Enumeration of marine microbial organisms by flow cytometry using near‐UV excitation of Hoechst 34580‐stained DNA

A flow cytometry method for enumerating marine heterotrophic bacteria and phytoplankton in a living or preserved sample using a low power solid state near‐ultraviolet laser is described. The method uses Hoechst 34580 to stain DNA in microbial cells in seawater samples. This stain is optimally excited at 375 nm unlike the similar Hoechst 33342, which requires ~ 350 nm excitation only available on more expensive lasers. Phytoplankton abundances from the Hoechst 34580 method are comparable to those of unstained samples and when analyzed by the Hoechst 33342 staining method. With this new method, nonpigmented marine bacteria and phytoplankton abundances are obtained simultaneously in a single sample as the Hoechst emission wavelength (~ 450 nm) is well separated from the emission wavelengths of chlorophyll and phycoerythrin fluorescence. Bacteria abundances are similar between this new method and those obtained with established Hoechst 33342 and SybrGreen I methods. Precision estimates (coefficient of variation) on populations with abundances near ~ 105 cells mL−1 are 1–3%, increasing to 3–9% at lower cell concentrations of 103 cells mL−1. The Hoechst 34580 method is simple, requiring no heating or pretreatment with RNAse, can be used on unpreserved and formaldehyde‐preserved cells, and is amenable to at‐sea use with portable, compact, low power‐requiring flow cytometers.

Marine microbial population abundances have been estimated using flow cytometry (FCM) for the last~40 yr (Olson et al. 1983;Yentsch et al. 1984). This technique uses the intrinsic properties of cells and fluorescent probes to detect particles in a fluid stream that are subsequently separated by their properties into microbial groups. Phytoplankton and nonpigmented marine bacteria have been enumerated using FCM in habitats ranging from the open ocean to coastal areas, as well as in laboratory cultures (Veldhuis and Kraay 2000;Marie et al. 2005).
Flow cytometers used in marine research have the capability to detect phytoplankton chlorophyll fluorescence, usually using lasers with 488 nm excitation and detectors configured to gather emitted light at~680 nm (Olson et al. 1983). One notable oceanographic discovery made with FCM was the detection of Prochlorococcus, which are cyanobacteria found in the euphotic zone of marine waters and of great importance to overall primary production (Chisholm et al. 1988;Partensky et al. 1999). Phycoerythrin, and phycobiliproteins in general, are naturally occurring pigments that are easily detected in FCM samples, and allow the enumeration of Synechococcus, cyanobacteria commonly found in coastal and open-ocean waters, as well as its separation from Prochlorococcus, as phycoerythrin expression is usually absent or minimal from the latter (Hess et al. 1996). Nonpigmented bacteria are usually detected with DNA dyes; this staining also facilitates the separation of living from dead cells or inorganic particles (Gasol and Del Giorgio 2000).
Most DNA stains used in oceanography are excited with a 488 nm laser, and often emit at green wavelengths (~525 nm, e.g., SybrGreen I; Marie et al. 1997), although stains emitting at longer wavelengths may be used, such as propidium iodide (Taylor and Milthorpe 1980). For samples containing phytoplankton, however, staining of DNA which results in fluorescence emission overlapping that of phycoerythrin or chlorophyll is of limited use. Probes optimally excited by UV light (~300 to < 400 nm) with emission (< 500 nm) are better separated from natural pigments than those emitting in green wavelengths. For instance, Hoechst 33342 binds to the minor groove of ds-DNA preferring AT-rich sequences (Portugal and Waring 1988) and is best excited by wavelengths near 350 nm, emitting at~450 nm (Arndt-Jovin and Jovin 1977;Shapiro 1981;Monger and Landry 1993;Bucevičius et al. 2018). This probe has been used extensively to characterize phytoplankton and marine bacteria in many open ocean habitats (Campbell et al. 1994;Binder et al. 1996;Landry et al., 2008Landry et al., , 2011Taylor and Landry 2018). However, this approach has been limited by the expense of flow cytometers with ultraviolet (UV) lasers emitting~350 nm required to optimally excite this probe.
Described here is the use of Hoechst 34580, which is excitable by a near-UV (375 nm) or violet (405 nm) laser, with a stain excitation maximum of 368 nm, and emission maximum of 437 nm. These laser excitation wavelengths are achieved with relatively inexpensive solid-state lasers (Shapiro and Pellmutter 2001) found on commercially available instruments (Shapiro and Telford 2018). I will describe the use of this probe in detecting marine bacteria and phytoplankton in a single sample, saving time and sample volume. Presented here are data comparing microbial population abundances in preserved and unpreserved samples, as well comparing samples stained with Hoechst 34580, Hoechst 33342, and SybrGreen I.

Materials and procedures
Samples were collected from a variety of locations for method validation and comparisons (Table 1). Open-ocean samples were collected from six depths in the euphotic zone (5-125 m) at the Hawaii Ocean Time-series Stations ALOHA and Ka'ena during HOT 312 (KM19-09, June 2019). Depth profile samples were also obtained from the Gulf of Mexico, aboard the NOAA Ship Nancy Foster during May 2017, designated as NF17 samples (5-115 m, further details found in Selph et al. 2021). In addition, surface samples were obtained on 18 December 2020 from Waim analo Beach (Oahu, Hawaii).
Samples (2 mL) were preserved within 1 h of collection with paraformaldehyde (0.5% final concentration), then flash frozen in LN 2 and stored at À80 C until batch analysis. On the run date, samples were thawed, and 275 μL aliquots were stained with 7 μL of 40 μL mL À1 Hoechst 34580 (1 μg mL À1 , final concentration) for 1 h in the dark at room temperature. Hoechst is potentially mutagenic and carcinogenic, so gloves should be worn as protection; also it may bind to plastic so staining should be in glass tubes. For stain stability tests, samples were thawed, stained (as above), then analyzed starting at 1 min poststaining and at frequent intervals till~6 h, stored in the dark in the refrigerator (4 C), and then analyzed again at~24 h.
Separate sample aliquots were stained with 1X SybrGreen I using the method of Marie et al. (1997) without RNAsetreatment or heating, for comparison of nonpigmented bacteria abundances relative to the Hoechst-stained samples. Briefly, 5 μL of 1% SybrGreen I solution (2 μL of 10,000X SybrGreen I diluted with 200 μL of distilled water) was added to 45 μL of 300 mM potassium citrate and 500 μL of sample. Another aliquot of sample was analyzed without staining, using intrinsic pigment fluorescence (i.e., chlorophyll and phycoerythrin) to detect phytoplankton cells. Finally, since the Hoechst molecule is small enough to enter unpreserved cells, freshly collected surface samples from Waimanalo Beach were stained with Hoechst 34580 (as above) or Hoechst 34580 plus trifluoperazine dihydrochloride (TFD; 15 μM final concentration).
A Beckman-Coulter CytoFLEX S flow cytometer was used to analyze all samples. This instrument is equipped with four lasers emitting light in spatially separated pathways at 375 nm (60 mW), 405 nm (80 mW), 488 nm (50 mW), and 561 nm (30 mW). Acquisition software was CytExpert (v. 2.3.1.22), generating listmode (FCS 3.0) files. Samples (115 μL) were delivered to the instrument from a 96-well plate and analyzed at 30 μL min À1 , with the exception of stain stability test samples which were run in single tube mode at the same acquisition rate. Daily flow rate calibrations (by weight) were done to ensure accurate delivery volume. For Hoechst 34580-stained samples, signal discrimination (with logic as "OR" rather than "AND") was on chlorophyll (EX 488 nm, EM 690 AE 50 nm, gain 500, discrimination 310 area) or DNA (EX 375 nm, EM 450 AE 45 nm, gain 500, discrimination 1950 area). For SybrGreen I-stained samples, the DNA discrimination was on green fluorescence (EX 488 nm, EM 525 AE 40 nm, gain 300, threshold 2000 area). Synechococcus was excited for both sample types using the yellow laser (EX 561 nm, EM 585 AE 42 nm, gain 500). Forward scatter gain was 100, while right-angle (90 ) light scatter was collected from both the 405 and 488 nm lasers (gains 220 and 100, respectively). For unstained samples, signal discrimination was on chlorophyll fluorescence (EX 488 nm, EM 690 AE 50 nm, gain 500, discrimination 310 area). 0.5 μm-UV and YG beads (Polysciences) were used to normalize the fluorescence and scatter signals (e.g., mean fluorescence/bead fluorescence = normalized signal). I also present data comparing nonpigmented bacteria in samples collected from off the coast of California (CCE-LTER, cruise P1908, August 2019; Table 1), preserved (as above), LN 2 frozen, stored at À80 C, thawed in batches, and then divided into two aliquots. One aliquot was stained with Hoechst 34580 (as above), whereas the other aliquot was stained with Hoechst 33342 (1 μg mL À1 final concentration). The Hoechst 34580 aliquot was analyzed on the CytoFLEX S (as above). The Hoechst 33342 aliquot was analyzed on a Beckman Coulter EPICS Altra flow cytometer, equipped with dual beam colinear excitation by two Coherent I-90 argon gas, watercooled lasers, emitting at~350 nm (200 mW) and 488 nm (1 W), with quantitative sample delivery controlled by a syringe pump (50 μL min À1 ). Chlorophyll (680 nm), phycoerythrin (575 nm), and DNA (450 nm) fluorescence, as well as forward and 90 light scatter signals, were collected (more details in Selph et al. 2011).
Linear regressions for data comparisons were reduced major axis regressions, calculated using RMA for Java v.1.21 (Bohonak and van der Linde 2004). This program also calculates 95% confidence intervals using Jackknife estimates. Difference plots (values from one method subtracted from the other method, plotted as a function of the mean of the two values) were used to determine if systematic bias was inherent in the data, and to examine if data fell within the 95% confidence limits of agreement (Hollis 1996).

Population definitions
Hoechst 34580-stained sample listmode files were analyzed using a series of two-parameter fluorescence and scatter plots to separate each microbial population using their signatures (Fig. 1). Shown here to illustrate this procedure is a 5 m station ALOHA-preserved sample. First, Hoechst 34580-bound DNA (discrimination parameter, thus including all events detected) was plotted as a function of acquisition time, which allows assessment of sample delivery rate, a key parameter for quantitative sample analysis (Fig. 1a). The CytoFLEX S flow cytometer has an uneven sample flow rate during the first ~30 s of sample delivery, so this was excluded in subsequent population analyses, resulting in a sample file representing 100 μL of sample (i.e., "included" particles).
Phycoerythrin fluorescence as a function of DNA fluorescence was plotted for the included particle population, and subdivided into those particles with (e.g., Synechococcus) or without phycoerythrin fluorescence (Fig. 1b). It is critical for quantitative purposes to include all ≤ 0 events on the phycoerythrin axis for downstream analyses or non-phycoerythrincontaining particles will be excluded completely from further consideration. Next, chlorophyll and DNA fluorescence are used to separate heterotrophic bacteria, Prochlorococcus and photosynthetic eukaryotes from each other (Fig. 1c). The Prochlorococcus population DNA fluorescence magnitude is within the range found for heterotrophic bacteria, as is expected since all are prokaryotes. However, Prochlorococcus has chlorophyll fluorescence, so it is separable from heterotrophic bacteria by that parameter. Photosynthetic eukaryotes have more chlorophyll and DNA fluorescence than Prochlorococcus.
Further refinement of the Synechococcus population was achieved by plotting chlorophyll fluorescence as a function of phycoerythrin fluorescence (data not shown). Typically, Synechococcus shows less DNA fluorescence with Hoechst than Prochlorococcus or heterotrophic bacteria despite having a somewhat larger genome size. This may be due to Synechococcus' lower A-T content (Zhaxybayeva et al. 2009), which is the target of the Hoechst molecule. Photosynthetic eukaryote populations were refined to omit machine or other particle noise by plotting chlorophyll fluorescence as a function of 90 light scatter, since they will have both higher light scatter (i.e., they are larger) and chlorophyll fluorescence (data not shown).
The heterotrophic (nonpigmented) bacteria population was refined (from the subset indicated in Fig. 1c) by plotting DNA fluorescence as a function of 90 light scatter (Fig. 1d). Particles with very high light scatter (e.g., > 10 4 arbitrary scatter units in Fig. 1d), but no higher DNA fluorescence than the bulk of the cells, were omitted, as were particles with very low light scatter and low DNA fluorescence. Prochlorococcus was further defined using chlorophyll fluorescence as a function of 90 light scatter, excluding large and small particles that do not have appropriate chlorophyll fluorescence (Fig. 1e). Samples collected deeper in the water column typically have more chlorophyll per cell; this is particularly apparent for Prochlorococcus (Fig. 1f).
Notably, DNA signals from Hoechst 34580-stained cells showed little difference whether excited with a violet (405 nm) laser or a 375 nm laser (data not shown) using the CytoFLEX S cytometer, despite the expectation that more DNA signal should be obtained from 375 nm excitation based on the inherent Hoechst 34580 excitation-emission properties (Shapiro and Permutter 2001). However, violet-emitting lasers may not always emit at the expected wavelength (Shapiro and Telford 2018), so Hoechst 34580 signals may vary depending upon the instrument laser specifications and their response should be verified.
A second unstained aliquot of the preserved 5 m station ALOHA sample was analyzed within 1 h of the Hoechststained aliquot (Fig. 2). Similar to the stained sample, a plot of chlorophyll fluorescence (discrimination parameter) as a function of time was used to omit the first 30 s of uneven sample flow rate (Fig. 2a). Then, chlorophyll fluorescence as a function of phycoerythrin was used to separate Synechococcus from Prochlorococcus and photosynthetic eukaryotes (Fig.2b). The Synechococcus population was further refined by plotting phycoerythrin fluorescence as a function of 90 light scatter (data not shown), while photosynthetic eukaryotes and Prochlorococcus were delineated separately with chlorophyll fluorescence as a function of 90 light scatter (data only shown for Prochlorococcus, Fig. 2c).

Precision and stability tests
To estimate the precision of the Hoechst 34580 method, replicate (10) samples of a single preserved aliquot from HOT   Fig 2. Microbial population designations for a second aliquot of the same 5 m sample from station ALOHA of HOT 312 as the sample shown in Fig. 1a-e, however this aliquot was not stained. SYN, Synechococcus; PRO, Prochlorococcus; PEUK, photosynthetic eukaryotes. Prochlorococcus, coefficients of variation (CVs) on the average cell abundance were 1%-3% at the three depths compared (5, 75, and 125 m). Similarly, nonpigmented, or heterotrophic bacteria had very low CVs (1-3%). In the 100 μL aliquot analyzed for both of these populations, at least 1000 and usually more than 10,000 cells were enumerated. In contrast, for the photosynthetic eukaryotes and Synechococcus, the 5 and 75 m samples had CVs between 3% and 9%, whereas the deepest sample (125 m) had very high CVs (22% for photosynthetic eukaryotes and 45% for Synechococcus). These very high CVs are likely due to these populations' low abundances, i.e., 10s or 100s of cells in a 100 μL aliquot. For lower variances, a higher volume of sample can be analyzed at the same rate (30 μL min À1 ) for a longer time. However, increasing the sample flow rate to analyze more volume must be done with extreme caution in any system with hydrodynamic focusing, as this broadens the sample stream and may compromise single-cell detection.
To determine how stable the DNA fluorescence was from Hoechst 34580, preserved station ALOHA samples (three casts, with samples from 45 to 100 m) were stained and analyzed in a time course, starting from the initial staining time (~1 min), then re-analyzed at intervals over a 24-h period. Prochlorococcus abundance time points averaged from 15 min to 6 h were within the precision of repeated measurements (2-3% CV), except for two of the 45 m samples where repeated measurements had CVs of 4-5% (Fig. 3a). In contrast, the average nonpigmented bacteria abundances were within precision estimates (1-2% CV) for all samples from 15 min to 6 h (Fig. 3b). At 24 h after staining, abundance measurements were lower for all samples, but still 83-94% of the mean value for Prochlorococcus and 94-98% of the mean value for nonpigmented bacteria (Fig. 3a,b, respectively).
DNA fluorescence in Prochlorococcus and nonpigmented bacteria increased after staining from 1 to 15 min, and was relatively stable after 30 min of staining, varying little over 24 h (Fig. 4a,b, respectively). Chlorophyll fluorescence in Prochlorococcus was highest initially, but decreased slightly over the first 2 h, stabilizing after that, with the 24-h value still 73-85% of the initial value. One exception was a 45 m sample (1a, Fig. 4c), which increased in chlorophyll Table 2. Precision estimate of Hoechst 34580-staining method demonstrated with HOT 312 station ALOHA samples from 5, 75, and 125 m depth. Separate aliquots of each preserved sample (n = 10) were analyzed after staining with Hoechst 34580. Populations enumerated are Prochlorococcus (PRO), Synechococcus (SYN), photosynthetic eukaryotes (PEUK), and nonpigmented (heterotrophic) bacteria (HBACT). Shown are means AE 1 standard deviation (n = 10) for cell abundances (Â 10 3 cells mL À1 for SYN and PEUK and Â 10 5 cells mL À1 for PRO and HBACT), as well as the coefficient of variation (CV, %) for each population. fluorescence in the first hour before decreasing. However, that sample was an outlier in several respects, showing the most abundance variance (5%) and greatest decrease in abundance and chlorophyll fluorescence after 24-h (1a, Fig. 3a).

Method comparisons
The Hoechst 34580 method was compared the SybrGreen I method using a difference plot (Hollis 1996), with the difference in nonpigmented bacteria abundance measured by each method plotted as function of the average of the two values (Fig. 5a). The data show a mean AE standard deviation of the difference of À0.01 AE 0.44 Â 10 5 cells mL À1 , with a standard error of the mean (n = 72) of 0.05 Â 10 5 cells mL À1 . The 95% confidence interval of these data is À0.01 AE 0.89 Â 10 5 cells mL À1 , which includes zero, showing no systematic bias. The percent of values falling between the confidence intervals is 94.4%, suggesting good agreement between the methods. Further, reduced major axis regression of these data have a slope overlapping 1.0 (range, 0.9881-1.198) and an intercept encompassing zero (range, À1.127 to 0.06624), within 95% confidence limits (Fig. 5b).
The Hoechst 34580 method was also compared to the Hoechst 33342 method using a difference plot (Fig. 5c). These data show a mean AE standard deviation of the difference of 0.177 AE 0.591 Â 10 5 cells mL À1 , with a standard error of the mean (n = 42) of 0.091 Â 10 5 cells mL À1 . All data fall within the 95% confidence interval which includes zero, indicating no systematic bias. Reduced major axis regression of these data have a slope of 1.124, with 95% confidence intervals of 1.046-1.127, indicating that the Hoechst 34580 method gives a slightly higher abundance than the Hoechst 33342 method (Fig. 5d). The intercept is À0.6199 Â 10 5 cells mL À1 , and its 95% confidence interval is below zero (range, À0.7025 to À0.2198 Â 10 5 cells mL À1 ). While the single high outlier in this regression (~30 Â 10 5 cells mL À1 ) does increase the slope slightly (1.124 relative to 1.061 Â 10 5 cells mL À1 ), its omission still results in a slope with 95% confidence limits higher than 1.0 at the hundredth decimal place (range, 1.026-1.088 Â 10 5 cells mL À1 ).
Depth profiles from HOT 312 and NF17 show that phytoplankton enumerated after staining with Hoechst 34580-stained samples were similar to unstained samples (Fig. 6).
Prochlorococcus abundances were not significantly different from each other (Fig. 6a). At Stations ALOHA and Ka'ena, where Synechococcus concentrations are very low (≤ 2 Â 10 3 cells mL À1 ), means of stained and unstained samples were also similar (Fig. 6b). However, in the Gulf of Mexico, where Synechococcus were present at higher concentrations in the euphotic zone and only single casts are compared between methods, Hoechststained samples were somewhat higher than unstained samples at a few depths, reflecting the lower precision of the method when applied to low population abundances.  Unpreserved Waim analo Beach surface samples were stained with Hoechst 34580 (Fig. 7). While Hoechst alone stained the live samples adequately, the addition of TFD greatly increased the cellular fluorescence of the eukaryotic phytoplankton (Fig. 7a). TFD functions to block efflux of Hoechst from living cells, for a brighter fluorescence signal (Krishan 1987). In contrast, Synechococcus had only slightly higher DNA fluorescence with TFD (Fig. 7b), while nonpigmented bacteria showed no overall change in fluorescence (Fig. 7c).

Discussion
The Hoechst 34580 method, which can be applied to preserved or living cells, has a precision of ≤ 3% for cell populations at ≥ 10 4 cells mL À1 , and < 10% for lower cell concentrations of~10 3 cells mL À1 when 100 μL of sample is analyzed. Staining is also relatively stable, with the same DNA signal obtained over~24 h period. However, preserved phytoplankton samples gradually lose pigment fluorescence on shorter time scales so it is advisable to analyze thawed samples as soon as possible for maximum signal. Abundance estimates in preserved samples also decrease somewhat after 6 h at room temperature. The method compares well to other methods, in particular Hoechst 33342 and SybrGreen I staining.
The SybrGreen I staining method is quite effective at separating Prochlorococcus from nonpigmented bacteria when the cellular red fluorescence of Prochlorococcus is high enough (e.g., deeper natural samples or cultures). However, for natural samples from shallow depths, SybrGreen I fluorescence interference into the chlorophyll fluorescence detector prevents clear separation of these populations, as previously noted by Marie et al. (1997). Marie et al. (1997) also recommends analyzing each sample twice: first, live and unstained for phytoplankton, and second preserved and stained for nonpigmented bacteria. With Hoechst 34580, only a single sample run is required regardless of Prochlorococcus' intrinsic chlorophyll fluorescence; and if a suitable instrument is available at the time of collection, this dye also may be used to stain living cells, so again a single sample can be analyzed for obtaining information on all populations.
The instrument used here, the Beckman Coulter CytoFLEX S, is a compact, easily transportable, sensitive flow cytometer, suitable for at-sea use. In addition to the availability of the 375 nm laser on this instrument, it has very sensitive avalanche photodiode detectors that allow the detection of natural populations of surface Prochlorococcus, which is not possible for many instruments (Gérikas Ribeiro et al. 2016). Avalanche photodiodes have a higher quantum efficiency than photomultiplier detectors at wavelengths greater than 650 nm (Lawrence et al. 2008). Previously, surface Prochlorococcus has usually only been detectable using very expensive, power-intensive flow cytometers which are comparatively difficult to take to sea and require highly trained operators to maintain and use.
The Hoechst 34580 DNA staining method allows enumeration of phytoplankton and nonpigmented bacteria in a single sample, using relatively low power 375 nm laser excitation. This is a distinct improvement over methods requiring expensive, power-intensive UV lasers for excitation, or the use of stains emitting in the green with emitted fluorescence overlapping natural pigment fluorescence. This methodology is effective for staining living or preserved cells, making it versatile in application for shore-and ship-based analyses. The simplicity of the method is also attractive, with no preincubation of sample in RNAse or heating. This new methodology, using a simple to operate and relatively inexpensive instrument, will allow more researchers to have the capability to study microbial populations, and particularly picoplankton, on shore or at sea.