Characterization of sub-pollen particles in size-resolved atmospheric aerosol using chemical tracers.

Pollen grains may contain allergens that exacerbate allergic respiratory diseases like asthma and rhinitis. In the presence of water, pollen grains (10-100 μm) can rupture to produce sub-pollen particles (SPP) with diameters <2.5 μm, which in comparison to intact pollen grains, have longer atmospheric lifetimes and greater penetration to the lower lung. The current study examines SPP, fungal spores, and bacteria in size-resolved atmospheric particulate matter (PM) using chemical and biological tracers. During springtime tree pollen season in Iowa City, Iowa, fine particle (PM2.5) concentrations of fructose (a pollen chemical tracer) increased on rainy sampling periods, especially during severe thunderstorms, and peaked when a tornado struck nearby. Submicron fluorescent particles, measured by single-particle fluorescence spectroscopy, were also enhanced during rain events, particularly thunderstorms in agreement with the chemical tracer measurements. PM2.5 sucrose (a pollen chemical tracer) concentrations were higher in early spring when nighttime temperatures were closer to freezing, while fructose concentrations were higher in late spring with warmer temperatures, consistent with chemical tracers being sensitive to seasonal temperature influences. The first co-located measurements of fructose and Bet v 1 (birch pollen allergen), indicated that SPP ranged in diameter from <0.25 to 2.5 μm during rainy sampling periods and that allergens and carbohydrates exhibited distinct size distributions. Meanwhile, mannitol (a fungal spore tracer) peaked on warm, dry days following rain and was primarily in supermicron particles (>1.0 μm), which is consistent with intact fungal spore diameters (1-30 μm). Bacterial endotoxins in PM also increased during extreme weather events, primarily in supermicron particles. While the concentrations of fructose, mannitol, and endotoxin all increased in PM2.5 μm during thunderstorms, the greatest relative increase in concentration was observed for fructose. Together, these observations suggest that SPP containing starch granules and allergens (Bet v 1) were released during rainy sampling periods. This study advances the use of chemical tracers to track SPP and other bioaerosols in the atmosphere, by providing new insight to their size distribution and response to extreme weather conditions.

Introduction

The impacts of bioaerosols on human health and the environment depend highly on their size. Generally, intact pollen grains range 10–100 μm, SPP range 0.25–2.5 μm (Hughes et al., 2020), fungal spores span 1–30 μm, bacteria are 0.25–8 μm and viruses are typically <0.3 μm in diameter, but often exist as larger aerosols (Després et al., 2012; Jones and Harrison 2004). Bioaerosols can be pathogens (Després et al., 2012), inflammatory agents (Thorne 2021), and/or carriers of allergens that can exacerbate asthma (WHO 2017). For example, endotoxin, found in the cell walls of gram-negative bacteria, is implicated in asthma, bronchitis, toxic pneumonitis, chronic obstructive pulmonary disease, and pulmonary function decline (Thorne et al., 2010). A particle’s size impacts its atmospheric lifetime, transport, and the extent to which it penetrates the human respiratory tract. Typically, particles >10 μm in aerodynamic diameter are trapped in the upper respiratory tract, particles ≤10 μm can reach the thoracic region, and particles ≤2.5 μm can penetrate the gas exchange region of the lungs (Brown et al., 2013; Hinds 1999). Bioaerosols can also impact the Earth’s climate and precipitation as ice nuclei (IN) and cloud condensation nuclei (CCN) (Murray et al., 2012; Sun and Ariya 2006). Due to the health and environmental impacts of bioaerosols, it is important to understand the factors affecting their atmospheric size distribution and abundance.

Anemophilous pollen grains are released from flowering plants in large quantities to transport male genetic material to the female flower for reproduction. The abundance and type of pollen in the Midwestern United States varies seasonally. Tree pollen (e.g. birch, oak, alder, and hazel) are released during springtime, whereas weed and grass pollen are elevated in summer (Després et al., 2012; Targonski et al., 1995). Meteorologically, warm and dry weather conditions favor the active release of pollen grains, while rainfall typically lowers the overall pollen count by scavenging (Taylor et al., 2004).

In response to rain and/or high humidity, many pollen types have ruptured to release SPP including Chinese elm (Miguel et al., 2006), grass (Taylor et al., 2002), giant ragweed (Stone et al., 2021), birch, alder, and hazel (Grote et al., 2003). Processes of pollen rupture include osmotic rupturing, abortive germination, and mechanical rupturing (Miguel et al., 2006; Grote et al., 2003; Taylor et al., 2004; Visez et al., 2015). Pollen rupturing in the atmosphere can increase bioaerosol number concentrations by three orders of magnitude, as a single giant ragweed pollen grain has been shown to rupture into 1400 SPP (Stone et al., 2021). While intact pollen grains are retained in the head-airway region, allergen-loaded SPP in the size range of 0.6–2.5 μm can penetrate the lower region of the lungs causing severe immunological responses in allergic individuals (Suphioglu et al., 1992; Wilson et al., 1973; Schäppi et al., 1997a¨). “Thunderstorm asthma” or epidemic asthma outbreaks coinciding with thunderstorms when pollen concentrations are high are hypothesized to involve SPP (D’Amato, 2017). Individuals affected by these outbreaks were commonly outdoors when the thunderstorm struck and were sensitive to the abundant pollen type as demonstrated by the high levels of pollen-specific Immunoglobulin (IgE) (D’Amato, 2019; Price et al., 2021). The proposed mechanism for these asthma outbreaks involves thunderstorm updrafts lofting pollen grains into the base of the clouds where high relative humidity triggers osmotic rupture to release SPP that are transported to ground level by downdrafts (Taylor and Jonsson 2004). This hypothesis is supported by our recent publication demonstrating that a significant increase in the SPP with diameters 0.25–2.5 μm during thunderstorms and rain events in the springtime tree pollen season, with peak concentrations occurring during convective thunderstorms with strong downdrafts, high rates of rainfall, and lightning (Hughes et al., 2020).

Fungal spores are produced sexually or asexually by fungi enabling them to colonize new areas in the environment. Ambient fungal spore concentrations vary with season, time of the day, and weather (Elbert et al., 2007; Bowers et al., 2013). The most abundant airborne fungal spore genera are Cladosporium, Alternaria, Penicillium, Aspergillus, Epicoccum, yeasts, rusts, and basidiomycetes (Rockett and Kramer 1974; Rotem and Aust 1991; Ataygul et al., 2007; Janine et al., 2009). Species like Cladosporium depend on dry discharge by agitation during dry and warm weather conditions (Herrero and Zaldivar 1997). In contrast, some species of fungi (e.g., Ascomycota) disperse spores by wet discharge (via liquid jets or droplets) under humid conditions at night, early mornings, or after rain. Fungal spores have also been found to rupture into particles ranging in diameter from 0.01 to 1 μm in size after exposure to ~98% relative humidity for ~10 h followed by drying, but not in response to short hydration and drying cycles (China et al., 2016). Some of the affected individuals during epidemic thunderstorm asthma events were sensitized to fungal spores like Alternaria alternata suggesting a possible involvement of fungal spores in these events (Pulimood et al., 2007).

Like pollen and fungal spores, the variability of airborne bacterial concentrations depends on seasonality and meteorological conditions. Airborne bacteria are those commonly found in soil, such as Proteobacteria and Actinobacteria, but can also include Cyanobacteria, Verrucomicrobia, Acidobacteria, and Bacillus (Després et al., 2012; Shaffer and Lighthart 1997). Airborne bacteria are frequently attached to other particles such as soil, vegetative detritus, or bacterial agglomerates (Lighthart and Shaffer 1997). Typically, atmospheric bacterial counts are higher in spring and summer compared to winter due to warmer temperatures favoring the growth of vegetation which provide surfaces for bacterial growth (Carty et al., 2003). However, bacterial endotoxin concentrations can also peak when crops are harvested, due to aerosolization of bacteria from plant surfaces and soil (Pavilonis et al., 2013; Rathnayake et al., 2016). Artemisia sp. pollen grains (16.5–36.5 μm) have been recorded as another carrier of bacterial endotoxin in coarse PM during summer (Oteros et al., 2019). Bacteria can also be transferred from soil to aerosol (<10 μm) via splashing caused by impacting raindrops, contributing to elevated airborne bacterial concentrations during and after rain events (Joung et al., 2017, Robertson and Martin, 1994).

The time-varying concentrations of bioaerosols in the atmosphere can be assessed via chemical and biological markers. Endotoxins, lipopolysaccharides present in Gram-negative bacteria can be used to track bacteria in atmospheric PM samples (Thorne et al., 1996). Mannitol is a chemical tracer for fungal spores (Bauer et al., 2008). It is used to store energy in fungal spores, accounting for approximately 4–7% of spore mass and its concentration correlates with airborne fungal spore counts (Bauer et al., 2008; Witteveen and Visser 1995). Glucose, sucrose, and fructose are energy storage materials found in pollen grains that contributes to 5–14% of pollen mass and they have been used as chemical tracers to quantify airborne pollen concentrations (Fu et al., 2012; Rathnayake et al., 2017; Barnard 1975; Jia et al., 2010; Medeiros et al., 2006). The concentrations of these carbohydrates in pollen grains vary with developmental stage and external environmental conditions (Vesprini et al., 2002). For example, the distribution of these molecules across mono-, di-, and polysaccharide impacts the pollen water content and turgor pressure (García et al., 2017; Pacini et al., 2006). Sucrose further resists formation of ice crystals in freezing temperatures, preserving pollen viability (Firon et al., 2012). Pollen antigens also can be used to quantify specific pollen species in the atmosphere. For example, Bet v 1 is quantified to find the airborne birch allergen concentration (Schäppi et al., 1997b; Buters et al., 2012). Further insight to the size-distribution of bioaerosols is gained by analysis of airborne particle samples separated by size.

Our prior publication (Hughes et al., 2020) discusses the release of SPP during spring rain events with a primary focus on single-particle fluorescent concentrations and meteorology. The current manuscript expands upon our prior study by Hughes et al. (2020), with a focus on the use of chemical tracers to track SPP in the atmosphere and new assessments of convective thunderstorms on the airborne concentrations and size distributions of fungal spores and bacteria. The size resolved measurements of SPP discussed in this manuscript are mainly focused on five rainy sampling periods (May 8, 16, 17, 18, and 24) that were selected based on the largest relative increase in chemical tracers and fluorescent biological particles. The objectives of this study are to 1) establish the distribution of pollen grains, SPP, fungal spores and bacteria across five particle size ranges (aerodynamic diameters <0.25, 0.25–0.50, 0.50–1.0, 1.0–2.5, >2.5 μm) during rain and exteme weather events, 2) determine the seasonal and meteorological conditions coinciding with elevated bioaerosol concentrations, and 3) provide recommendations for the use of chemical tracers to track bioaerosols in future atmospheric research. Together, these objectives advance the scientific tools and knowledge surrounding the size distribution of atmospheric bioaerosols and their response to changing meteorological conditions. A separate manuscript is in preparation (Hughes et al. in preparation) to describe the single-particle fluorescent data and meteorology during other rain and non-rain events in spring 2019.

Methodology

Collection of ambient particulate matter samples and field data

Site description

Field measurements were conducted from April 06 – May 30, 2019, at the University of Iowa Air Monitoring Site, Iowa City, Iowa, USA (+41.6647, −91.5845). This peri-urban site is located on the edge of the urban areas of Iowa City and Coralville and is surrounded by woods, meadows, prairie undergoing reconstruction, a miscanthus field, and a parking lot.

Online measurements

Daily meteorological data with 1 min time resolution were collected using a meteorological station (Vantage Vue, Davis Instruments) installed on the rooftop of a trailer 3.5 m above ground level. Real-time fluorescent particle concentrations with optical diameters 0.5–20 μm were analyzed using Wideband Integrated Bioaerosol Sensor (WIBS-NEO, Droplet Measurement Technologies) installed inside a wooden shed at the field site. The particles were sampled using a total suspended particulate sampling head (Mesa Laboratories) at 5 L min−1 installed on the shed roof 4 m above ground level. The sampled air was drawn through stainless steel tubing. The WIBS NEO isokinetically sampled at its designated flow rate of 0.3 L min−1, while 4.7 L min−1 make up air was drawn by an auxiliary pump. Single particle data collected were analyzed using WIBS toolkit in Igor Pro. Additional details on the WIBS NEO set up and data analysis are available in the supporting information to Hughes et al. (2020). PM mass measurements for PM2.5 and PM10 were obtained using beta attenuation monitors (BAMs) located at the field site, following the methods described in Vaughn (2009).

Active sampling

23-h filter samples were collected from 08:00 to 07:00 the following day. PM2.5 samples were collected onto pre-baked 90 mm quartz fiber filters (Pall Life Sciences) using a medium volume sampler (URG Corp.) equipped with a sharp-cut cyclone to select PM <2.5 μm at a flow rate of 90 L min−1. The sampler was mounted on to a wooden platform with the inlet 2.5 m above ground level with a flow rate of 90 L min−1 Size-resolved PM samples were collected over 23 h by three 5-stage Sioutas Cascade impactors (SKC Inc.) that were installed on the wooden platform 2 m above ground level and collected particles with aerodynamic diameters >2.5, 1.0–2.5, 0.50–1.0, 0.25–0.50, and <0.25 μm (50% cutoff diameters). Samples were collected at a flow rate 9 L min−1 onto 25 mm Teflon filters (SKC Inc.) for chemical and biological analyses. A third impactor operating at a flow rate 8.5 L min−1 collected particles onto 25 mm polycarbonate filters (Isopore, Fisher Scientific, pore size 0.1 μm) for scanning electron microscopy (SEM) analysis. Pre-baked quartz fiber filters (37 mm, Pall Life Sciences) were used as after-filters in all three impactors. Teflon and quartz filters were stored in a freezer at −20°C while polycarbonate filters were stored in a desiccator (relative humidity <20%) at ambient temperature.

Intact pollen collection and analysis

Intact pollen grain concentrations were measured using a volumetric spore trap (Burkard Manufacturing Company) co-located with other measurement platforms. The samples were collected on to a microscopic slide (75 × 25 × 1.0 mm, Micro Slides, VWR) coated with a thin layer of grease (Lubriseal, Thomas Scientific). The collected samples were mounted and imaged using Olympus BX-61 light microscope with an automated slide scanner (Leica Aperio Ariol). All pollen taxa are included in the intact pollen grain concentrations. Longitudinal counting method was used to obtain 1 h pollen counts. Birch pollen identification was conducted for May 8, 16, 17, 18, and 24 by re-analyzing higher resolution images of 1 h intervals starting at 15:00, 22:00 and 5:00 to differentiate birch/mulberry pollen grains from previous publication (Hughes et al., 2020). Further description of slide mounting, instrumental analysis, pollen identification and determination of pollen counts are detailed in the supporting information of Hughes et al. (2020).

Analysis of carbohydrates

Glucose, sucrose, fructose, and mannitol were quantified using high-performance anion-exchange chromatography with pulsed amperometric detection (HPAEC-PAD, Dionex ICS 5000, Thermo Fisher, Sunnyvale, CA, USA) following the methods described by Rathnayake et al. (2017).

Sample preparation

All glassware was cleaned with tap water (5 times), deionized water (5 times) and ultrapure (UP) water with resistivity 18.2 MΩ cm−1 (Barnstead EasyPure II, 7401; 5 times) and baked for 5.5 h at 500°C. Plastic vials were cleaned with UP water (5 times). Half of the PM2.5 filter samples were subsampled using filter punches and placed inside the plastic vials. Teflon filters were pre-wetted with 50 μL of acetone (Sigma Aldrich). Substrates were extracted into 4.00 mL of UP water by rotary shaking for 10 min at 125 rpm, sonication for 30 min at 60 Hz (Branson 5510, Danbury, CT, US) and rotary shaking again for 10 min. The extract was then filtered through 0.45 μm polypropylene syringe filters (Whatman, GE Healthcare Life Sciences) before analysis.

Instrumental analysis

Samples were injected into the HPAEC-PAD equipped with a CarboPac PA20 analytical column (Dionex), a guard column (Dionex), and an electrochemical detector consisting of a disposable gold working electrode and a pH-Ag/AgCl reference electrode. Waveform A was used for the pulsed amperometric detection (Jensen and Johnson 1997; Rocklin et al., 1998). Instrumental control, data acquisition, and analysis were executed using Chromeleon 7 software. Mannitol (Sigma Aldrich; > 98%), glucose (Sigma Aldrich; > 98%), sucrose (Fisher Scientific; > 99%), and fructose (Sigma Aldrich; > 99%) were separated isocratically using 10 mM sodium hydroxide with a flowrate of 0.5 mL/min. Initially, a 200 mM NaOH solution was prepared by diluting (with UP water) a 50% w/w NaOH (Fisher Scientific) solution and the needed 10 mM concentration was achieved using UP water with the mobile phase proportioning in the instrument.

Eight-point calibration curves of fructose, glucose, sucrose, and mannitol were linear from 0.0100 to 5.00 ppm with squared correlation coefficient (r2) values > 0.995. Samples were analyzed in batches, consisting of 7–8 samples, a field blank, a lab blank, and a spike recovery sample. Spike recovery was assessed by spiking a blank filter with 50.0 μL of 25.0 ppm carbohydrate solution. The recovery percentages were in the range of 86–114%. Uncertainty of the samples was propagated from the standard deviation of the field blanks to account for precision error and 14% of the measurement as a conservative estimate of inaccuracy indicated by spike recovery analysis.

Analysis of endotoxins

For endotoxin analysis, substrates were transferred to pyrogen-free 15 mL tubes and stored at −20 °C until analysis. All glassware was made endotoxin-free by heating overnight at 200 °C prior to use. Filters were extracted following a method previously described by (Sauvé et al., 2020). Filters were extracted at room temperature using 2.00 mL of Limulus Amebocyte Lysate (LAL) reagent water with shaking for 30 min. This was followed by 30 min of sonication at 26 °C, 10 additional min shaking. and 5 min centrifuging (600×g, 4 °C). A 0.45 mL aliquot of the extract was transferred to pyrogen-free cryovials for endotoxin analysis and the remaining extract was stored for allergen assay. The filter extract was evaluated for endotoxin using the kinetic chromogenic LAL assay as previously described by Thorne (2000). Briefly, eluents were diluted into 4-fold serial dilutions from full strength; 1:4, 1:16. Absorbance was measured at 405 nm, every 30 s for 90 min using a microplate reader (Molecular Devices SpectraMax Plus 384, Sunnyvale, CA, with Softmax PRO 5.4 analysis software) and evaluated against a 12-point standard curve using E. coli 055:B5 for standard endotoxin control. The minimum acceptable r2 value for the standard curve was 0.995.

Analysis of allergens

For the analysis of Bet v 1, a 0.450 mL aliquot of the PM extract and 0.05 mL sterile 10 × Phosphate Buffer Saline (PBS) at pH 7.2 with 0.5% Tween-20 was added to yield 1x PBS with 0.05% Tween-20 and vortexed for 5 min. The extract was stored at −80 °C until the assay was performed. A fluorescent multiplex array kit (MARIA, MRA-C3; Indoor Biotechnologies, Inc.) was used following the manufacturer’s procedure. Sample extracts were thawed, vortexed, and centrifuged for 2 min at 4 °C at 16000×g. Each sample was assayed at full strength and at 1:5 dilution. A 12-point standard curve was prepared using 2-fold dilutions. The plate was read by using an xMAP instrument (BioPlex200, BioRad, Inc.). Each filter extract was assayed for the following allergens: Amb a 1 (ragweed pollen), Bet v 1 EP (birch pollen), and Phl p 5 (timothy grass). Only Bet v 1 was detected.

Analysis of particles using SEM

Particles were analyzed by SEM in the Central Microscopy Research Facility at The University of Iowa by a scanning electron microscope with a field emission gun as the electron source (Hitachi S-4800). Approximately 9 × 9 mm2 of PC filter from each impactor stage was mounted on SEM stubs using carbon tape (Ted Pella Inc.). For morphological analysis, the samples were coated with gold (Emitech sputter coater). The microscope was operated at an accelerating voltage of 5 kV for imaging.

Results and discussion

Pollen and sub-pollen particles (SPP)

This study was conducted from April 6-May 30, 2019, during the tree pollen season in Iowa City, Iowa, USA, and coincided with the transition from spring to summer. According to the National Allergy Bureau scale for tree pollen (AAAAI 2020), the daily average intact pollen grain concentration was high on 64% of the days (ranging 90–1499 grains m−3) and a very high pollen count (>1500 grains m−3) occurred on April 16 at 1514 pollen grains m−3 (Fig. 1b). Of the 55 sampling periods having 23 h each, rain was recorded in 28 periods (Fig. 1a). The intact pollen grain concentrations decreased with rain, likely due to scavenging (Lo and Levetin 2007; Pérez et al., 2009). SPP were observed during many of these rain events, especially thunderstorms, by increases in submicron fluorescent particles detected by single-particle fluorescence spectroscopy and by increases in fructose in size resolved particles <2.5 μm as described by Hughes et al. (2020). The influx of fluorescent particles with the rain indicates that the rain delivered the submicron particles (Figs. S1, S2, and S3) either in the downdraft or suspension of biological material from the surface by splashing of rain. Elevated concentrations of fluorescent particles for limited time after rain suggests a short window of exposure (Figs. S1, S2, and S3). In comparison offline measurement of chemical tracers with 23-h samples lacks the ability of providing the time-resolved information required to assess exposure risks. Fluorescent particle data along with the chemical tracer and allergen data provide evidence for release of SPP during thunderstorms supporting the Taylor and Jonsson’s hypothesis of thunderstorm asthma (Taylor and Jonsson 2004).

Fig. 1.

Time series of a) cumulative rain and daily mean temperature, b) (total) intact pollen grain concentrations, PM2.5 concentrations of c) fructose, and d) sucrose. Missing data are marked by an asterisk. PM2.5 fructose concentrations from Apr 17-May 30 were previously discussed in Hughes et al. (2020).

The presence of SPP in daily PM2.5 samples was examined through measurements of birch allergen (Bet v 1; Fig. S4) and chemical tracers previously associated with SPP (Fig. 1). The Bet v 1 allergen concentrations were low in the extracted PM2.5 samples, near to the lowest calibration point of 0.1 ng mL−1. The lack of variability in PM2.5 Bet v 1 (Fig. S4) is likely due to the low birch pollen concentration leading to low Bet v 1 in the atmosphere. While our prior study suggested birch and/or mulberry pollen were among the major pollen types (Hughes et al., 2020), reanalysis of microscope images indicated that only 0.14% of total pollen were birch with an estimated daily average of <2 grains m−3. In addition to the low birch pollen counts, another reason for low PM2.5 Bet v 1 concentrations observed in extracts of PM samples is the relatively low volumes of air analyzed (approximately 12 m3). To quantify allergens like Bet v 1 in future studies, larger air volumes are needed.

Seasonally, fructose and glucose concentrations increased from early spring to late spring (Figs. 1 and 2). This trend likely results from increasing temperatures that promote maturation of pollen grains and the interconversion of starch into soluble carbohydrates (Firon et al., 2012). Sucrose concentrations were highest in early spring and decreased in the late spring. This trend was opposite to the seasonal variation observed for fructose and glucose and is likely driven by the production of sucrose to prevent freezing. During early spring sampling periods, daily average temperatures were low compared to late spring (Fig. 1a) with night-time temperatures reaching below 0 °C (Fig. S5). Sucrose in pollen grains inhibits ice crystallization, preserving the pollen viability in freezing temperatures and can be metabolically synthesized from glucose and fructose (Firon et al., 2012). Thus, the prevalence of sucrose early in the spring is likely due to freezing temperatures.

Fig. 2.

Time series of a) cumulative rain and daily mean temperature, PM2.5 concentrations of b) mannitol and c) endotoxin. Endotoxin was analyzed in a subset of samples collected from May 3–30.

Daily PM2.5 fructose concentrations ranged 0.05–7.29 ng m−3 and glucose concentrations ranged from 0.27 to 40.9 ng m−3 with the maximum concentrations of both tracers occurring on May 24 when a strong convective storm struck and generated a tornado nearby (Hughes et al., 2020). Meanwhile, sucrose concentrations ranged 0.07–12.9 ng m−3. The maximum sucrose concentration occurred on April 16, which was a day with burning of prairie grasses nearby which could affect the carbohydrate concentration in PM and rainwater (Medeiros and Simoneit 2008; Mullaugh et al., 2014). Among the carbohydrates measured, fructose showed the greatest relative increase during rainy sampling periods, especially May 18 and 24 when severe storms occurred (Hughes et al., 2020). Elevation of submicron fluorescent particles with rain and concurrent decreases in intact pollen concentrations were also observed on April 17, 22, 27–29, and May 28–29. On five of these seven days, fructose in PM2.5 was elevated relative to background days (Fig. 1c). The lack of increase in PM2.5 fructose concentration on April 27 is likely due to moderate rainfall rate not significantly impacting the ambient PM2.5 fructose concentrations when integrated over 23-h. Despite the heavy rainfall rate, the ambient PM2.5 concentration did not increase on May 28. Similar observations were made during spring 2013 showing that fructose is more sensitive to rain compared to glucose and sucrose at this location (Rathnayake et al., 2017). On May 22, fluorescent particles spiked at 09:27, persisted for 10 min, and particles >2 μm remained elevated for 14 h (Fig. S7). No rain fell on this day and grass cutting nearby the sampling site were expected to be the source of these fluorescent particles. Relatively high concentrations of glucose, sucrose, and fructose in PM2.5 (Fig. 1), were likely due to pollen, plant debris and dust resuspension (Simoneit et al., 2004).

The mass of SPP in PM2.5 was roughly estimated from the fructose enhancement on rainy days relative to background periods and by assuming that the fructose above background levels was associated with SPP. Mass concentrations of SPP were estimated by dividing the fructose concentrations associated with SPP by the fructose-to-pollen mass ratios observed for oak pollen in the literature (3.3% by average mass for red and pin oak) (Rathnayake et al., 2017), due to the relatively high abundance of oak trees in Iowa City area (Treeplotter Inventory, 2019). Estimated SPP mass concentrations (averaged over 23 h) ranged 1.4–180 ng m−3 and accounted for 0.1–2.2% of PM2.5 mass (Table S2). The number of pollen grains ruptured into PM2.5 was approximated by dividing the estimated SPP mass concentration by the measured mass of an English oak pollen grain (10.8 ng grain−1; Brown and Irving (1973)). The estimated number of ruptured oak pollen grains during rain events ranged from <1 to 17 grains m−3, corresponding to rupturing of <1–11% of pollen grains (Table S2). These estimates are limited by available data on pollen grains and knowledge of what pollen types undergo rupture. As an indication of the uncertainties in these estimates, using data for the less abundant birch pollen (with a fructose-to-pollen mass ratio of 1.8% by weight (Mueller et al., 2016) and pollen grain mass of 7.9 ng grain−1 (Schäppi et al., 1997b), would roughly double the SPP mass concentrations and triple estimates of pollen grains ruptured. Thus, these calculations provide only rough approximations of ambient concentrations of SPP and extent of pollen grain rupture.

Even though glucose has previously been suggested as a pollen tracer because of its relatively high mass fraction in pollen grains (Medeiros et al., 2006; Speranza et al., 1997; Rathnayake et al., 2017), the data collected during this study (Fig. 2b) suggest contributions from other sources to its atmospheric concentration. The glucose concentration strongly correlated with mannitol (R = 0.904, P < 0.001; Table S1). The contribution from many other sources (e.g. growing leaves in spring, fungal spores, and soil particles) (Simoneit et al., 2004; Medeiros et al., 2006; Tereshina et al., 2000; Jia and Fraser 2011; Pashynska et al., 2002) to glucose could lead to a diurnal variation of peak daytime glucose and peak nighttime mannitol concentrations (Graham et al., 2003) in 23 h samples. Together, these observations suggest that among the measured pollen tracers, fructose is the most sensitive to rain during late spring in Iowa City; however, further measurements are needed prior to generalizing these findings to other locations.

To more accurately determine the size distribution of bioaerosol classes, PM samples collected by a Sioutas impactor were analyzed for Bet v 1 and carbohydrates in five size ranges (with 50% cutoff diameters): >2.5 (stage A), 1.0–2.5 (stage B), 0.5–1.0 (stage C), 0.25–0.50 (stage D), and <0.25 μm (E, after-filter) (Fig. 3). These measurements were applied to five sampling periods with precipitation, elevated PM2.5 pollen tracer concentrations, and increases in ambient fluorescent biological particle concentrations (May 8, 16, 17, 18, and 24) (Hughes et al., 2020). Additionally, three background periods (May 7, 15, and 23) were analyzed, which had relatively low fluorescent particle and chemical tracer concentrations, no rain or light rain events (Table 1), and close proximity in time to observations of SPP (Hughes et al., 2020). In the impactor samples, pollen tracers observed on stage A were interpreted as intact pollen grains and large SPP >2.5 μm, while pollen tracers on stages B-E were interpreted as SPP <2.5 μm (Fig. 3). These designations are supported by SEM experiments that showed a few intact pollen grains on stage A (Fig. 4a and b), but no pollen grains on stages B-D. Stage E was not analyzed by SEM due to measurement incompatibility with QFF. Our observation of only a few intact pollen grains on stage A is consistent with this 5-stage impactor having low collection efficiency for particles >10 μm (Misra et al., 2002).

Fig. 3.

Mass concentrations of fructose (a chemical tracer for pollen), Bet v 1 (birch allergen), mannitol (a fungal spore tracer), and endotoxin from Gram-negative bacteria, measured in ambient PM collected by a 5-stage impactor on 5 days with local rain events. Measurements are reported relative to background periods corresponding to dry days. The mannitol concentration on May 18, and fructose concentrations on these 5 days were previously reported by Hughes et al. (2020).

Table 1

Summary meteorology, daily average pollen counts, PM2.5 tracer concentrations, and background sampling periods in May 2019.

Sampling period[a]	Description of meteorology[b]	Temperature range (°C)	Cumulative rainfall (mm)	Daily average pollen counts (grains m−3)	PM2.5 fructose concentration (ng m−3)	PM2.5 mannitol concentration (ng m−3)	PM2.5 endotoxin concentration (EU m−3)
May 7 (background)	Early morning light rain on May 8	7.2–16.6	3.8	611	0.31	0.71	0.012
May 8	Multiple rain events with moderate to heavy rainfall, and an afternoon thunderstorm	6.8–20.5	16.8	185	0.67	5.55	0.120
May 15 (background)	No rain events	10.7–25.0	0	145	1.10	6.18	0.013
May 16	Multiple rain events with light to heavy rain, thunderstorm around mid- night and a severe thunderstorm at night	17.5–33.1	18.2	449	2.17	11.8	0.183
May 17	Continued rain event from prior sampling period and light rain in the afternoon	9.7–20.8	6.4	159	1.15	7.70	0.152
May 18	Multiple rain events with moderate to heavy rain and several thunderstorms	12.5–25.3	20.6	117	5.70	25.0	0.213
May 23 (background)	Early morning light rain on May 24	13.4–23.7	10.4	350	1.21	4.8	0.084
May 24	Morning and evening rain events with heavy rainfall and severe evening thunderstorm with a tornado	14.4–27.7	21.6	627	7.29	24.3	0.313

Sample period begins at 08:00 on the date listed and ends at 07:00 the following day.

Rain events are defined based on rainfall rate (Bluestein and Howard, 1993)

Fig. 4.

Scanning electron microscopy images of a) and b) Mulberry pollen grains c) ascospore d) basidiospore (Particles a, b, c, and d were observed on the stage A of cascade impactor on May 18). Intact pollen grains appear deflated as the images were taken under vacuum or due to being stored in the desiccator until analysis.

Bet v 1 was elevated in PM < 2.5 μm on the five selected rainy sampling periods, compared to background periods in which Bet v 1 was not detected (Fig. 3). These observations provide additional evidence of SPP release during springtime thunderstorms and further suggests that birch pollen grains contributed to SPP, although to a small extent. Bet v 1 was most frequently detected in the PM size range <0.25 μm with highest concentration in 0.5–1.0 μm size particles on May 24. The appearance of Bet v 1 on rainy sampling periods agrees with prior observations in which light rainfall resulted in an increase in Bet v 1 in particles <7.2 μm (Schäppi et al., 1997b). Bet v 1 was also observed in submicron and <2.4 μm particles previously (Pehkonen and Rantio—Lehtimäki 1994) agreeing with our observations of Bet v 1 in particles with diameters <2.5 μm (Fig. 3). Additionally, Grote et al. (2003) indicates that Bet v 1 in birch pollen grains is mainly found in the cytoplasm and is released from the pollen grain during rupture in the presence of rainwater, agreeing with the observation of the allergen in the submicron particles during rainy sampling periods. Taylor et al. (2004) observed that rupturing of birch pollen grains release allergen loaded SPP distributed in size ranges 0.4–2 μm and 2–4 μm, agreeing with our observations of Bet v 1 in these size ranges. Bet v 1 concentrations were low or below detection limit on top impactor stage, likely due to low collection efficiency for particles >10 μm (Misra et al., 2002) coupled with the low atmospheric birch pollen concentration. In contrast, Buters et al. (2012) collected size resolved samples with a higher flow rate (800 L min−1). They reported ~90% of the Bet v 1 in PM > 10 μm fraction when atmospheric birch pollen counts were higher than our study and no recorded thunderstorms. These observations suggest that the thunderstorms are the likely reason for the increase in Bet v 1 concentration in submicron particles in Iowa City.

The fructose concentration was elevated in <0.25–2.5 μm particles during the five selected rainy sampling periods compared to background days (Fig. 3). The highest fructose concentration was recorded on May 18 in 0.25–0.5 μm size particles (Fig. 3). These observations agree with previous observations of fructose concentration increasing in PM < 2.5 μm during periods with rain (Rathnayake et al., 2017). As fructose is associated with starch granules (0.5–2 μm) in the pollen grains (Speranza et al., 1997, Matikainen and Rantio-Lehtimäki 1998), the appearance of fructose in PM < 2.5 μm during rainy sampling periods suggests the release of SPP due to pollen rupturing. The concentration of fructose and Bet v 1 often peaked in different PM size ranges. Bet v 1 was most frequently detected in the PM size range <0.25 μm, while fructose concentrations often peaked in particles 0.25–1.0 μm. This difference in particle size likely occurs for several reasons. Bet v 1 is unique to birch pollen, while fructose is common to many pollen types. Thus, the size distribution of fructose is expected to be impacted by other pollen types during the sampling period (Hughes et al., 2020), which is supported by the relatively low concentrations of Bet v 1 compared to fructose (Fig. 3). These data suggest that fructose is not an indicator of birch allergens, even when associated with SPP. Another reason for difference in size distribution could be the distribution of fructose and Bet v 1 in the pollen grain. As previously mentioned, fructose is mainly found in the starch granules, while Bet v 1 is present in the cytoplasm of pollen grains (Grote et al., 2003). Although Bet v 1 is coated on starch granules it may also be transferred into other types of particles and aerosolized during rain events (Schäppi et al., 1997a) resulting in a differing size distribution.

Size-resolved sucrose concentrations increased in particles <2.5 μm on May 8, 16, 17, and 18 (Fig. S6). The increase in tracer concentration in similar size particles indicates both fructose and sucrose are originated from the same source, likely SPP that are released due to pollen rupturing events during the rainy sampling periods. Variations in the relative concentration of fructose and sucrose are likely to result from numerous factors, including seasonal temperature, pollen type, and developmental stage (Carrizo García et al., 2017; Firon et al., 2012, Carrizo García, 2010). The increase in concentration of both pollen tracers in submicron particles provides further evidence of the release of respirable SPP during rain events.

Fungal spores

During the spring 2019 sampling period, mannitol was used as a tracer of fungal spores following Bauer et al. (2008). PM2.5 mannitol concentrations gradually increased from April to May following the increase in daily average temperature (Fig. 2c), agreeing with the warmer temperatures promoting fungal growth (Jones and Harrison 2004). The mannitol concentrations peaked during sampling periods after rain due to favorable conditions for fungal growth (Jones and Harrison 2004; Van Osdol et al., 2004). PM2.5 mannitol concentrations ranged 0.10–32.5 ng m−3 with the maximum concentration on May 30, a dry and warm sampling period after several consecutive rainy sampling periods (Fig. 2c).

According to the size-resolved mannitol concentrations during the selected rainy periods, fungal spore mass was observed in the upper 3 size bins with more than 90% of mannitol mass in particles >1 μm (Fig. 3). These observations are consistent with SEM observations of fungal spores on the upper impactor stages (Fig. 4c and d) and intact fungal spore diameters that range 1–30 μm (Després et al., 2012). Light microscopy analysis also indicated the presence of ascospores and basidiospores in the atmosphere. However, spores could not be accurately quantified because of limitations to magnification used. Elevated mannitol concentrations during rainy sampling periods have been previously documented (Lyon et al., 1984; Pinkerton et al., 1998; Van Osdol et al., 2004) and can be explained by active spore release during rain (Lagomarsino Oneto, 2020). Mannitol observed in particles <1 μm is likely due to the fungal spores reaching lower stages of the impactor as only 50% of the particles on each stage are within the designed cutoff diameter (Misra et al., 2002; Hinds 1999).

Co-located time-resolved fluorescent measurements showed an influx of supermicron BC and ABC type particles, primarily at night and early mornings during rainy sampling periods (Figs. S1, S2, S3, and S10). Similar florescent pattern was observed on May 22 and May 30, a warm and dry sampling periods following a rainy period with the high PM2.5 mannitol concentrations (Figs. S7 and S8). The fluorescent signature and the size of the particles (optical diameter of 2–9 μm) agrees with previous fluorescent measurements of fungal spores (Hernandez et al., 2016; Savage et al., 2017).

Bacteria

Endotoxin concentrations in PM2.5 samples ranged 0.008–0.332 EU m−3 and the concentrations increased during rainy sampling periods (Fig. 2d). Size-resolved endotoxin concentrations show that, on background sampling periods endotoxin was primarily present in particles >2.5 μm (Fig. 3) whereas on the rainy sampling periods there was a bimodal distribution of endotoxin mass concentrations. The size mode with particles <0.25 μm could be due to bacterial fragments (Wang et al., 2007). Endotoxin in particles >0.5 μm most likely from intact bacteria (Górny et al., 1999; Després et al., 2012).

The highest endotoxin concentrations were recorded on May 14 (0.332 EU m−3) and May 24 (0.313 EU m−3), which were both rainy days (Fig. 2d). Prior to the rain event on May 14, there was an increase of the wind speed which led to resuspension of particles evident by the increased PM10–2.5 concentrations and an influx of fluorescent particles (Fig. S9). Resuspension of particles due to high wind speeds during the severe thunderstorm and tornado is a possible reason for the high endotoxin concentration on May 24. On May 16, a sampling period with severe thunderstorms, endotoxin concentrations elevated in particles <2.5 μm, (Fig. 3). Coarse particle mass and fluorescent particle number were elevated prior to the start of rain and dropped at the start of the rain (Fig. S10). Meanwhile, the fine fluorescent particle concentration increased (Fig. S10). Airborne bacteria are frequently attached to other particles like soil, vegetative detritus, or agglomerates of bacterial cells (Lighthart and Shaffer 1997) and it is expected that the resuspension of such particles led to the elevated endotoxin concentrations in coarse particles prior to rain. Additionally, rainfall can also transfer of bacteria from soil and other surfaces by splashing and impaction of raindrops releasing aerosols <10 μm (Joung et al., 2017), which provides a source of fine particle endotoxin on rainy days. It is expected that the relative increase in PM2.5 endotoxin concentration on May 22 (non-rainy sampling period) (Fig. 2d) resulted from local grass cutting aerosolizing the bacteria attached to plant matter and/or dust.

Conclusions

Through the combination of chemical and biological markers and high time-resolution fluorescent particle measurements, new insight can be gained on tracking pollen and SPP in the atmosphere. Elevated fructose concentrations in fine particles during rainy sampling periods provide evidence of the release of SPP by rupturing of pollen grains, especially during extreme weather conditions like thunderstorms and tornados. Due to its high sensitivity to rain and not showing any interference from other sources, we suggest fructose as a tracer for SPP in the atmosphere. Sucrose concentrations in fine particles depend on temperature. Low atmospheric temperature could lead to high sucrose concentrations in pollen grains making sucrose a more sensitive tracer when temperature is closer to freezing. As this temperature dependance of sucrose was observed in Iowa City, IA, USA during spring, more studies are needed to assess how generalizable these measurements are in other locations. Although glucose has been used as a pollen tracer in previous studies, our results from Iowa City suggest that it is affected by other sources, suggesting glucose is not a reliable SPP tracer. As this study focuses on spring tree pollen season and more analyses are required to extend these observations to other locations and seasons.

Rain and thunderstorms affect SPP, fungal spores, and endotoxins concentrations differently. While Yue et al. (2016) reported the contribution of fungal spores and bacteria to the biological particles concentrations during rain, for the first time our study records elevation of fructose and Bet v 1 concentrations in ambient PM providing evidence to support the contribution of SPP to biological particles <2.5 μm during rain events. SPP are highest in particles <2.5 μm during rain, while fungal spore concentrations typically increased after rain events and overnight, in particles >1 μm. Bacterial endotoxin concentrations increased primarily in particles >1 μm prior to rain, likely due to dust elevated by wind, and in particles <1 μm during rainy periods. Based on the timing of the rain events, co-exposure to fungal spores, bacterial endotoxin, and SPP are likely to occur. Other co-pollutants in thunderstorms, like ultrafine particles (Wang et al., 2016), nitrogen oxides produced by lighting (Choi et al., 2005), and/or ozone (Betts et al., 2002) may further exacerbate the effects of allergenic bioaerosols. With thunderstorms predicted to increase in intensity and severity (Maupin et al., 2021), there is a need to advance public health preventive measures for susceptible individuals.