Corresponding author: A. J. Prenni, Department of Atmospheric Science, Colorado State University, Campus Delivery 1371, Fort Collins, CO 80523, USA. (email@example.com)
 With 18% of the total U.S. landmass devoted to croplands, farmland and farming activities are potentially major sources of biogenic particles to the atmosphere. Farms harbor large populations of microbes both in the soil and on plant surfaces which, if injected into the atmosphere, may serve as nuclei for clouds. In this study, we investigated two farms as potential sources of biological ice nuclei (IN): an organic farm in Colorado and a cornfield in Nebraska. We used a continuous-flow diffusion chamber (CFDC) to obtain real-time measurements of IN at these farm sites. Total aerosol particles were also collected at the sites, and their temperature-dependent ice nucleating activity was determined using the drop freezing method. Quantitative polymerase chain reaction and DNA sequencing of 16S rDNA clone libraries were used to test aerosols and washings of local vegetation for abundance of theinagene in ice nucleation active bacteria (from the well-known group within theγ-Proteobacteria) and to identify airborne primary biological aerosol particles. The vegetation in each of these farms contained 6 × 105 to 2 × 107ina genes per gram vegetation. In contrast to the vegetation, airborne ina gene concentrations at the organic farm were typically below detectable limits, demonstrating a disconnect between local vegetative sources and the air above them. However, for measurements made during combine harvesting at the Nebraska corn field, ina gene concentrations were 19 L−1, with maximum IN concentrations of 50 L−1 determined from the CFDC at −20°C and above water saturation. At both farms, there was also an apparent biological contribution to the IN population which did not contain the ina gene.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 In a single year, farmers lose millions to billions of dollars due to frost-related injury to crops [Snyder and de Melo-Abreu, 2005]. Epiphytic microorganisms that nucleate ice near 0°C are one causative factor for generating ice crystals that cause frost damage to plants. To date, a few bacterial species have been identified as agents for such damage, including Pseudomonas syringae [Maki et al., 1974], Pseudomonas fluorescens [Kaneda, 1986], Pantoea agglomerans [Lindow et al., 1978b] and Xanthomonas campestris [Kim et al., 1987]. These bacteria can express an ice nucleation active (Ina or INA, where Ina refers to the protein and INA refers to bacteria) protein which, when bound into large aggregates on the outer cell membrane, enables them to nucleate ice at temperatures typically ranging from −4 to −1.5°C [Kim et al., 1987; Lindow et al., 1989; Morris et al., 2008]. Because the large aggregates are labile, a variable and typically small fraction of all cells will possess activity at these warm temperatures. For example, O'Brien and Lindow  observed the fraction of INA P. syringae strains grown on plants able to nucleate at −5°C to range from 1 in 30 to 1 in 20,000 for a variety of experimental conditions. At colder temperatures clusters of a few proteins are enough to trigger freezing, so that by −10°C to −12°C the nucleation frequency may be 1 in 10 cells or higher [Hirano and Upper, 2000; Kim et al., 1987]; a single Ina protein nucleates at −12 to −13°C [Govindarajan and Lindow, 1988].
 With 1.5 billion hectares of land committed to crop production globally, the global arable landmass provides an enormous potential source of microorganisms with ice nucleation activity. Although fluxes from vegetation are a major source of airborne microorganisms [Lindemann et al., 1982] and airborne primary biological particles are thought to represent as much as 25% of the total number concentrations for particles larger than 0.2 μm [Jaenicke et al., 2007], little is known about the number concentrations of biological IN in the atmosphere [DeMott and Prenni, 2010], and their importance is not well established on a global scale [Hoose et al., 2010]. Their atmospheric abundance is particularly relevant for clouds and precipitation, as biological particles may represent a major source of atmospheric IN in modestly supercooled clouds; that is, those with regions warmer than approximately −15°C [Christner et al., 2008; Möhler et al., 2007].
 In this study, we present real-time and offline analyses of ice nucleating activity from aerosols collected at two farms. Samples were collected during harvest from a cornfield grown for feed crop production near Maywood, Nebraska, and samples were collected in fields of several different crops grown organically near Fort Collins, Colorado. The organic vegetable farm is largely dependent on hand harvesting operations that provide minimal disturbance, while the corn field in Nebraska represents a virtual monoculture being subjected to high disturbance during the sampling period. Sampling was done at the two farms during different months and the farms were located ∼370 km apart. As such, these measurements are not meant to provide a direct comparison of farming practices, but rather provide data from two potential IN sources. Measurements were conducted in October and November, as it is expected that in fall there will be a higher concentration of bacterial aerosol due to decay processes [Jaenicke, 2005; Matthias-Maser, 1998]. In addition to measurements of ice nucleation activity from airborne particles, we also quantified the number of ina (inarefers to the gene) genes that code for the ice nucleating protein in INA bacteria (the well-known group within theγ-Proteobacteria) present on the vegetation at each field and in the air above, and identified components of airborne primary biological aerosol particles (PBAP). To this end, we used a combination of the drop freezing method to identify IN spectra (IN concentration versus temperature), quantitative polymerase chain reaction (qPCR) to count theinagene, and thiamine-adenine cloning (TA cloning) and Sanger sequencing of 16S rDNA to identify species of bioaerosols (plants, bacteria and fungi). Finally, using cloning (TA cloning) and Sanger sequencing, we attempted to identify the biological particles responsible for ice nucleation at these sites by collecting and characterizing the IN activated in the real-time IN instrument.
2.1. Mobile Laboratory
 Measurements were conducted using the CSU mobile air quality laboratory (Brown Specialty Vehicles, Inc., Lawrence, Kansas). It uses a 2008 Isuzu chassis equipped with an 8 foot high × 8 foot wide × 14 foot long aluminum body, which creates an enclosed sampling shelter. Measurement locations were selected based on forecasted winds but were restricted based on road access. Honda generators (EU6500is, EU3000i) located downwind of the parked vehicle were used for power. Samples collected are summarized in Tables 1 and 2.
Local observations were supplemented with temperature, wind speed, and RH data for the sampling periods from nearby weather stations. For Grant Farms this was Fort Collins (20 km south), and for Schmidt's farm it was McCook (40 km south).
Grant Family Farms, Larimer County, Colorado
29 Sep 2010, 12:12–16:12
Clear and winds locally estimated at 25 km h−1 from the north. At Fort Collins: 26–28°C, winds 18–32 km h−1 from the north to ENE, 20% RH.
Wheat, harvested and disked.
Grant Family Farms, Larimer County, Colorado
4 Oct 2010, 11:05–15:05
Clear. At Fort Collins: 24–27°C, winds 8–15 km h−1 from SSE to SSW, ∼25% RH.
Potatoes and leeks and beyond that cauliflower (aphid infested).
Grant Family Farms, Larimer County, Colorado
8 Oct 2010, 09:52–13:37
Clear with later clouds. Winds initially gentle from the south and west across ley field, then more strongly from the north across the corn field. At Fort Collins: 18–22°C, winds 6–13 km h−1, 27–42% RH.
To the west and south a ley field of common rye, Austrian winter pea, hairy vetch, and ground cover of grasses. Further to the west a field of squash. To the north a conventionally farmed corn (Zea mays) crop of fully grown dry plants.
Grant Family Farms, Larimer County, Colorado
20 Oct 2010, 10:09–16:09
Clear. At Fort Collins: 22°C, winds ∼15 km h−1 from the east, 20% RH.
To the east the squash field and beyond that the ley field.
Schmidt's Farm, near Maywood, Nebraska
3 Nov 2010, 10:40–15:40
Clear with scattered cumulus in afternoon, 10–15°C, winds estimated at 15–30 km h−1 from the north. At McCook: 14–17°C, 39–45 km h−1 gusting to 56 from the north to NNW, 26–39% RH.
Fully grown corn crop being harvested (plants and surface soil dry).
Table 2. Summary of Results From Each Sampling Sitea
ina Genes in Leaf Washings and Soil (g−1 Fresh Weight Vegetation/Soil)
Values for the drop freezing method are estimated for cases in which data are not available.
BDL = below detection limit.
In DNA extracted from one Biosampler, only one PCR reaction in nine produced a product (confirmed by sequencing). This single positive equated to ∼0.035 genes L−1. In DNA extracted from a second Biosampler no reactions out of nine produced a product.
The qPCR result is for primer 3463r only and is a low estimate due to mispriming. The reaction failed using primer 3462r due to bad mispriming. The soil sample qPCR products were not confirmed using sequencing.
29 Sep 2010 harvested and disked wheat field: wheat straw
3 Nov 2010 corn field being harvested: corn preharvest, corn fragments postharvest (including corn husks), dry surface soild
3.2 × 106, 6.6 × 106, 1.5 × 105
2.2. Aerosol Collection
 Particles were collected in water using sterilized BioSamplers® (SKC Inc.). SKC Biosamplers are similar to glass impingers used to trap airborne microorganisms for subsequent analysis, however the Biosampler nozzle ejects particles at an angle to the sampler's inner wall, significantly reducing particle bounce, preserving aggregates of organisms, and preventing reaerosolization. The SKC Biosamplers collect particles larger than 500 nm at greater than 90% efficiency. Prior to sampling, the Biosamplers were washed with 10% bleach, rinsed in deionized water, autoclaved and baked at 500°C for 24 h to remove any contamination. Particles were collected using four Biosamplers running simultaneously for 4–6 h at a flow rate of 12.5 L min−1 into 20 mL of tissue grade deionized water (Sigma). Water level in the Biosamplers was maintained between 15 and 20 mL throughout the measurement period to ensure efficient particle collection. In all cases, an aliquot of the water was placed in a clean Biosampler, set aside in the field as a blank and analyzed to ensure neither it, the reagents nor the methods introduced contamination.
2.3. DNA Extraction From Biosamplers
 We developed and optimized DNA extraction methods to meet specific goals of the study; as such, more than one extraction method was used. From each site, the contents from one or more Biosamplers were concentrated to 200 μL for 16S rDNA analysis using an Amicon Ultra-15 Centrifugal Filter Unit (Millipore). For one of the samples (29 September 2010), DNA was extracted using the MP FastDNA SPIN Kit for soil (MPBIO). For all of the other samples, DNA was enzymatically extracted using Proteinase K, by adding 200μL of a 200 μg mL−1solution dissolved in 10 mM Tris-HCl at pH 8.5. Samples were then distributed equally among four 200μL PCR tubes, which were held at 55°C for 50 min, activating digestion, followed by Proteinase K inactivation at 95°C for 10 min. Samples were stored at −20°C.
 The suspension from another Biosampler was divided into 2 aliquots: 2–3 mL were removed for estimating the IN concentration via droplet freezing analysis and the remaining 12–15 mL were used for DNA extraction and quantification of the ina gene using qPCR. In this second aliquot, the suspended material was concentrated either by centrifugation at 22,500 g for 3 min (mixed ley on 8 October and cornfield during harvest on 3 November) or by filtering, dropwise, through a 0.2 μm pore diameter Nuclepore™ track-etched polycarbonate membrane (Whatman). Membranes were decontaminated before use by soaking in 15% H2O2 for 5 min followed by two rinses in deionized water. Deionized water had been filtered through a 0.45 μm pore diameter Nalgene sterile filter unit (Thermo Scientific) and autoclaved. For extraction blanks, 30 mL of the same batch of deionized water used in each Biosampler was filtered through a polycarbonate membrane. The portion of the membrane exposed during filtering was removed, cut into pieces and placed in a DNA extraction tube. DNA was extracted from Biosampler samples and extraction blanks using the PowerLyzer™ UltraClean® Microbial DNA Isolation Kit (Mo Bio Laboratories, Inc.). For the 8 October mixed ley and 3 November harvested corn field the standard method was used, but for the other sites we adopted a modified method that increased recoveries. That is, omission of the step to precipitate non-DNA organic and inorganic material (solution MD2), use of two instead of one ethanol washes (solution MD4) and use of two instead of one DNA elution steps (2 × 50μL of solution MD5). This increased DNA recovery from ∼25% to ∼90% (tested using 120 and 1200 P. syringae Cit7 cells added directly to extraction tubes). Homogenization was performed using a FastPrep® bead beater (BIO 101, Carlsbad, CA) at setting 4 for 5 min.
2.4. DNA Extraction From Vegetation
 One bulked sample was obtained from each vegetation type and each comprised approximately 50 small pieces of representative material taken from a broad area upwind of the Biosamplers. Samples weighed from ∼30–150 g. Vegetation varied from photosynthetically active and actively growing to brown, senescent plants awaiting harvest. At the harvested wheat field, sampled on 29 September, only remnant surface straw and stalks remained. At the cornfield under harvest on 3 November a bulked soil sample was also collected; >50, ∼0.5 g pinches of the loose, dry surface soil were taken from exposed patches in the harvested area. Disposable gloves were used throughout and samples stored in new ziplock plastic bags at 4°C for 1–3 d before processing.
 Each vegetation sample was placed in a wide-necked, 4 L polypropylene container that had been decontaminated by soaking in 5% H2O2 for 1 h, and then rinsed twice with deionized water and autoclaved. To this, 500 or 1000 mL of 0.01 M sodium phosphate buffer (pH 7) with 0.1% Difco Proteose Peptone No. 3 (Beckton, Dickinson and Company) was added, and the sample shaken vigorously by hand intermittently over 1 h. A 2 mL aliquot was centrifuged at 22,500 g for 3 min, the supernatant removed and the pellet used for DNA extraction. The soil sample was thoroughly mixed and 0.07 g used for DNA extraction. The deionized water used for rinsing containers and preparing phosphate buffer had been prefiltered through a 0.45 μm pore diameter, Nalgene sterile filter unit and autoclaved. DNA was extracted using the standard protocol for the PowerLyzer™ UltraClean® Microbial DNA Isolation Kit with homogenization as described above.
2.5. DNA Analysis of Biosampler and Vegetation Samples
 From the Biosampler used for 16S rDNA cloning and sequencing, 2 μL of the extracted DNA were PCR amplified using the universal 515F (5-GTGCCAGCMGCCGCGGTAA) and 1391R (5-GACGGGCGGTGWGTRCA) primers [Lane, 1991]. After amplification, the PCR products were cloned into a plasmid vector using the TOPO TA Cloning Kit® (Invitrogen) according to the manufacturer's instructions. We selected 25–70 clones for sequencing from each experiment. Each clone was then Sanger sequenced at the Colorado State University Proteomics and Metabolomics Facility and identified using a National Center for Biotechnology Information (NCBI) Blast search.
 Real-time qPCR was also used to detect and quantify the number ofina gene copies in DNA extracted from Biosamplers, leaf washings and soil. There is likely a direct correlation between the ina gene copy number and the number of INA bacteria, since only one copy of the gene per genome has been found in 10 fully sequenced strains of INA bacteria. These include two P. syringae, four P. ananatis and four X. campestris pathovars (GenBank accession numbers CP000058, CP000075, AP012032, CP001875, CP003085, HE617160, AE008922, AM920689 CP000050 and CP002789, respectively). Counting the ina gene directly will detect live bacteria, including viable but nonculturable cells, and also DNA from dead bacterial cells in which the ina gene sequence remained intact. However, we note that ina gene detection does not guarantee that the Ina protein is present, and conversely that the Ina protein may persist on/in dead bacteria that have lost the ina gene through the action of DNAses. DNA primers that amplified all published ina alleles except some possessed by Xanthomonasspp. (T. C. J. Hill et al., unpublished manuscript, 2012) were used for qPCR. Amplification was performed on a Bio-Rad DNAEngine® fitted with a Chromo4™ Real-time PCR Detector. qPCR products from all vegetation washings and positive Biosampler samples were electrophoresed and sequenced to confirm they wereina genes. All contained a predominant ina allele. When mispriming occurred, bands of approximately correct size were excised, reamplified and sequenced: none were ina genes. For standards we used DNA extracted from known numbers of P. syringae Cit7 cells (enumerated by dilution plating). P. syringae Cit7 was originally isolated from the surface of a healthy navel orange leaf growing near Exeter, CA [Lindow, 1985]. Aliquots of 118 to 1.18 × 107 cells were added to MicroBead tubes and DNA was extracted using the PowerLyzer Kit. Two standard series were made, one with aliquots extracted using the kit's standard protocol for DNA extraction and the other using the high recovery method described above. Each series was used with samples extracted using the corresponding method. A minimum of two replicate reactions with each primer pair were performed for each sample. Gene counts using different primer pairs differed by <35%, and were averaged. Primer pair 3308f and 3462r cleanly amplified the product in reactions containing an estimated 1–3 gene copies (when using standards). Primer pair 3308f and 3463r were consistently able to amplify 10 gene copies, but below five the reaction often failed. All ina gene number concentrations were corrected to standard temperature and pressure.
2.6. Laboratory Measurements of Ice Nuclei
 An aliquot (2–3 mL) from the Biosamplers was used to characterize the temperature spectrum of IN concentration using the drop freezing method. While real-time IN measurements were made at near constant operating conditions to better capture temporal variability, the drop freezing method allows for measurements of IN activity across a range of temperatures, including temperatures inaccessible by real time methods. We note that for the drop freezing method the sample may differ somewhat from what was in the atmosphere, in that all individual particles are pooled into one aqueous mixture/suspension, cell lysing is possible which may lower the apparent number of high-temperature IN in subsequent droplet freezing analyses, and soluble and potentially surface active compounds likely become mixed with previously uncoated particles. Such coatings have the potential to modify the ice nucleation activity of particles. However, recent laboratory work has shown that simple solutes do not impede heterogeneous ice nucleation on IN surfaces acting in the condensation/immersion freezing modes unless certain chemical reactions are involved [Sullivan et al., 2010a, 2010b]. Despite this limitation, the drop freezing method has several advantages for characterizing biological IN: it allows for measurements of larger particles, ice nucleation activity can be characterized over a range of temperatures, and the low detection limit of the drop freezing method is ideally suited for measurements of warm temperature IN.
 IN concentrations were estimated by counting the number of 30 or 50 μL aliquots (n = 32–72) frozen in 96-well polypropylene PCR trays. Trays were cooled in steps to −9.0°C in a thermal cycler (PTC-200, MJ Research). Temperature variation across the head was ±0.2°C of the true temperature measured using a thermistor (VPT-0300, Bio-Rad). The cycler was programmed to descend in 0.5 or 1°C increments, and after 5 min at each temperature the number of frozen wells was counted and the temperature lowered to the next increment (transitions took a few seconds). Once at −9°C the tray was transferred to a 96-well aluminum incubation block (catalog number 13259–260, VWR) cooled to ∼−12°C inside a foam box in a freezer. The thermistor was inserted into a side well and after 10 min the block temperature and number of frozen wells was recorded. The tray was then transferred to two further blocks at sequentially colder temperatures (down to ∼−20°C). Aliquots of the same batch of tissue grade deionized water used in each Biosampler were used as negative controls. The water had 2–5 IN per milliliter at −16 to −20°C, and these background values were subtracted from sample counts. All number concentrations were corrected to standard temperature and pressure.
 For heat treatment effects on IN spectra, trays were transferred back to the PTC-200, thawed at 16°C for 5 min and then heated at 98°C for 20 min to denature proteins and organics. It was then centrifuged at 300 g to pool the contents and retested across the full temperature range as described above. The effectiveness of 98°C treatment was tested onP. syringae isolate Cit7 plus five local INA bacterial isolates with different ina alleles. In five, activity was eliminated (tested to −18°C), while in the sixth (X. campestris) the onset of freezing was lowered from −2.5°C to −9°C. Previous work has shown that fungi (2 Fusarium spp.) as IN also lose all activity above −12°C when heated to 90°C for 10 min [Pouleur et al., 1992], and lichen as IN are sensitive to heating to >70°C [Kieft and Ruscetti, 1990], suggesting their ice nucleating proteins are similarly heat labile. While heat treatment is expected to inactivate biological IN [Christner et al., 2008], it is not expected to affect the ice nucleation activity of inorganic materials, such as mineral dust [e.g., Conen et al., 2011]. Cumulative numbers of IN per milliliter Biosampler water were estimated using the formula where f is the proportion of droplets not frozen and V the volume of each aliquot [Vali, 1971], and using the total water volume and volume of air sampled converted to IN per standard liter of air. Binomial sampling confidence intervals (95%) employed the formula (number 2) recommended by Agresti and Coull .
2.7. Real-Time Measurements of Ice Nuclei
 A ground-based version of the Colorado State University continuous-flow diffusion chamber (CFDC) [Petters et al., 2009; Rogers et al., 2001b] was used for real-time measurements of IN concentrations. In the CFDC, sampled air is directed vertically between two concentric ice-coated cylinders held at different temperatures, creating a zone supersaturated with respect to ice in the annular region. The sample flow (1.5 vLPM), ∼15% of the total flow, is injected between two particle-free sheath flows. As the particles in the sample flow are exposed to ice supersaturations for several seconds, those particles active as IN under the sample temperature and humidity conditions are nucleated and grown to ice crystals larger than a fewμm in size. These larger particles are distinguished from small non-IN aerosols by an optical particle counter at the outlet of the instrument. Physical impaction of larger particles (>2.4μm; 50% cut point diameter for unit density particles, as validated in the laboratory) in advance of the CFDC and reduction of humidity conditions to ice saturation in the lower third of the chamber, via setting the ice walls to equivalent temperatures, prevent false detection of large particles or cloud drops as IN. We note that removing particles larger than 2.4 μm upstream of the CFDC likely removes some particles that can potentially serve as IN. Temperatures (±1°C) and humidities (±3% RH with respect to water maximum uncertainty at −30°C) are well controlled in the instrument [DeMott et al., 2009]. For this study, measurements were made at −20°C at relative humidities typically in the range of 103 to 105%. At these conditions, contributions are expected to the IN population from both dust and biological particles. Condensation and immersion freezing mechanisms of ice nucleation are believed to be captured under these operating conditions. All number concentrations were corrected to standard temperature and pressure.
2.8. Biological Ice Nuclei
 After processing in the CFDC, ice crystals activated as IN and grown to >3 μm diameter were impacted onto a glass slide, which was coated with 5 μL of molecular grade mineral oil (Bio-Rad), using a method similar to that used by our group previously for collections of activated IN onto transmission electron microscopy (TEM) grids [e.g.,Prenni et al., 2009; Rogers et al., 2001a]. The glass slides were sterilized and analyzed in a manner identical to that described for the Biosamplers, including analysis of blank slides. Proteinase K solution was added directly to the top of the glass slide, the solution and mineral oil were mixed using the tip of the pipette, and the resulting suspension was placed into a sterile 50 mL falcon tube and incubated in a water bath for 15 min at 55°C. The sample was then centrifuged at 3750 g for 10 min and transferred into a sterile PCR tube for analysis, as described above. Although we collected IN onto glass slides for four of the five sampling periods, we were able to identify biological IN in only one case (Maywood, Nebraska, farm), when IN concentrations were relatively high. On one day, 4 October, IN were collected onto a TEM grid and were analyzed using scanning electron microscopy.
3.1. Organic Vegetable Farm, NE Colorado
3.1.1. Wheat: 29 September 2010
 The initial measurements came from a wheat field that had already been harvested and subjected to surface tillage at Grant Family Farms (40.77°N, 105.1°W). The surface was a mix of remnant straw, dead plant crowns and associated roots and bare soil. We sampled from a mobile laboratory located south of the field, in clear and warm conditions with winds estimated at ∼25 km h−1 from the north (Table 1).
 Ice nuclei were first detected at −7°C using the drop freezing method, with concentration increasing with decreasing temperature from 0.33 L−1 air at −12°C to 5.4 L−1 air at −20°C (Table 2 and Figure 1). Here and throughout we interpolate drop freezing data to −12°C, as this is the approximate temperature at which a single Ina protein triggers freezing. The average IN concentration detected by the CFDC at −20°C was 1.4 L−1 over a 3.5 h sampling period, with a maximum concentration of 13 L−1. The slightly greater mean concentration indicated by the drop freezing method compared to the CFDC is expected, because drop freezing samples a broader particle size range and allows for freezing over a longer time period, which may result in a higher fraction of particles that nucleate [Broadley et al., 2012; Murray et al., 2011]. For each method, IN concentrations active at −20°C were typical of those reported in ambient air at other locations [DeMott et al., 2010]. Heating the Biosampler sample to 98°C decreased IN concentrations over the entire spectrum, from a 25-fold decrease at −8.7°C to 5- to 10-fold reductions at colder temperatures. The large reduction in IN activity upon heating suggests that the bulk of measured IN were organic, as heat treatment is expected to denature the proteins and other biogenic macromolecules responsible for ice nucleation activity.
 The surface debris remaining on the wheat field harbored 3.4 million ina gene copies per gram fresh weight vegetation as determined by qPCR analysis of their washings (Table 2). This number could include living and viable but nonculturable bacteria, as well as dead INA bacteria for which DNA was at least partially intact. With a substantial population of INA bacteria assumed present on the ground combined with warm and dry conditions and moderate winds, which are generally correlated with upward fluxes of bacteria [Burrows et al., 2009; Lindemann et al., 1982], it was surprising that we failed to detect the ina gene using qPCR analysis of the atmospheric sample (Table 2). As a test of qPCR sensitivity, sample DNA was spiked with genomic DNA from P. syringae cells (isolate Cit7). The lowest spike, the equivalent of 4–16 ina gene copies (>95% confidence interval of the Poisson distribution for a mean of 10 copies added), amplified unambiguously along with one minor misprimed band with both pairs of primers. Assuming a lower detection limit of 5 ina gene copies per qPCR reaction, the method's sensitivity was ∼0.1 ina genes L−1 air, less than half of the IN detected at −12°C. This limit is similar to the concentration of INA bacteria recovered by culturing (0.02–0.18 CFUs L−1 air) above heading wheat in Wisconsin [Lindemann et al., 1982].
 Failure of the qPCR could be due to the qPCR not amplifying ina genes in INA bacteria present at the site, such as inaX in X. campestris. Alternatively, other IN sources may exist, including those from soils or from plant, fungal or bacterial material that do not possess the ina gene. DNA sequencing of the total aerosol sample confirmed that a variety of airborne microbes were present. Of the 70 clones sequenced, the most abundant species belonged to the genus Aureobasidium, a fungus found in multiple habitats that include roots and soil (Table 3). In addition to Aureobasidium, a number of other fungi and bacteria were identified, including two clones from a species that contains INA strains (Pseudomonas syringae).
Table 3. Genera of Bacteria and Eukarya in the Low-Level Atmosphere at Grant Farms for the Four Days Studied in 2010a
Bacteria and Eukarya (mostly fungi) are given to genus level. When more than one, numbers in parentheses indicate the number of clones identified. Potential INA bacteria are shown in bold.
3.1.2. Cauliflower, Potatoes, and Leeks: 4 October 2010
 The second set of measurements was at an organic cauliflower field infested with aphids (Table 1). For these measurements our mobile laboratory was set up to the north, with crops of potatoes and leeks located between the mobile laboratory and the cauliflower field. Drop freezing estimates from the Biosampler revealed an IN concentration of 0.33 L−1 air at −12°C, the same as the organic wheat field. At colder temperatures, however, IN concentrations were much greater, with average values of 13–20 L−1 air at −20°C for both methods used, and a maximum CFDC value of 43 L−1 over a 2 h time period, the highest for any measurements made at Grant Family Farms. On this date, residual IN from the CFDC were captured directly onto a TEM grid. Scanning electron microscopy analysis of 23 IN from this grid indicated contributions from both mineral dust and carbonaceous particles. Heat treatment of the Biosampler particle suspensions to 98°C lowered IN concentrations over the range tested (Figure 2), suggesting that many of the IN were biological or organic. At −8°C there was a 30-fold decrease, while at colder temperatures the reduction was ∼10-fold.
 High numbers of ina gene copies were observed from leaf washings: qPCR detected 19 million ina gene copies g−1 fresh weight of vegetation. Potential INA bacteria were also detected among DNA sequences from the total aerosol sample, with five P. syringae observed (Table 3). Despite the relative abundance of potential INA bacteria in the sequencing results, qPCR measurements of DNA extracted from the combined contents of two Biosamplers failed to detect ina genes. Although strong mispriming by both primer pairs occurred with this sample, spiking tests showed that low levels of ina genes were still detectable (Figure 3), and the method's sensitivity was ∼0.03 ina genes L−1 air, less than one tenth of the number of IN detected at −12°C. Thus, similar to the wheat field, despite high concentrations of ina gene copies on the ground, the presence of potential INA bacteria in the 16S rDNA clone libraries of airborne bacteria, and the negative effect of heat treatment on the IN activity of Biosampler samples, no ina gene copies were detected in the atmospheric samples. These data again show that the number of ina gene copies is poorly correlated with the number of IN detected via drop freezing, and suggest the predominant biological IN are those that do not have the γ-Proteobacterialina gene.
3.1.3. Mixed Ley and Corn: 8 and 20 October 2010
 The final set of measurements at Grant Family Farms came from a mixed ley field. On 8 October, measurements were made northeast of the field, with winds initially out of the south and shifting westerly from across a green, mixed ley field and beyond that a crop of squash. After 2 h the wind turned north and increased in strength coming across a conventionally farmed corn crop containing fully grown dry plants awaiting harvest. On 20 October, measurements were made on the west side of the field, with winds from the east blowing across the squash and mixed ley fields.
 For collections made on 8 October, drop freezing results from the Biosamplers indicated IN concentrations at −12°C of 0.79 L−1 air, which was modestly greater than observed for the two fields sampled previously. At −20°C the average IN concentrations were ∼5 L−1 determined by the drop freezing method and ∼1 L−1 from the CFDC, with peak CFDC values near 5 L−1. Heat treatment at 98°C reduced IN concentration about tenfold at warm temperatures but had no significant effect at temperatures colder than −12°C, as shown in Figure 4. The absence of an effect at colder temperatures suggests that the majority of IN at −20°C likely had a different source than the IN measured at warmer temperatures.
 Washings from the bulked mixture of vegetation from the ley field vegetation yielded 580,000 ina gene copies g−1 fresh weight and dry corn leaves in the adjacent field 820,000 g−1 fresh weight. The smaller population of bacteria with the ina gene on the ground corresponded to an absence of potential airborne INA bacteria from the cloning and sequencing results (Table 3). On this day, nearly 20% of clones were identified as plant material, and among the remainder no single species of bacteria or fungi dominated. For quantification of the ina gene copies from the atmospheric sample, DNA was extracted separately from two Biosamplers. In one, ina gene copies were detected by qPCR, but in only one out of nine reactions. No qPCR products were detected in DNA processed from the second Biosampler. The INA bacteria were thus detected, but not at a concentration high enough for quantification. As such, we ascribe an estimated concentration of bacteria with the ina gene of 0.04 L−1, equivalent to our limit of detection on this day. This value is comparable in magnitude to INA concentrations reported previously at agricultural fields [Constantinidou et al., 1990; Lindemann et al., 1982]. Given this value, INA bacteria made up at most ∼10% of the IN population at −10°C.
 Results from 20 October, when sampling nearer to the squash, were generally consistent with the previous measurements, and again suggested a biogenic contribution that is not dominated by INA bacteria. IN concentrations at −12°C were lower than on the previous occasion, at 0.13 L−1, but higher at −20°C, at nearly 7 L−1 using the CFDC. Heat treatment caused a ∼fivefold or greater reduction in IN down to the limit of testing at −15°C. Again bacterial, fungal and plant DNA were all present in the total aerosol sample from the Biosamplers, but on this date fungi were in greatest abundance (Table 3). No potential INA bacteria were observed, either via DNA sequencing or qPCR from the Biosampler collections. A spike of eight ina genes added to the qPCR was readily amplified and detected, indicating a detection limit of around 0.05 ina genes L−1 air, about half the number of IN L−1 air at −12°C.
3.2. Conventional Farm in Southwestern Nebraska
3.2.1. Cornfield During Harvest: 3 November 2010
 The final set of measurements came from a conventional farm near Maywood, Nebraska (40.55°N, 100.7°W). This particular corn crop varied from ones at many conventional farms in that it was not actively irrigated. For these measurements we sampled from south of the field, with winds out of the north northwest measured at 39–45 km h−1, gusting to 56 km h−1, at the nearest weather station. The combine harvester worked its way north, following east-west tracks back and forth across the field. Approximately 12 ha were harvested during the measurements. The corn harvesting process produced a cloud of dust and fragmented debris that was observed during each pass. A large fraction of the dust and biological particles generated during corn harvesting is expected to be larger than 2μm, with variability in PBAP composition with size [Lee et al., 2006]. Thus, differences between measurements made using the CFDC, which is limited to sampling particles smaller than 2.4 μm, and the Biosamplers, which should capture the entire size distribution, were to be expected.
 Atmospheric IN concentrations were higher than at other sites across the temperature range tested. Aliquots from two Biosamplers were analyzed using the drop freezing method. In both, the highest temperature of activity was −5°C, with a relatively steep increase in IN concentration with decreasing temperature, to a mean of 8 L−1 air at −12°C. Heat treatment of one of the Biosampler aliquots greatly reduced IN concentrations. At −7°C the reduction was 99.5% while at −14°C it was 98.5%. qPCR analysis from one Biosampler indicated a mean of 19.1 ± 6.7 (SE) ina genes L−1 air during the sampling period, in relatively good agreement with the drop freezing analysis (Figure 5), and ina genes were detected in abundance on the surrounding vegetation and soil surface (Table 2). Despite the presence of the ina gene predicted from the qPCR, no potential INA bacteria were observed from DNA sequencing of 53 clones (Table 4). The absence of potential INA bacteria in the sequencing data may be the result of the small sample size; more clones or more advanced methods are needed to fully capture the entire population.
Table 4. Genera of Bacteria Collected at Maywood, Nebraska, During Corn Harvest on 3 November 2010a
Total aerosol designates the sample collected using the SKC Biosampler. Biological IN designates IN collected on the glass slide impactor downstream of the CFDC. Identical techniques (DNA extraction, cloning, and sequencing) were utilized to identify the bacteria from both samples. When more than one, numbers in parentheses indicate the number of clones identified.
 At −20°C the IN concentrations as determined by the drop freezing method reached 53 L−1. Real-time CFDC IN data from this day are shown inFigure 6. The spikes in the IN concentrations correspond to time periods when the combine passed upwind of the mobile laboratory, with maximum particle concentrations present at these times. As shown, IN concentrations were highly variable, with maximum concentrations reaching nearly 50 L−1 when the combine was nearest the mobile laboratory. Average concentrations were 6 L−1, about nine times less than that determined from the drop freezing method, at −20°C. Given this large discrepancy, we infer that the drop freezing results were likely influenced by a significant population of larger particles that the CFDC was not able to measure (i.e., particles >2.4 μm).
 Measurements from this farm included the only identification of biological particles directly from the IN population processed by the CFDC. For these measurements, IN were collected utilizing the glass impactor at the CFDC outlet. The most abundant bacteria in the IN population captured on the CFDC glass impactor were Sphingobacteria, shown in Table 4, which are found in sediments [Vogel et al., 2009], agricultural fields [Mehnaz et al., 2007] and also atmospheric samples at sites in Colorado [Bowers et al., 2009, 2011]. The remaining phyla identified were from the Proteobacteria and Firmicutes. No known INA bacterial species were among the sequenced clones. The nondetection of known INA bacteria from the CFDC samples is surprising, given that the ina gene was detected in significant numbers via qPCR of the Biosampler particle suspensions, and the size of bacteria such as P. syringae should fall within the cut point of the CFDC. However, previous work has shown that bacterial bioaerosol particles are often present at larger diameters than a single bacterium [Lighthart et al., 1993], likely due to aggregates of bacteria or associated plant debris or other material. Given the nature of aerosol production during combine harvesting, it is likely that the INA bacteria in this case are associated with larger plant debris, making many particles too large for measurement using the CFDC.
4. Discussion and Conclusions
 Biological particles are ubiquitous in the atmosphere, with microbial concentrations often falling between 10 and 10,000 L−1 of air [Bowers et al., 2009; Huffman et al., 2010; Kasper-Giebl et al., 2002]. However, airborne bacteria that encode an ina gene, such as P. syringae, P. fluorescens, Pa. ananatis, Pa. agglomerans and X. campestris, are much less common [Bowers et al., 2009]. In this study, we quantified the INA bacteria and other biogenic particles in atmospheric IN populations from two locations thought to be potential sources of INA bacteria to the atmosphere: Grant Family Farms, a certified organic farm in Colorado, and a conventionally farmed corn field in Nebraska. Real-time IN measurements were made using a CFDC, and collected samples were postprocessed to characterize the temperature dependence for ice nucleation using the drop freezing method and to determine the number of airborne bacteria possessing theina gene using qPCR. Finally, 16S rDNA sequencing of the total aerosol population was used to identify biological aerosol particles at both farms, and in a single collection of IN captured by the CFDC.
 For all measurements conducted at Grant Family Farms, onset of freezing was observed at −6 to −7°C, and IN concentrations at −10°C ranged from 0.09 to 0.46 L−1 (30–40% fewer than at −12°C). Observed ice nucleation in this temperature range is often assumed to result from biological ice nuclei [Bowers et al., 2009; Christner et al., 2009]. These IN concentrations are twofold to tenfold higher than those observed by Bowers et al.  for five agricultural sites sampled in northern Colorado in summer. At −20°C, IN concentrations ranged from 5 to 20 L−1 measured by the drop freezing method and 1 to 13 L−1 measured by the CFDC. Despite the differences in the two methods, discussed above, and the fact that for the droplet freezing method many particles are pooled into an aqueous mixture, in all cases measurements from the CFDC were comparable to, albeit slightly lower than, those determined by the drop freezing method (factor of 1.5–5 at Grant Farms and a factor of 9 at the corn field in Nebraska). The lower concentrations in the CFDC likely resulted from differences in the aerosol sizes measured and differences in timescales for the two instruments. These concentrations span the range of typical IN concentrations at this temperature [DeMott et al., 2010]. Reduction in ice nucleation activity with heat treatment provided evidence for the predominance of biological/organic ice nuclei at Grant Family Farms. Ice nucleation activity following heat treatment decreased dramatically in all cases at warmer temperatures, and in all but one sample at −20°C. These data indicate not only the presence of apparent biogenic IN, but also that these particles likely play an important role as atmospheric IN at colder temperatures, where mineral and soil dusts are usually expected to dominate.
 In all cases at Grant Family Farms, the vegetation harbored large populations of bacteria with the ina gene, ranging from 6 × 105 to 2 × 107 per gram of vegetation. Despite this, and evidence from the drop freezing experiments that biological IN were present in the airborne samples, qPCR analysis found the ina gene to be at or below the detection limit for all samples collected from air at Grant Farms. These data imply a poor correlation between INA bacterial populations in the atmosphere and neighboring vegetation. For the one case in which the ina gene was detected, INA bacteria comprised no more than 10% of the IN at −10°C, suggesting that a large portion of the biological IN observed in this study do not have the ina gene, or that the ina gene had been digested in dead INA bacteria that still retained the protein. Compared to the total bacterial population expected for an agricultural site in this region, which we estimate to be 1500 L−1 based on previous measurements [Bowers et al., 2011], we estimate INA bacteria make up only 0.002% of all airborne bacteria for the specific sampling conditions at this site. Although this fraction is lower than previous estimates [e.g., Constantinidou et al., 1990; Lindemann et al., 1982], these previous studies estimated total bacterial counts based on culturing methods, rather than the DNA staining methods used by Bowers et al. , which likely impacted their calculated fractions. DNA sequence analysis of total PBAP revealed the presence of potential INA bacteria, such as P. syringae and P. fluorescens, making up ∼4% of the 171 DNA sequences analyzed. Given that only a fraction of these strains will possess the ina gene, the sequencing results also suggest a contribution of INA bacteria to the total population of less than 1%, consistent with the qPCR results during sampling at Grant Family Farms.
 Leaf washing from the Nebraska corn field once again indicated large populations of bacteria with the ina gene both before and after the harvest. However, at this corn field, the action of the combine harvester actively injected detritus into the atmosphere, and concentrations of airborne bacteria with the ina gene were at the highest levels from any field, at ∼19 L−1. These high concentrations, combined with the enhanced IN concentrations measured with the CFDC when the instrument was exposed to particles generated from the combine harvester (15 L−1 average IN concentration when sampling within the harvester plume; 4.6 L−1 outside of the plume) suggest that such agricultural activities likely enhance atmospheric IN emissions from vegetation and soil, as has been observed for other biological particles generated during harvesting [Lee et al., 2006; Lighthart, 1984]. The ina gene concentration was consistent with, albeit slightly higher than, measured IN concentrations from the drop freezing method at −12°C, the temperature at which a single Ina protein can nucleate (Figure 5). Finally, heat treatment resulted in ∼98% reduction in IN numbers at −12°C. Combined, these data suggest that most of the warm temperature nucleation was the result of INA bacteria on this day.
 Compared to all sites at Grant Family farms, IN concentrations from the corn field were significantly higher at −20°C based on drop freezing method, with an average value of 53 L−1. These data suggest that another population of IN was active at colder temperatures, likely also tied to the mix of particles generated by the combine. Further, a reduction in IN activity measured at colder temperatures following heat treatment suggests many of these IN at −20°C were biological. Some of these potential biological particles that nucleated ice at colder temperatures, but did not have the ina gene, were identified by directly extracting and sequencing DNA from ice nucleating particles generated during harvesting and collected using a glass impactor downstream of the CFDC. Sphingobacteria were found to be the most abundant biological particles in our IN sample. These bacteria do not express an Ina protein, and so the mechanism for nucleation is not known. One possibility involves interactions with minerals from soil, in which the IN detected may have been the result of bacteria and minerals bound together as a single particle. IN composed of internally mixed PBAP and mineral dust have been inferred previously [Pratt et al., 2009].
 Considering all measurements, bacteria with the ina gene were only observed in relatively large numbers for air samples collected during combine harvesting. These particles appear to have been quite large (>2.4 μm), and so may have been associated with large plant fragments and soil dust generated from the action of the combine. At Grant Family Farms, IN concentrations were lower, and bacteria with the ina gene were at or below the detection limit, and below the concentration of IN at −12°C for the four cases studied. Despite the relative absence of INA bacteria, the data suggest the abundant presence of apparent biological IN [Conen et al., 2011; Schnell and Vali, 1976]. These biological particles remain to be completely identified.
 This work was funded by the NSF (ATM-0841602). E. Garcia was supported by the NSF (ATM-0919042). The authors wish to thank Lew Grant and the staff of Grant Family Farms, Dionne Cozzola, and Lowell Schmidt for providing access to the farms for measurements. The authors are also indebted to Dimitri Georgakopoulos for preliminary experiments during a 2008 Visiting Scholarship from the Fulbright Foundation that helped guide the present studies.