship) were noted to occur from 5 to 12 October. Conversely, a notable orange coloration, indicative of local dust intrusion, was observed on HV filters from 14, 15, 17, 21, 22, and 23 October (Figure 2). The high dust load during these dates was also confirmed by NAAPS surface level modeling with a peak dust level in the region indicated on 14–15 October and 23 October (~80 μg m⁻³) (Figure 3). Aerosol particle counts were significantly higher on dust days (3.487 × 10⁷ particles l⁻¹ ± 3.761 × 10⁷ particles l⁻¹ [SEM; n = 6]) than on clear days (1.163 × 10⁷ particles l⁻¹ ± 3.568 × 10⁵ [SEM; n = 4]) (P = 0.0003).

Because of the likely influx of diesel exhaust into the HV filter between 5 and 11 October, the relationship between Vibrio and aerosol particulate counts was examined excluding this period. Two dates, 12 and 13 October, which also showed evidence of exhaust in the filters, but preceded dust arrival into the area (confirmed by NAAPS and low particle counts) were retained as low dust days in this time series (Figures 2, 3). After elimination of samples that exhibited PCR inhibition, or were chronologically isolated (making determination of Vibrio population dynamics difficult to interpret), a total of 14 samples were used for comparative analysis. This provided a 2-week long daily data set of samples collected in the mid-Atlantic 12–25 October 2011, including as noted above, two distinct increases in dust level in the region indicated on 14–15 and 23 October.

A daily average of 3,782 Vibrio 16S rRNA gene copies ml⁻¹ were measured during this time period. Vibrio showed a high correlation with aerosol particle counts with a 24 h lag, to account for a one-day delay in Vibrio growth (r = 0.56; p = 0.04) (Figure 4). There was no significant correlation with particle counts when incorporating 48 and 72 h lag times. Additionally, Vibrio gene copy ml⁻¹ increased with NAAPS dust aerosol optical depth (AOD) (r = 0.58; p = 0.03) (Figure 4) as did genome equivalents. We also examined possible associations between Vibrio and aerosol counts on days with no dust but soot present on filters (e.g., 5–11 October) and found no significant relationship. The highest Vibrio concentration of 9,145 gene copies ml⁻¹ (15 October) was measured 24 h after a spike in aerosol particle counts (3.68 × 10⁷ particles l⁻¹; 14 October) (Figure 4).

Bacteria and VLP direct counts were also analyzed between 12 and 25 October (Figure 5). Total bacterial counts averaged 3.59 × 10⁵ ml⁻¹ and ranged from 2.85 × 10⁵ ml⁻¹ on 14 October to a high on 22 October of 4.60 × 10⁵ ml⁻¹, associated with an influx of dust (NAAPS dust model; Figure 5). VLP were almost an order of magnitude higher than bacterial counts, averaging 2.66 × 10⁶ ml⁻¹, and ranged from 1.66 × 10⁶ ml⁻¹ on 12 October to a high of 3.37 × 10⁶ ml⁻¹ on 23 October, 24 h after the peak bacterial count. Bacterial concentrations increased with aerosol particle counts, with a 24 h lag response (r = 0.76; p = 0.001). VLP concentrations were also significantly correlated with aerosol particle counts, with a 24 h lag (r = 0.37; p = 0.02). VLP also increased with bacterial concentration (r = 0.67; p = 0.01) (Figure 5). Vibrio gene copy numbers were not significantly correlated with either the total bacterial nor VLP direct counts, with or without a time lag.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.