Our data illustrate that, in the absence of strong seasonal variation in water temperature in the tropics, variation in salinity driven by rainfall becomes a primary controlling variable on V. vulnificus abundance. There is thus a tendency for a rainfall-driven seasonal cycle in V. vulnificus abundance which is inverted from the temperature-driven seasonal cycle at higher latitudes.

Study Site

Sampling took place in the Ala Wai Canal, a 3.1 km long, engineered waterway located on the southern coast of Oʻahu that separates Waikīkī and urban Honolulu (36). A watershed that covers 42.4 km² drains into the Ala Wai Canal via the Mānoa and Pālolo Streams, which merge to form the Mānoa-Pālolo Stream prior to entering the canal, and the Makiki Stream, all of which run through urban areas before reaching the canal. Consequently, the streams are contaminated with a variety of anthropogenic substances and their convergence in the Ala Wai Canal has contributed to its pollution and eutrophication (36, 37). The influx of fresh water from the streams creates a salinity gradient with a typical salt-wedge structure. Tidal flow causes seawater to flow landward on the flood tide and seaward on the ebb tide and remain at depth. The freshwater streams flow seaward on all tides creating a freshened water surface layer estimated to extend to 0.5 m depth on average, but which is highly variable both in salinity and thickness (38). Sediments are continually deposited in the canal at the mouth of the Mānoa-Pālolo Stream causing the build-up of a sill that restricts flushing of deep water in the uppermost section of the canal.

Sampling locations, dates, and times

Sampling of the Ala Wai Canal spanned 13 months beginning March 17, 2008 and concluding on March 10, 2009, covering the nominal dry summer (April–September) and rainy winter (October–March) months. Samples were collected monthly at twelve sites in the Ala Wai Canal numbered (1 and 5–15) by distance from the shallow, upper section of the canal (Site 1) to the Ala Wai Harbor (Site 15). Site 9 was just inside the mouth of Mānoa-Pālolo Stream and Site 12 was at the mouth of Makiki Stream. Missing site numbers 2–4 referred to other samplings at Site 1 that were not used in this study. Sampling at a higher temporal resolution was also conducted in the dry and rainy seasons to assess changes on shorter time scales. Samples were collected weekly at all sites for four weeks from June 26–July 17, 2008 and again for three weeks from February 22–March 10, 2009. Samples were also collected at a reduced number of sites (Sites 5, 9, 12, 14) daily for six days from July 10–15, 2008 and daily for five days from March 2–6, 2009, and once every three hours (trihoral) for twenty-four hours at Sites 5, 9, and 14 from July 15 to July 16, 2008.

Rainfall and streamflow

Rainfall data collected by National Weather Service rain gauges (part of the Hawaiʻi Hydronet System) at 15-minute intervals were retrieved from the online resource (https://www.weather.gov/hfo/hydronet-data). Data from two gauges were selected for analysis. The first was HI-18 (NOAA# MNLH1), which is located near the origin of Mānoa Stream (N21.3161 W157.8142) at an elevation of 150 m in Manoa Valley (“Valley” rainfall). The second is HI-26 (ALOH1), which is located at Aloha Tower (N21.3060 W157.8662) in downtown Honolulu near sea level (15 m) at the coast (“Coastal” rainfall). From these data, average daily rainfall for all sampling months was determined, as well as total rainfall from each 24-hour period prior to sampling. Data on tidal flux were obtained from the National Ocean Service, using tide gauge number 1612340 (https://tidesandcurrents.noaa.gov/noaatidepredictions.html?id=1612340). Stream flow data were obtained from the United States Geological survey for the Mānoa-Pālolo Stream gauge #16247100 (https://waterdata.usgs.gov/usa/nwis/uv?16247100).

Water sample collection and processing

Whole water samples were collected from the top 10–30 cm at all sites in acid-washed bottles with a pole sampler and stored on ice (except for samples used for culturing, which were kept at ~15 °C with cold packs) and transported to the laboratory within three hours of collection. Subsamples (ca. 25 mL) for nutrient analysis (n = 207–211) were frozen and shipped on dry ice to the Oregon State University nutrient analysis facility for determination of dissolved silica, phosphate, nitrate plus nitrite, nitrite, and ammonium concentrations. Nutrient concentrations were measured during every sampling event excluding two weekly sampling events in July 2008 (July 3 and 7).

For particulate carbon (PC) or nitrogen (PN) and chlorophyll a (chl a) measurements, subsamples (25–200 ml) were filtered onto pre-combusted glass-fiber filters (GF/F, Whatman) in duplicate and stored frozen until analysis. For PC and PN (n = 199), filters were pelletized and combusted in a high-temperature combustion CN analyzer, the CE-440 CHN elemental analyzer (Exeter Analytical). Filters for chl a analysis (n = 194) were extracted in 100% acetone at -20°C for 7 days. Fluorescence of extracts and standards were measured using a Turner AU10 fluorometer before and after acidification.

Samples for bacteria counts (n = 219) were fixed with filtered (0.2 µm) formaldehyde (10% w/v final concentration) in a cryovial (Nalgene) and stored at -80 °C. Total bacteria were counted by thawing samples, staining with SYBR Green I, and analyzing on an acoustic focusing flow cytometer (Attune; Thermo Fisher Scientific).

Samples for molecular analysis (n = 243) were pressure filtered via peristaltic pump through 0.22 µm polyethersulfone filter capsule (Sterivex, Millipore), then stored at -80 °C until extracted. Most of the samples (94%) were filtered to the target range of 500–550 mL, but the volume was smaller for 15 samples (100–400 mL) as a function of increasing particulate carbon concentrations which prematurely fouled the filters and drastically reduced flow rate.

Cultivation on vibrio selective medium

For five of the monthly samplings (Mar, Jun, Sep, Dec 2008, and Mar 2009), water samples from every site were assayed for colony counts on a chromogenic, vibrio-selective medium, CHROMagar Vibrio (DRG Intl.). Samples were diluted 10 to 125-fold in sterile peptone water (3% NaCl, 0.15% peptone) and a fixed total volume of 5 mL was filtered immediately through 0.45 µm pore size, mixed cellulose ester filters (47 mm, GN-6; Pall) to achieve plated volumes corresponding to 0.04–0.5 mL of original sample. Filters were placed face-up on the medium and incubated at 37 °C. After overnight incubation (12 to 18 h), blue colonies were enumerated as putative V. vulnificus on plates with most appropriate numbers of colonies (ca. 5–200 colony-forming units; CFUs).

DNA extraction and purification

DNA was extracted from the Sterivex filters using the Masterpure™ Nucleic Acid Extraction Kit (Epicentre). Six-hundred microliters of Masterpure™ Tissue and Cell Lysis Solution containing recommended quantities of proteinase K were added to each Sterivex filter. The ends of the filters were sealed, and the filters incubated on a rotisserie in a hybridization oven at 65 °C for 15 minutes. Fluid was recovered from filter housing by aspiration with a syringe. The filling with buffer, incubation, and buffer recovery steps were repeated twice more and the combined extract from all three rounds was pooled (total volume ca. 1.8 ml). Three-hundred microliters of the pooled extract was processed according to the Masterpure™ Kit guidelines and the remainder was archived. Accounting for all the raw extract volume, total DNA yields ranged from 1–540 µg L^-1 of canal water (geometric mean of 30 µg L^-1). Following initial purification, the resuspended DNA (200 µL TE) was passed through a spin column containing acid-washed polyvinylpolypyrrolidone (PVPP) to remove any residual inhibitors (43). DNA concentration in each sample was quantified fluorometrically (Quant-iT Broad Range DNA kit, Life Technologies) both before and after the PVPP purification step to account for losses incurred during the purification stage (average recovery 60%). Geometric mean concentration of DNA in the final purified extracts was 7 ng µL^-1 (range 0.1–54 ng µL^-1). Extracts were stored at -80° until assayed.

Quantitative PCR

Total V. vulnificus was estimated by TaqMan qPCR targeting the hemolysin gene, vvhA, using previously published primer and probe sequences (Campbell MS, Wright AC. 2003. Real-time PCR analysis of Vibrio vulnificus from oysters. Appl Environ Microbiol 69:7137–7144). Quantification of C-type V. vulnificus used primers and probes targeting the virulence-correlated gene variant, vcgC (Baker-Austin C, Gore A, Oliver JD, Rangdale R, McArthur JV, Lees DN. 2010. Rapid in situ detection of virulent Vibrio vulnificus strains in raw oyster matrices using real-time PCR. Environ Microbiol Rep 2:76–80). Both assays were prepared as 25-µL reactions with 12.5 µL of TaqMan Universal PCR Master Mix (Applied Biosystems), 1.5 µg µl^-1 final concentration of non-acetylated bovine serum albumin (Applied Biosystems) and 0.25–0.9 µM each of the appropriate primers and probe, 2–5 µl of DNA template (equivalent to 0.01 to 2 mL of original sample after accounting for DNA loss and dilution), and water as needed. For vvhA assay, primers were added at 0.9 µM each and the probe at 0.25 µM. For the vcgC assay, primers and probe were each added at 0.5 µM final concentrations. Cycling conditions consisted of initial denaturation at 95 °C for 10 min, then 40 cycles of denaturation at 95 °C for 15 s and annealing/extension at 60 °C for either 60 s (vvhA) or 90s (vcgC). All qPCR reactions were performed in triplicate with DNA template in the final replicate diluted 10-fold to check for inhibition (46) and with additional replication and dilution (up to 50-fold) performed on samples showing inhibition. The amplified PCR product was detected by monitoring the increase in fluorescence signal generated from the 6-carboxyfluorescein-labeled probe using a Realplex² Mastercycler (Eppendorf). Data were analyzed using realplex software (Eppendorf) to determine quantification cycle (C_q) values. Standard curves for both assays consisted of serial 10-fold dilutions (500,000 to 5 genome copies per reaction) of genomic DNA from V. vulnificus strain YJ016 (vvhA+ and vcgC+) assayed in duplicate in each run. Efficiency of amplification based on the standard curves across independent qPCR runs ranged from 97 to 104% (n = 11; vvhA) or 100-110% (n = 11; vcgC). Standard curve intercepts varied little (< 1%) among curves from independent runs of each assay, but differed for vvhA (40.1± 0.3) vs. vcgC (42.3± 0.3).

Reporting limits were based on a maximum cycle number of 38 (equivalent to 4–6 gene copies per reaction) for vvhA or 39 (8–13 gene copies per reaction) for vcgC. This translates into concentration reporting limits for the original sample of 2–374 (vvhA) or 4–220 (vcgC) gene copies mL^-1 depending on initial volume filtered, DNA recovery, and extract dilution factor. At least two (up to four) replicate assays inferred to be free of significant inhibition and above the reporting limit were averaged. Out of 243 total qPCR assays for V. vulnificus abundance (vvhA), seventeen (ca. 7%) had issues that made them unreliable or unavailable (inhibition, below the reporting limit for the assay, or absence of data). In thirteen of these instances, abundances were instead inferred from blue colony counts on CHROMagar Vibrio medium based on the strong correlation (r = 0.8) between log-transformed concentrations of blue colony counts and vvhA gene copy numbers. No corresponding colony counts were available for the remaining samples and they were excluded.

Missing values were replaced with -99.99