Emigration of the plant pathogen Pseudomonas syringae from leaf litter contributes to its population dynamics in alpine snowpack

Emigration of the plant pathogen Pseudomonassyringae from leaf litter contributes to its populationdynamics in alpine snowpackemi_2680 1..14

Caroline L. Monteil,1 Caroline Guilbaud,1

Catherine Glaux,1 François Lafolie,2

Samuel Soubeyrand3 and Cindy E. Morris1*1INRA, UR407 de Pathologie Végétale, Domaine StMaurice, BP. 94, 84140 Montfavet, France.2INRA, UMR 1114 EMMAH, Domaine Saint-Paul, SiteAgroparc, 84914 Avignon, France.3INRA, UR546 Biostatistique et Processus Spatiaux,Domaine Saint-Paul, Site Agroparc, 84914 Avignon,France.

Summary

The recently discovered ubiquity of the plant patho-gen Pseudomonas syringae in headwaters and alpineecosystems worldwide elicits new questions aboutthe ecology of this bacterium and subsequent conse-quences for disease epidemiology. Because of themajor contribution of snow to river run-off duringcrop growth, we evaluated the population dynamicsof P. syringae in snowpack and the underlying leaflitter during two years in the Southern French Alps.High population densities of P. syringae were foundon alpine grasses, and leaf litter was identified as themain source of populations of P. syringae in snow-pack, contributing more than the populations arrivingwith the snowfall. The insulating properties of snowfoster survival of P. syringae throughout the winter inthe 10 cm layer of snow closest to the soil. Litter andsnowpack harboured populations of P. syringae thatwere very diverse in terms of phenotypes and geno-types. Neither substrate nor sampling site had amarked effect on the structure of P. syringae popula-tions, and snow and litter had genotypes in commonwith other non-agricultural habitats and with crops.These results contribute to the mounting evidencethat a highly diverse P. syringae metapopulation isdisseminated throughout drainage basins betweencultivated and non-cultivated zones.

Introduction

The development of disease management practices hasbeen almost exclusively based on the biology, ecologyand genetics of plant pathogens in relation with the crophost or the environment surrounding the crop (agriculturesensu stricto). However, plant pathologists are realizingthe insufficiency of this agro-centric view and its incom-patibility with sustainable agriculture. Extending the para-digms of plant pathogen life history and evolution ofparasitic fitness beyond agricultural boundaries is seen asa means to an alternative view of plant disease epidemi-ology (Morris et al., 2009). Recent publications on the lifehistory of the plant pathogen Pseudomonas syringae arethe first to formally link the ubiquity of a plant pathogen innon-agricultural substrates to a life history framework wellbeyond agricultural contexts, in this case in close asso-ciation with the water cycle (Morris et al., 2008; 2010). Inaddition to cultivated plants, P. syringae is frequentlyfound in rain and cloud water, in alpine rivers, in wildplants and in seasonal snowpack. Headwaters worldwideharbour P. syringae populations with different combina-tions of phenotypes in different climatic and geologic con-texts. The whole of these observations has raised novelquestions about how each substrate of the water cyclestructures the pathogenic population and how transferfrom one substrate to another can affect the populationdynamics and diversification of P. syringae.

In this present study we have focused on the sub-alpineenvironment and especially on the fate of P. syringae insnowpack covering leaf litter. Snowfall and the resultingsnowpack are immense reservoirs of water and thereforeare potential reservoirs for the bacterium. Snowfalls rep-resent 12.5 ¥ 103 km3 of water per year, which corre-sponds to between one-quarter and one-third of the totalannual freshwater discharge (Oki and Kanae, 2006). Incertain valleys of the USA, such as the Green Valley inColorado, snowfalls can constitute as much as 80% oftotal precipitation (Brooks and Williams, 1999). Approxi-mately one-third of the Northern Hemisphere land area(40 million km2) is blanketed with snow cover for at least3 months per year. Moreover, in middle and high latitudeareas, snowmelt is a major source of runoff for irrigatingcrops. In Europe in particular, alpine headwater streams

Received 26 August, 2011; revised 17 November, 2011; accepted 22November, 2011. *For correspondence. E-mail [email protected]; Tel. (+33) 4 32 72 28 41; Fax (+33) 4 32 72 28 42.

Environmental Microbiology (2011) doi:10.1111/j.1462-2920.2011.02680.x

© 2011 Society for Applied Microbiology and Blackwell Publishing Ltd

mailto:[email protected]

mailto:[email protected]

contribute water to four major rivers, the Danube, Rhine,Rhone and Po, draining a total land area of 1.1 millionkm2 entering coastal waters and discharging 4.37 ¥ 1015

litres of water annually (see review by Edwards et al.,2007).

Numerous studies have reported on microbial presenceand activity in glacial ice and snowpack (reviewed in detailby Hodson et al., 2008). In stark contrast, few studieshave considered non-psychrophilic microbial communi-ties in seasonal snow in sub-alpine and alpine regions.Furthermore, the natural occurrence and growth of meso-philic bacteria, considered to be associated primarily withother ecosystems, have never been explored in theseenvironments. After artificial inoculation, Reynolds andRingelberg (2008) showed a successful transfer of P. sy-ringae inoculum from snow to soil. Their results suggestthat both indigenous and non-indigenous microbial popu-lations could survive the winter season in association withsnowpack. Furthermore, P. syringae has a measurabledoubling time at 0°C (about 22 h) (Young et al., 1977)typical of psychrophiles.

Much attention has been paid to the role of snowpack inthe composition and functioning of the underlying soilmicrobial communities (Zinger et al., 2009) and its impacton plant litter decomposition (see review by Gavazov,2010), gas fluxes (see review by Williams et al., 2009b)and biogeochemical cycles (Brooks and Williams, 1999;Lipson et al., 1999; Schmidt and Lipson, 2004; Williamset al., 2009a). Some of these studies have reported a riseof bacterial biomass in winter and a high heterotrophicmicrobial activity under the snow. This phenomenonoccurs because snowpack can maintain unfrozen condi-tions at the soil-snow interface due to its capacity toeffectively insulate (Brooks et al., 1996). Energy balancemeasurements of the soil-snow-atmosphere system andexperimental observations have shown that a snow depthof 30–40 cm dissociates soil temperature from air tem-perature (Cline, 1995; Brooks and Williams, 1999). Micro-scopic films of free water at the surface of soil particlesand substrate availability foster the proliferation of largeand diverse cold-adapted microbial communities (Leyet al., 2004). Thus snowpack engenders conditions thatcould permit the survival and diversification of residentmicroorganisms including P. syringae.

Based on these assumptions and on our previousobservation of P. syringae in snowpack (Morris et al.,2008), our objective was to elucidate the populationdynamics of P. syringae in the snowpack at sites repre-senting the drainage basin of the Southern French Alpsand to evaluate the contribution of snowfall to the popu-lations in snowpack. We observed a marked density gra-dient of bacterial populations in snowpack, which led us toinvestigate the role of the underlying ground cover in thepopulation dynamics in snowpack and thereby to reveal

the importance of alpine prairie leaf litter as an unsus-pected reservoir of P. syringae.

Results

P. syringae populations in snowfalls, alpine grasses andlitter all contribute to the presence of this bacterium inthe drainage basin

Prior to winter snowfalls, average population densities ofP. syringae in senescent plants and leaf litter were3.39 ¥ 109 and 8.51 ¥ 108 colony-forming units (cfu) m-2 ofground surface, respectively, and generally constitutedwell below 1% of the total mesophilic population (Fig. 1A).The population densities of P. syringae were higher insenescent plants than in the leaf litter (P < 0.05), but thepopulation densities of total culturable bacteria were lowerin the phyllosphere than in the litter (P < 0.05).

Ten per cent of all of the snowfalls occurring in thedrainage basin during the two winter seasons between2008–2010 had detectable populations (> 5 cfu l-1) ofP. syringae whereas nearly all (92%) of the snowfalls con-tained detectable levels of total mesophilic bacteria(Fig. 1B). When detected, P. syringae densities in freshsnowfall were 80 cfu l-1 to 1.66 ¥ 104 cfu l-1 of meltedsnow, whereas total mesophilic bacteria were present at166 cfu l-1 to 9.88 ¥ 108 cfu l-1.

P. syringae survives in snowpack during the winter andthe population density is stratified

Pseudomonas syringae was systematically detected inthe snowpack of all sites at all dates. Throughout thesampling periods the study sites were characterized by anabiotic environment that was stable in terms of pH(5.2–6.2 for melted snow) and electrical conductivity(0.4–6.5 mS cm-1). Population densities of P. syringaeranged from 44.6 cfu l-1 to 1.11 ¥ 105 cfu l-1 of meltedsnow and represented 0.008% to 8% of the total meso-philic bacterial population. The sampling of triplicates ofeach snow layer at the Ceillac and Lautaret sites revealedthat there was a significant effect of depth in the snowpack on the population density of bacteria (Fig. 2). Onaverage, total mesophilic bacteria and P. syringae werenearly 1000 times more abundant in the 10 cm layer ofsnowpack that was in contact with the ground than in thelayers above. This stratification along the snowpack wassystematically observed at all sampling sites and dates[14 observations in addition to those for which triplicateswere sampled from the Ceillac and Lautaret sites(Table S1)]. Whereas P. syringae was always detected inthe first layer in contact with the ground at densities from102 to 106 cfu l-1, it was not always detected in the otherlayers and particularly for the snow on the top layer (from

2 C. L. Monteil et al.

© 2011 Society for Applied Microbiology and Blackwell Publishing Ltd, Environmental Microbiology

30 cm above ground level to the top) even though totalculturable bacteria were almost always detected.

Emigration of P. syringae from litter to the snowpackcontributes to the stratified population density inthe snowpack

To compare population densities of total mesophilic bac-teria and of P. syringae in snow and litter, samples werecollected from March to April 2010 over a delimited

surface. As presented in Fig. 3, triplicate samples fromCol du Lautaret and Ceillac showed a decreasing densitygradient of P. syringae and total bacterial microflora(expressed as cfu m-2) that was also systematicallyobserved for the four others dates and sites where onlysingle samples were collected (Table S1). In the underly-ing litter, the mean population densities across all siteswere 2.77 ¥ 1011 cfu m-2 of ground for total mesophilicbacteria and 3.83 ¥ 108 cfu m-2 for P. syringae. Thesevalues, which represent population sizes after establish-ment of the snowpack, were not significantly different fromdensities in leaf litter determined before the first snowfalls

Fig. 1. Population densities of P. syringae and total culturablebacteria in alpine grasses, litter and snowfalls in the drainage basinbefore and during snowpack formation.A. Mean population sizes in the leaf litter (grey bars) andsenescent plants (white bars) for all sampling sites in September2009 and September 2010. Population densities on leaf litter weresignificantly different from those on senescent plants for each ofthe populations (P. syringae and total bacteria) enumerated(P � 0.05) as indicated by the letters associated with the bars.B. Number of snowfalls in which P. syringae (white bars) and totalmesophilic bacteria (grey bars) were detected or below thedetection threshold (5 bacteria per litre) (ND) from November toMarch 2008–2009 and 2009–2010. Error bars representedstandard error.

Fig. 2. Population densities of P. syringae and total mesophilicbacteria in two layers of snowpack in the Southern French Alps.For sites where triplicates were sampled (A) Col du Lautaret and(B) Ceillac, values represented the mean of three replicate samplesfor the first 10 cm of snow (grey bars) and from 10 cm to the top ofthe snowpack (white bars) at each of the sampling dates (March2010 for Col du Lautaret, and April 2010 for Ceillac). For eachbacterial population (P. syringae or total) the values associated withthe same letter are not significantly different (P � 0.05). Error barsrepresent standard error.

P. syringae in snowpack 3


in September 2010 (t-test, P < 0.05). For all sites pooled,population densities of P. syringae varied between4.03 ¥ 105 and 5.74 ¥ 1010 cfu m-2 for the leaf litter,1.11 ¥ 104 and 2.62 ¥ 108 cfu m-2 for the 10 cm of snowin contact with the ground, and 4.74 ¥ 102 and1.13 ¥ 107 cfu m-2 for the upper 20 cm of snow.

To determine if a transfer of bacterial populations fromlitter to snowpack was possible, two sets of microcosmswere set up in 2010 with litter and snow and buried at the

Col du Lautaret site. Although no P. syringae was detect-able in the snow at the time the microcosms were con-structed, approximately 107 P. syringae per m2 of snowwere detected at the first sampling date (3 weeks ofincubation) (Fig. 4). Thereafter, population sizes in snowand litter were stable over the 9 weeks until the end of theexperiment. Those in the litter were about 1010 cfu m-2. NoP. syringae was detected in microcosms without leaf litterat any of the sampling dates. The ratio between popula-tion densities in the two substrates showed that about0.01% of the population of P. syringae from litter trans-ferred to the snow if we assume that there was no growthof P. syringae in the snow during the period of incubation.

There is no consistent effect of substrate, site orsampling date on phenotypic structures of P. syringaepopulations in alpine prairies

Phenotypes of strains from all litter and snow sampleswere assessed. A wide range of intensities of the differentphenotypes was observed leading us to rank each phe-notype into a range of categories. The capacity to inducea hypersensitive reaction consisted of two categories(HR- or HR+). Aggressiveness on cantaloupe consisted ofthree categories: low (AGR1, less than half of the cotyle-dons showed symptoms or if more than half showedsymptoms the severity score was < 1), intermediate(AGR2, more than half of the cotyledons showed symp-

Fig. 3. Population densities of P. syringae and total mesophilicbacteria in litter and two layers of the covering snowpack insub-alpine meadows of the Southern French Alps for sites wheretriplicates samples of litter and snow were sampled. At (A) Col duLautaret and (B) Ceillac, values represent the mean of 3 replicatesamples for leaf litter (hatched bars), the first 10 cm of snow (greybars) and the overlaying 20 cm of snow (white bars) at two dates inMarch 2010 for Col du Lautaret and April 2010 for Ceillac.Densities were expressed in terms of numbers per surface ofground cover. For each bacterial population (P. syringae or total)the values associated with the same letter are not significantlydifferent (P � 0.05). Error bars represent standard error.

Fig. 4. Dynamics of population densities of P. syringae in litter(dashed line) and snow (solid line) in microcosms incubated for 9weeks under snowpack in the Southern French Alps. This dataillustrate the transfer of P. syringae from the litter to the snow.Values represent mean population densities for 3 replicates foreach of the two sampling date. The population density for the firstsample in snow was below the detection threshold (5 cfu l-1 ofsnowmelt which represents 4.6 ¥ 103 cfu m-2). For each variable,values were not significantly different (ANOVA, P � 0.05). Error barsrepresent standard error.



toms and the mean disease score was between 1 and 2),and high (AGR3, more than half of the cotyledons showedsymptoms and the mean disease score was > 2). Icenucleation activity consisted of three categories: weak(INA1, no ice nucleation activity at -8°C or warmer among106 cells), moderate (INA2, ice nucleation activity between-7°C and -5°C in a population of 106 cells), and intense(INA3, ice nucleation activity between -4°C and -2°C ina population of 106 cells). Intensity of production of asyringomycin-like toxin was classed into four categories:weak (SYR1, zone of inhibition < 2 mm), low (SYR2, zoneof inhibition between 2 and 6 mm), moderate (SYR3, zoneof inhibition between 6 and 10 mm) and intense (SYR4,zone of inhibition � 11 mm). Overall, about 67% of strainsinduced hypersensitivity on tobacco. Most (81%) of thestrains were ice nucleation active at -8°C or warmer andfor some samples between 20% and 67% of strains werein the INA3 class. On average, 14.6 % of strains werepathogenic on cantaloupe (frequency of cotyledonsinfected > 0.5) and 75% of these strains induced symp-toms on cantaloupe with a high average of intensity(AGR2,3). About 60% of the strains producedsyringomycin-like toxins, resulting in a zone of antibiosisof at least 3 mm (SYR>1) and in some cases up to 15 mm.

The frequency of occurrence of strains with each of thedifferent phenotypes varied with substrate, site and date(Table 1). Although there were significant differences inthe frequency of occurrence of the different phenotypesdue to substrate in some cases, these differences werenot consistent (Table 1). For the 14 sets of comparisons(different dates for the different sites) we made amongsubstrates (between layers of snow or among layers ofsnow and litter), substrate had a significant structuringeffect on the frequency of any of the individual pheno-types in only half or fewer of the cases. Substrate had nosignificant effect on the occurrence of highly aggressivestrains (AGR3), but there was a trend for the frequency ofhighly aggressive strains to be greater in leaf litter than insnow (in 4/5 comparisons).

We also determined the frequency of occurrence of thecombinations of the four phenotypes assessed for the 986strains characterized here (Fig. 5). A total of 45 combina-tions of the possible 72 combinations of the 4 phenotypesand their different intensities were observed. To determineif there was a significant effect of substrate (litter or snow)on the frequency of the different combined phenotypes weused a generalized linear model (GLM). For this approachwe used the combined phenotypes corresponding to the712 strains isolated from litter or the first 10 cm layer ofsnow. The observed frequencies of occurrence for eachcombination were compared with the expected frequencyunder the null hypothesis of no difference in frequency ofcombined phenotypes due to substrate. Figure 6 illus-trates an example of phenotypic structure comparison

with the GLM approach. The grey histogram representsthe observed frequencies for the phenotypic combinationsrelative to the probability that they are in the 0–10 cmlayer of snow. The white histogram represents the confi-dence interval of the frequency of occurrence for eachcombination to be in the snow for the entire collection ofstrains from litter and the 0–10 cm layer of snow under thenull hypothesis. A significant structuring effect of eithersnow or litter is revealed when the frequency of occur-rence of the phenotype is outside of the confidence inter-val. Hence, for the grey bars labelled ‘S’, the associatedphenotypes are more significantly frequent in snow thanexpected and for the bar labelled ‘L’ the associated phe-notypes are more frequent in litter than expected. Thisapproach was used to evaluate the effect of date and sitein addition to the effect of substrate.

Results of the whole GLM comparisons are presentedin Table S2. Overall, there was no consistent effect ofsubstrate, site or date on the combined phenotypic popu-lation structure. All the tests showed at least one com-bined phenotype for which differences were significant,but this was not consistently a particular phenotypic com-bination for substrate, site or date. All samples consistedof a mix of dominant and rare phenotypes and showed noconsistent trend in population structure. The most domi-nant combined phenotype in the collection [AGR1, HR+,SYR1,2,3,4, INA2,3] was found in all samples where leaf litteror snow from the 0–10 cm layer were collected and thesecond most dominant combined phenotype [AGR1, HR-,SYR1,2,3,4, INA1,2,3] was found in 66% of these samples.

P. syringae populations in leaf litter and snow aregenetically diverse, comprised of pandemic strains alsofound in water and crops and strains endemic tonon-agricultural habitats

Assessment of genotypic diversity by BOX-PCR revealed45 different BOX-PCR profiles at Ceillac, 11 at Col deVars, 18 at Col du Lautaret and 21 at Super Sauze. Withina site, some profiles were specific to the snow or litter, andabout 20–50% were shared between substrates and/ordates. An example of BOX-PCR profiles is shown inFig. S1. This high number of different genetic profiles wasnot readily compatible with statistical tests to quantify thesignificance of the differences in structure observedbetween snow and litter (Morris et al., 2002). Therefore,the effect of substrate on the genetic structure of P. syrin-gae populations was evaluated by a population geneticsapproach based on the sequence of the cts housekeepinggene.

The cts sequences of the 290 strains from snow andlitter collected in this study and those from 323 water andcrops strains from a previous study (Morris et al., 2010)represented 138 unique haplotypes, 59 of which were



Table 1. Frequency of strains of P. syringae from leaf litter and snow in different phenotypic classes.

Year Month Sitea Substrate

Frequency of strains in each classb

HR+c INA+d INA3e AGR+f AGR>1g AGR3h SYR+i SYR>2j

2009 March Ceillac Snow 0–10 cm 0B 0.95A 0A 0.14A 0A 0A 1.00A 0.77A

Snow 10–30 cm 0.60A 1.00A 0.20A 0.2A 0.07A 0A 0.80A 0.27B

May Vars Snow 0–10 cm 0.36 1.00 0.23 0.23 0.04 0 0.95 0.76Sauze Snow 0–10 cm 0.90 1.00 0.05 0.62 0.38 0.09 0.81 0.67Ceillac Snow 0–10 cm 1.00 1.00 0.45 0.15 0 0 0.15 0.05Sauze Snow 0–10 cm 0.50A 0.80A 0.15A 0.35A 0.10A 0A 0.20A 0.10A

Snow 10–30 cm 0.33A 0.89A 0.17A 0.05B 0.05A 0A 0.05A 0.05A

2010 March Ceillac Leaf litter 0.42A 0.61B 0.12A 0.05A 0.03A 0.02A 0.55B 0.28A

Snow 0–10 cm 0.57A 0.90A 0.25A 0.15A 0.15A 0.1A 0.9A 0.55A

Vars Leaf litter 0.98A 0.78A 0.22A 0.12A 0.10A 0A 0.05A 0.17A

Snow 0–10 cm 0.92A 0.35B 0.05A 0A 0.10A 0A 0.05A 0.05A

Sauze Leaf litter 0.92A 1.00A 0.36A 0.45A 0.32A 0.03A 0.93A 0.83A

Snow 0–10 cm 0.69B 0.70B 0.30A 0.05B 0B 0A 0.85AB 0.60AB

Snow 10–30 cm 0.75AB 0.65B 0.15A 0.05B 0B 0A 0.60B 0.30B

April Ceillac Leaf litter 0.79B 0.82B 0.28A 0.05A 0.03A 0.02A 0.52A 0.23B

Snow 0–10 cm 0.98A 0.95A 0.17A 0,00A 0,00A 0,00A 0.60A 0.25B

Snow 10–30 cm 0.92A 0.76B 0.03B 0.07A 0.02A 0,00A 0.63A 0.28B

Snow > 30 cm 0.80B 0.90AB 0.25A 0.03A 0.02A 0,00A 0.68A 0.43A

Vars Leaf litter 0.95A 0.88A 0.12A 0.64A 0.47A 0.10A 0.81A 0.78A

Snow 0–10 cm 0.53B 1.00A 0.30A 0.10B 0.05B 0.05A 0.90A 0.45B

Sauze Leaf litter 0.28B 0.47B 0.07B 0.28A 0.10A 0,00A 0.38B 0.17B

Snow 0–10 cm 0.67A 0.52B 0.14AB 0.05B 0,00A 0,00A 0.50B 0.30AB

Snow 10–30 cm 0.40AB 0.95A 0.25A 0.05B 0.05A 0,00A 0.85A 0.50A

2010, replicates March Lautaret (r1) Leaf litter 1.00A 1.00A 0.40A 0A 0A 0A 0.95A 0.70A

Snow 0–10 cm 1.00A 1.00A 0.35A 0A 0A 0A 1.00A 0.75A

Snow 10–30 cm 1.00A 0.94A 0.35A 0A 0A 0A 0.82B 0.58A

Snow > 30 cm 1.00A 1.00A 0.30A 0.05A 0A 0A 1.00A 0.75A

(r2) Leaf litter 1.00A 0.85A 0.50A 0A 0A 0A 0.80A 0.55A

Snow 0–10 cm 1.00A 1.00A 0.15B 0.05A 0.05A 0.05A 0.95A 0.60A

(r3) Leaf litter 1.00A 1.00A 0.20A 0A 0A 0A 0.47B 0.26B

Snow 0–10 cm 1.00A 1.00A 0.45A 0A 0A 0A 0.85A 0.60A

Snow 10–30 cm 1.00A 1.00A 0.50A 0A 0A 0A 1.00A 0.75A

Snow > 30 cm 1.00A 1.00A 0.44A 0A 0A 0A 1.00A 0.66A

April Ceillac (r1) Leaf litter 0.87AB 0.95A 0.25A 0A 0A 0A 0.55A 0.25A

Snow 0–10 cm 0.97A 0.90A 0.25A 0A 0A 0A 0.70A 0.35A

Snow 10–30 cm 0.93AB 0.85A 0.25A 0.10A 0.05A 0.05A 0.50A 0.10B

Snow > 30 cm 0.73B 1.00A 0.20A 0.05A 0A 0A 0.20B 0B

(r2) Leaf litter 0.80A 0.70B 0.35A 0.10A 0.1A 0.05A 0.70A 0.25AB

Snow 0–10 cm 0.97A 1.00A 0.15A 0A 0A 0A 0.65A 0.15B

Snow 10–30 cm 0.87A 0.84AB 0.16A 0.10A 0A 0A 0.70A 0.25B

Snow > 30 cm 0.85A 1.00A 0.45A 0A 0A 0A 0.90A 0.70A

(r3) Leaf litter 0.70B 0.80A 0.25A 0.05A 0A 0A 0.30C 0.10BC

Snow 0–10 cm 1.00A 0.95A 0.10A 0A 0A 0A 0.45BC 0.20B

Snow 10–30 cm 1.00A 0.55A 0.05A 0A 0A 0A 0.70AB 0.50AC

Snow > 30 cm 0.85AB 0.70A 0.10A 0.05A 0.05A 0.05A 0.95A 0.60A

a. The specific location of sites is indicated13. In 2010, the three replicates of snow from each layer collected at Ceillac and Lautaret at the samedate are labelled r1, r2 and r3.b. For each site and each date, Fisher tests (using the integer number of strains tested) were conducted to evaluate if there was a significant effectof substrate (snow layers and litter) on the frequency of strains for each phenotype. Values followed by the same letter are not significantly differentfrom the frequencies for the other substrates for the same sampling date and site. At some sampling dates only one substrate was available andhence no comparisons could be made at these dates.c. Frequency of strains capable of inducing a hypersensitive reaction on tobacco.d. Frequency of strains that were ice nucleation active at -8°C or warmer in a test of 106 cells.e. Frequency of strains that were ice nucleation active between -4°C and -2°C in a test of 106 cells.f. Frequency of strains inducing symptoms on more than half of the cotyledons and having a mean score severity � 1.g. Frequency of inducing symptoms on more than half of the cotyledons and having a mean score severity > 2.h. Frequency of strains for which the mean disease severity induced on cantaloupe seedlings was > 3.i. Frequency of strains for which there was a detectable zone of inhibition (> 0 mm) of G. candidum.j. Frequency of strains producing a zone of antibiosis against G. candidum > 2 mm.



each represented by a single strain, and the dominanthaplotype was represented by 81 strains. Among the 290strains from snow and litter considered apart, there were64 unique haplotypes. Bayesian analysis of the 138unique haplotypes revealed that they were most optimallygrouped into 3 clusters based on the K statistic of Evannoand colleagues (2005). A subsequent principal compo-nent analysis (PCA), followed by clustering based onWard’s method as conducted in a previous study (Morriset al., 2010) also revealed three clusters of P. syringaestrains. Axes 1 and 2 of the PCA accounted, respectively,for 28.3% and 16.6% of the total genetic variability.Assuming a threshold of Q = 0.8 for assignment to acluster, about 93% of the distinct haplotypes clearlybelonged to only one cluster, indicating that the threeclusters were highly divergent.

The three clusters represented strains with a specificrange of affiliation in the phylogenetic tree designed byMorris and colleagues (2010). Cluster 1 representedstrains from the Group 2a to Group 3 clades in the tree asdepicted by Morris and colleagues (2010), Cluster 3 rep-resented strains in the Group 3 clade to the UB-246 cladeand Cluster 2 represented only strains from the CC1524and UB-246 clades as well as 2 strains that could not beattributed to the previously described clades but had theclosest genetic distance to the UB-246 clade relative to allother clades. Strains in Cluster 1 and 3 were ubiquitous,present in crops, water, leaf litter and snow (Fig. 7), whileCluster 2 contained haplotypes only found in leaf litter,snow and water but not in crops. All three clusters wererepresented in the snow and leaf litter at all samplingsites.

Fig. 5. Phenotypes of P. syringae and their frequency distributionamong strains from leaf litter and the entire snowpack in the UpperDurance drainage basin in the Southern French Alps. Relativeabundances of strains were represented for each of the observedcombined phenotype for a collection of 986 strains characterizedfor each of the four phenotypes. AGR: Aggressiveness wasdetermined on cantaloupe ranked in three class; class AGR1 (whitesquares) included avirulent strains and those which the severityscore was < 1; AGR2 (grey squares) strains for which more thanhalf of the cotyledons showed symptoms and the mean diseasescore was between 1 and 2; AGR3 (black squares) strains for whichmore than half of the cotyledons showed symptoms and the meanscore disease > 2. HR: Black squares indicate strains that induceda necrosis on tobacco leaves. SYR: Four classes were consideredfor the production of syringomycin-like toxins represented bysquares with an increasing gradient of grey from the leastproductive (SYR1 in white) to the most productive (SYR4 in black).The class SYR1 included strains induced a zone of antibiosisagainst Geotrichum candidum � 2 mm. The zone of inhibition (inmm) induced by strains in classes SYR2, SYR3 and SYR4 wererespectively 2–6, 6–10 and � 11. INA: Ice nucleation activity wasdetermined with a population of 106 cells and consisted of threecategories; strains in class INA1 (white squares) were not icenucleation active at -8°C or warmer; INA2 (grey squares) showedice nucleation activity between -7°C and -5°C; and INA3 (blacksquares) showed ice nucleation activity between -4°C and -2°C.



Discussion

The ubiquitous species P. syringae has a life historyclosely linked to the hydrosphere through which water isnaturally circulating and stocked in reservoirs (Morriset al., 2008). The bacterium is found in several compart-ments of surface freshwater (Morris et al., 2008; 2010), incrop and wild plants (Hirano and Upper, 2000; Mohr et al.,2008; Morris et al., 2008), in aerosols and clouds (Linde-mann et al., 1982; Amato et al., 2007). However, the roleand significance of dead plant material of the soil–vegetation interface as a durable habitat has not beeninvestigated. One of the few works mentioning the pres-ence of P. syringae in decomposing vegetation was byMaki and colleagues (1974) whose strains were isolatedfrom decaying alder leaves (Alnus tenuifolia) leading tothe discovery of the ice nucleation activity of this bacte-rium. Crop debris has also been recognized as a reservoirof P. syringae for some disease epidemics but the debris-associated populations decline rapidly (McCarter et al.,

1983; van Overbeek et al., 2010). In our study, throughthe systematic detection of P. syringae in leaf litter ofalpine meadows, we have shown that dead plant materialis a common habitat. Populations of P. syringae were inthe size range of populations in senescent vegetation inthe prairies where litter was collected (103 to 107 cfu g-1 oftissue). Furthermore, these populations were not reducedin size by the snow cover in the winter.

The survival of P. syringae in such conditions raisesquestions about the role of snowpack as a protectiveenvironment for plant pathogen populations at the soilsurface. It is well known that the insulating properties ofthe snow permit the overwintering of microbial communi-ties and the realization of biogeochemical cycles in theseveral centimetres of soil near the surface (Brooks et al.,1996; Lipson et al., 1999; Ley et al., 2004; Schmidt andLipson, 2004). Most of the microbial communities adaptedto cold temperatures studied to date utilize complex sub-strates such as cellulose or vanillic acids, and simpleamino acids to a lesser extent (Lipson et al., 2002). Thismetabolic activity is essentially due to fungi including theplant pathogens known as snow moulds, which constitutethe dominant microorganisms (Lipson et al., 2002;Schmidt et al., 2009). The ability of these active fungi todegrade complex substrates in such environments couldbenefit Gram-negative bacteria such as P. syringae byproducing less complex nutrients. This relation was pre-

Fig. 6. Example of phenotypic structure comparison with the GLMapproach: frequency distribution of the probability for eachcombination of phenotypes to be in the first layer of snow. Greybars represent the observed frequencies (Fobs) of the phenotypiccombinations. The white area represents the confidence interval ofthe frequency distribution for 2000 permutations of the phenotypiccombinations in the whole data set. The thick vertical black linerepresents the mean of this distribution. By comparing theobserved frequencies with the confidence intervals of the expectedfrequencies, significant effects of the substrate on the frequency ofoccurrence of different phenotypic combinations can be identified.In this example, combinations whose probability > 0.5 tend to bemost represented in the first layer of snow, and those < 0.5 in theleaf litter. The frequency associated with ‘L’ was significantly higherin the litter (in this example this corresponds to [AGR1, HR-, SYR1,INA1]), and frequencies associated with ‘S’ were significantly higherin the snow (here they correspond to the phenotypes [AGR1, HR-,SYR2, INA2] and [AGR1, HR-, SYR4, INA2]).

Fig. 7. Distribution of cts haplotypes of P. syringae into threegenetic clusters inferred with the Bayesian clustering algorithm ofthe STRUCTURE software. The frequency of haplotypes assignedto each of the clusters is indicated for haplotypes of strains fromcrops (hatched bars), from leaf litter (solid black bars), fromsnowpack (dark grey bars) and of strains from water (light greybars). Of the 138 haplotypes analysed, 128 were assigned toclusters with probabilities of inferred ancestry between 0.75 and0.99. Eight haplotypes were assigned at equal probabilities (c.0.50) to clusters 1 and 3, and two haplotypes were assigned atequal probabilities to clusters 2 and 3.



viously demonstrated experimentally by the inoculation ofleaf litter with the soft rot pathogens Pectobacterium caro-tovorum (formerly Erwinia carotovora) and Aspergillusnidulans. Proteome analyses coupled to analyses ofextracellular degradative enzyme activities suggestedthat the bacterium had limited capabilities to degrade leaflitter and profited from the degradation products of thefungi (Schneider et al., 2010). Furthermore, the antifreezeactivity of snow moulds could modify the physical envi-ronment of the bacterial community by limiting ice forma-tion (Snider et al., 2000), thereby leading to greateravailability of free water. Temperature, moisture and nutri-ent availability may constitute favourable conditions forthe survival of P. syringae in leaf litter as well as in asso-ciation with living host plants. We evaluated survival ofP. syringae in terms of stability of total populations of thisspecies. However, this might mask an underlying dynamicof birth and death of clonal lines differing in their relativecapacity to persist in and under snowpack. We observedwhat might be perceived as changes in relative abun-dances of different clonal groups in snowpack and litterover time (as illustrated in Fig. S1). But, a more intensivesampling would be needed to distinguish a possible tem-poral dynamic from spatial heterogeneity.

A second major contribution of our results is the dem-onstration that, at treeline altitude in the Upper Durancedrainage basin, alpine grasses and leaf litter are the majorsources of P. syringae in the snowpack and are a moreimportant source than snowfalls. Because P. syringae isice nucleation active (Vali et al., 1976; Lindow, 1983;Morris et al., 2004) and found in clouds (Amato et al.,2007) and snowfalls (Morris et al., 2008), we expectedthat snowfalls would be the major source of P. syringae inthe snowpack. However, P. syringae was detected in only10% of samples of freshly fallen snow (essentially at thebeginning of the winter). Furthermore, our microcosmstudies clearly reveal that a large part of the P. syringaepopulations in the snowpack, and in particular in the10 cm layer in contact with the soil surface, emigratesfrom litter. Emigration from the litter contributed to anincrease in the density of P. syringae of several orders ofmagnitude in the 10 cm layer of snow near the soilsurface. Irregularity of presence of P. syringae in snow-falls accumulating on top of this 10 cm layer contributed tothe distinct stratification.

Theoretical considerations corroborate our hypothesisthat P. syringae emigrates from litter to snowpack. At 0°C,snow is a porous medium represented by the threephases of water (solid, liquid and gas), which foster theexistence of a water film on the surface of ice grains thatcan rise in the snow for several cm by capillary forces(Coléou et al., 1999). Given that unfrozen water at thesoil–snow interface carries bacteria and others microor-ganisms (Ley et al., 2004), it is possible that P. syringae

was transferred in the snow via the capillary rise of water.This transfer could also be coupled to snow driftingevents. Alpine environments are windy and the first snow-falls can be repeatedly suspended and deposited withaerosols scrubbed from the ground (Lehning et al., 2006).

Clustering analysis confirmed that the habitat compart-ments of the water cycle harbour ubiquitous strains thatare also found on crops but also strains that seem to bemore specifically associated with non-agricultural habitatsand related to the UB246 and CC1524 clades. These twogenomic groups are the most deeply rooted in the mostrecent phylogenetic tree of the P. syringae species (Morriset al., 2010). Their association with aquatic habitats andespecially to litter supports the hypothesis that thespecies is well adapted to a lifestyle that does not neces-sitate pathogenicity to plants. However, the clusteringanalysis led us to conclude that snow and leaf litter cannotbe differentiated in terms of the effect that they might haveon the genetic diversity. Population structure seemed tobe the result of a set of conditions whose relative impor-tance could not be distinguished between samples andlikely involved exchanges between the two compartmentsthrough water transfer (capillary rise/percolation), localplant diversity of the meadow, deposition of local aero-sols, and possible multiplication of the population broughtby the first snowfall. Indeed, Young and colleagues (1977)showed that P. syringae could grow at cold temperatures,especially at 0°C with a doubling time of 22 h. The popu-lation at each site from the Upper Durance basin seemedto share the same dominant strains and to be composedof rare strains specific to the sample. Local diversitydynamics likely reflects high dispersal links in the metapo-pulation and seemed to show the same regional invari-ance observed by Östman and colleagues (2010) amongmicrobial communities in freshwaters of Sweden. Weobserved that there was no significant probability for anytype (phenotype) of strain of P. syringae to be associatedwith a particular site, date or substrate. Neither the prop-erties linked to pathogenicity nor to ice nucleation activityseemed to be determinants, independently or together.Among the strains from leaf litter and snow tested in thisstudy, 85% induced a hypersensitive reaction on tobaccowithout being aggressive on cantaloupe, indicating thatthey have a restricted host range (Morris et al., 2000). Thedominance of relatively host-specific strains showed thattheir limited pathogenic relation with plants was not abarrier to their survival in an alpine drainage basin.

Together, the leaf litter and snow pack harboured > 108

pathogenic P. syringae per square metre. Seasonal snow-pack of mountain regions is one of the most importantwater stocks upstream of cultures in middle and highlatitudes and occupies about a quarter of the global ter-restrial land surface (Edwards et al., 2007). In earlyspring, melting of snowpack recharges groundwater



tables and supplies the hydrological network of drainageareas where P. syringae was found from worldwide head-waters (Morris et al., 2010) to rivers (Morris et al., 2008)and retention basins (Riffaud and Morris, 2002). Given thepopulation dynamics in alpine meadows and in water,these observations strongly suggest that snowpack andthe associated leaf litter constitute crucial habitats for thedissemination and diversification of P. syringae. Theslopes studied here were dominated by grasses, forbsand legumes, but lands can be occupied by other coverssuch as forests or shrubs. This raises questions on theassociation of population structure in leaf litter with landcover and how this in turn is affected by altitude andclimate in general. Hence, global change is likely to havean impact on a plant pathogen such as P. syringae viathe effects on alpine regions. As reviewed in detail byGavazov (2010), some predictions suggest that globalwarming may deeply affect the functioning of alpine eco-systems via its impact on precipitation regimes, plant andmicrobial community structure, leaf litter quality and itsdecomposition rates. For example, snow cover thicknessand soil moisture are predicted to decrease. The risk offreeze events in spring is likely to intensify but thermo-philic species may progressively replace the present plantcommunity. If we assume that the dissemination of P. sy-ringae is strongly related to these environments, we couldwonder how global warming may affect its populationdynamics and what could be the impact on diseaseepidemiology. We could extend this question to plantpathogens in general whose life histories in relationwith non-agricultural habitats are still unsuspected orunderestimated.

Experimental procedures

Field sites and substrate sampling

The field study was carried out between December 2008 andMarch 2010 in prairies dominated by forbs, grasses andlegumes in the Southern French Alps where land was notcultivated: Col de Vars (06°42′07″E, 44°32′12″N), SuperSauze (06°42′36″E, 44°20′57″N), Ceillac (06°47′21″E,44°38′06″N), and Col du Lautaret (06°24′00″E, 45°02′09″N)at altitudes of about 2100 m. Sites were on calc-schistsbedrock.

Prior to snow fall, senescent plants and leaf litter weresampled in September 2009 and 2010 from three replicateplots of 20 cm2. All vegetation from each 20 cm2 was cut withsterile scissors and put into a clean plastic bag. The vegeta-tion from each study site was pooled. Litter from each 20 cm2

was scraped from the soil surface with a sterile spatula andput into a clean plastic bag. Likewise, the litter from eachstudy site was pooled.

Freshly fallen snow was collected on clean plastic tarps(1 m2) that had been placed on the ground or on previouslyfallen snow before each anticipated snow event. Sampleswere collected by 3 h after the end of the snowfall event.

Snowpack was sampled five times in 2 years from eachsite by digging pits. At each site a fresh pit (c. 2 m2 dug to theground level), oriented NW-NE in zones that had been pre-viously delimited to prevent crossing by pedestrians, was dugat each sampling date and was oriented in the delimited zoneso that pits at subsequent sampling dates would not beinfluenced by run-off from pits dug at previous dates. Snowfrom the snowpack was collected from three layers: the10 cm layer in contact with the ground, the 20 cm layer justabove this first layer, and the remaining layer of snow up tothe surface. Samples were collected from the top of the packtoward the bottom layer. At least 500 g, sampled at varioussites within the pit, was collected from each layer with adisinfected plastic scoop. At most dates, one bulk sample perlayer per date was collected except in the cases when weevaluated within-pit variability (3 bulks per layer per pit perdate). In 2010, plant material and snowpack was sampledfrom a delimited surface (0.04 m2) and the leaf litter underly-ing the snowpack was collected by scraping it from the soilsurface as described above. Samples were kept in a cooler,transported to the laboratory and processed on the followingday.

Quantification and isolation of P. syringae

Samples of leaf litter and senescent plants were macerated ina stomacher (BagMixer, Interscience, St Nom-La Breteche,France) for 2 min in 0.1 M phosphate buffer (8.75 g K2HPO4

and 6.75 g KH2PO4 per litre, pH 6.8). The volume of bufferused depended on the mass of tissue processed (about10–15 g of tissue and 2 ml of buffer per gram). The maceratewas dilution-plated as indicated below.

Snow samples were thawed overnight at room temperatureand processed within 24 h as described previously (Morriset al., 2008). Samples were concentrated by a factor of 200by filtration across sterile nitrous cellulose filters (pore diam-eter 0.22 mm) before dilution plating. Before filter concentra-tion, electrical conductivity and pH were measured with anelectrochemical analyser (Consort C561, UK, reference tem-perature at 25°C).

Aliquots of the samples as prepared above were dilution-plated on KBC medium (Mohan and Schaad, 1987), a semi-selective medium for P. syringae used in our previous studies(Morris et al., 2008) and on 10% tryptic soy agar (TSA) toenumerate the total mesophilic bacterial flora, and plateswere incubated for up to 5 days at 22–25°C. The detectionthresholds were 5 cfu for P. syringae and 55 for total bacteriaper litre of melted snow.

Strains of P. syringae were collected from snow pack andleaf litter. From each of these samples strains were collectedby purifying at least 30 colonies chosen randomly from asingle dilution. Strains were tested for production of fluores-cent pigment on King’s medium B (King et al., 1954) and theabsence of arginine dihydrolase and of cytochrome c oxidaseas described previously (Morris et al., 2008). For strainsmeeting the phenotypic criteria, between 20 and 40 strainsper sample were stored in 0.1 M phosphate buffer at 4°C andin nutrient broth with 40% glycerol at -80°C for further char-acterization. A set of 986 strains was randomly selected forthis study for full phenotypic characterization as describedbelow. Of these strains, 712 were from leaf litter and the snow



in the 10 cm layer closest to the soil collected in 2009 and2010 (Table S1).

Evaluation of the transfer of P. syringae from leaf litterto snowpack

The transfer of indigenous P. syringae from plant litter to thesnowpack covering the ground was investigated by establish-ing microcosms in February and March 2010. Litter collectedfrom under the snowpack at the Col du Lautaret site wascollected as described above and finely minced with sterilescissors. About 10 g of the litter were carefully placed at thebottom of 12 sterile cylindrical screw-cap polypropyleneflasks (100 mm deep, 50 mm in diameter) and compactedwith create a layer of about 8–10 mm in thickness. This wasaccomplished without touching the inner sides of the flasksabove the height of the layer of leaf litter. As an additionalprecaution, the inner sides were swabbed with alcoholseveral times and allowed to dry after the litter was put intothe flasks. To obtain snow for the microcosms, snow coreswere collected at depths between 1.30 m and 1.50 m fromthe ground with sterile tools. Based on preliminary results,snow at this depth frequently had no detectable levels ofP. syringae (verified as described below). Aliquots of 60 g ofsnow were then put into each flask resulting in a snow layerof about 10 cm. Flasks with snow but without leaf litter wereused as controls. All flasks were then buried under 1–1.80 mof snow and incubated for 9 weeks.

Microbiological analyses of the snow and leaf litter in themicrocosms were conducted similarly to those describedabove. At 3, 6 and 9 weeks of incubation, all snow except thecentimetre layer in contact with the litter was removed fromeach flask, melted, concentrated by filtration and dilutionplated on KBC and TSA media. The remaining snow wasdiscarded and the litter was recuperated for analysis.

Phenotypic characterization of strains

At least 20 strains per each sample of snow and plant litterwere characterized for ice nucleation activity (INA), thecapacity to induce a hypersensitive reaction in tobacco (HR),aggressiveness on cantaloupe (AGR) and production ofsyringomycin-like toxins (SYR) as described by Morris andcolleagues (2008). Inoculum for all tests consisted ofaqueous suspensions of 48 h bacterial cultures from strainsstored at -80°C. Strains were tested in blocks of up to 30chosen at random from the entire collection. Strain CC0094of P. syringae, which is virulent on cantaloupe, ice nucleationactive, and produces a syringomycin-like toxin (Morris et al.,2000), was used as a control in all blocks. Sterile distilledwater was used as a negative control.

Phenotypic characterization was performed as describedpreviously (Morris et al., 2010). INA was evaluated by deter-mining the freezing temperature, between -2° and -8°C ofthree drops (30 ml) containing 106 cells. Strains were scoredpositive for freezing if at least two drops froze. Induction of ahypersensitive reaction was tested by injecting 50 ml of asuspension of 108 cells ml-1 into the lamina of leaves oftobacco (Nicotiana tabacum cv. Xanthi) at the 10-leaf stageand noting the development of necrosis at 24 h after inocu-

lation. Aggressiveness of strains on cantaloupe (Cucumismelo var. cantalupensis cv. Vedrantais) was determined byinfiltrating 50 ml of a bacterial suspension of 3 ¥ 108 cells ml-1

into the hypocotyl of plants at the cotyledon stage. Ten plantswere inoculated per strain and the intensity of symptoms wasassessed after 7 days of incubation at ambient conditions of17–25°C according to an index from 0 to 4. Production ofsyringomycin-like toxins was revealed on a minimal SRMmedium (Gross, 1985) by measuring the inhibition zonecreated by antibiosis against Geotrichum candidum (Grossand Devay, 1977) after 6 days of incubation.

Genotypic characterization

Genotypic characterization was based on profiles from rep-PCR targeting the BOX element of genomic DNA (Morriset al., 2000) and sequencing of genes in the core genome asdescribed previously (Morris et al., 2008; 2010). For rep-PCR, primer sequences (obtained from Genset SA, Paris,France) were as described by Versalovic and colleagues(1991). The amplifications were performed with a 480Thermal Cycler thermocycler (Perkin-Elmer, Norwalk, CT)using cycles described by Louws and colleagues (1994).

For sequencing, one or all of four genes in the coregenome (rpoD, gyrB, cts and gapA) were sequenced from apure suspension of each strain adjusted to 2 ¥ 108 cells ml-1,for which DNA was amplified with primers described byYamamoto and colleagues (2000) for the first two genes, bySarkar and Guttman (2004) for the third and by Morris andcolleagues (2010) for gapA. PCR reactions were performedwith a Qiagen Multiplex kit (Qiagen, Courtaboeuf, France)and their products were checked by electrophoresis in 2%agarose gels before sequencing.

Clustering analyses

Strains of P. syringae from crops, water, leaf litter and the firstlayer of snow (0–10 cm) were clustered as described byMorris and colleagues (2010) on the basis of their geneticrelatedness, in terms of cts sequences, using two differentindividual-based clustering methods: a Bayesian algorithmand a multivariate analysis. To represent the full range ofdiversity observed in this study, we used the BOX-PRC pro-files to select strains; 290 strains were chosen based onthese profiles. Then the cts sequences of these 290 strainswere aligned with all those from a previous global biogeo-graphic study of P. syringae (323 sequences) (Morris et al.,2010) using the ClustalW program and truncated to a lengthof 415 bp with DAMBE (version 5.2.18) (Xia and Xie, 2001).Redundant sequences were eliminated, thereby leading tothe identification of 138 haplotypes among the total of 613sequences. The sequences of these haplotypes were con-verted into the Structure format using xmfa2struct, which wasdeveloped by Didelot and Falush and is available at theClonalFrame website (http://www.xavierdidelot.xtreemhost.com/clonalframe.htm). The Bayesian clustering approachwas performed using the software STRUCTURE 2.3 (Prit-chard et al., 2000; Falush et al., 2003) and the estimation ofgenetic clusters (K) was done as described by Morris andcolleagues (2010). The optimal number of clusters was



http://www.xavierdidelot.xtreemhost.com/clonalframe.htm

http://www.xavierdidelot.xtreemhost.com/clonalframe.htm

further evaluated by PCA using the procedure available in thepackage adegenet (Jombart, 2008) for the statistical free-ware R version 2.9.1 (The R Development Core Team, 2009).Principal component analysis was followed by a clusteringanalysis using the classical Ward method available in R,which is a hierarchical method designed to optimize minimumvariance within clusters.

Phylogenetic classification of strains

The objective of the phylogenetic analysis was to assignstrains isolated from the plant litter and the ground level layerof snow in 2010, to the clades described by Morris andcolleagues (2010). This was accomplished for the 290 strainsused for the cluster analysis above. Determination of theclade of each strain was based on the approach used in ourprevious study of P. syringae from non-agricultural and agri-cultural habitats (Morris et al., 2010), revealing that strains ofP. syringae could be accurately assigned to one of the cladesdefined in that work if the dissimilarity of the cts genesequence with others in the clade did not exceed a thresholdof 1.8%. Dissimilarity between strains was calculated withMEGA, version 4.0 (Tamura et al., 2007). Via this approachonly 8 of the 290 strains could not be assigned to one of thepreviously described clades (Morris et al., 2010).

Statistical analysis of genotypic and phenotypic diversity

Frequencies of individual phenotypes (INA, HR, AGR, SYR)between samples were compared with Fisher’s test using theR software version 2.9.1 (The R Development Core Team,2009). Frequencies of the different combinations of all fourphenotypes among samples were compared with a general-ized linear model. One of 72 possible combined phenotypeclasses was attributed to each of the 712 strains isolated in2009 and 2010 from the 10 cm snow layer closest to theground and the litter. For each combined phenotype class,the probability for a strain with this combination to be in a siteor in a substrate was assessed by fitting a generalized linearmodel (GLM; McCullagh and Nelder, 1989) to the data. In theGLM, the site or the substrate is the binary response variablemodelled with the Bernoulli distribution, the phenotypes andtheir interactions are the explanatory variables and the logis-tic function is used for the link function. The explanatoryvariables were selected with a backward minimization of theAkaike’s information criterion (AIC), which allowed us tochoose the best model fitting the data (Burnham and Ander-son, 2002). This procedure provides the distribution of prob-abilities DPobs for the observed combined phenotype classesto be in a given site or in a given substrate; this distributioncan be plotted as a histogram Hobs. One can test if DPobs issignificantly different from the distributions expected underthe null hypothesis of no difference in the phenotype combi-nations between the sites or the substrates. For this purpose,a permutation test (Manly, 1997) was implemented: the testprovides 95% confidence envelopes for the histogram Hobs byassuming that this histogram is obtained under the nullhypothesis (see Appendix S1 for details). If Hobs does not lie inthe envelopes, then the null hypothesis can be rejected at therisk level of 5%. In such a case, the combined phenotype

classes with high probabilities correspond to those whichhave a large propensity to be in the site or in the substrate ofinterest. Those with low probabilities are more likely to be inthe alternative site or substrate.

Acknowledgements

We thank Daniel Granier and his team from the ski resort ofSuper Sauze, for their help with collection of samples. We aregrateful to Odile Berge, Eric Michel and Pierre Amato for theirparticipation in sampling and for rich discussions, and Josse-lin Montarry for clusters analyses.

References

Amato, P., Parazols, M., Sancelme, M., Laj, P., Mailhot, G.,and Delort, A.M. (2007) Microorganisms isolated from thewater phase of tropospheric clouds at the Puy de Dome:major groups and growth abilities at low temperatures.FEMS Microbiol Ecol 59: 242–254.

Brooks, P.D., and Williams, M.W. (1999) Snowpack controlson nitrogen cycling and export in seasonally snow-coveredcatchments. Hydrol Process 13: 2177–2190.

Brooks, P.D., Williams, M.W., and Schmidt, S.K. (1996)Microbial activity under alpine snowpacks, Niwot Ridge,Colorado. Biogeochemistry 32: 93–113.

Burnham, K.P., and Anderson, D.R. (2002) ModelSelection and Multimodel Inference: A PracticalInformation-Theoretic Approach, 2nd edn. New York, USA:Springer-Verlag.

Cline, D. (1995) Snow surface energy exchanges and snow-melt at a continental alpine site. In Biogeochemistry ofSeasonally Snow-Covered Basins. Tonnessen, K.A., Will-iams, M.W., and Tranter, M. (eds). Colorado, USA: Inter-national Association for Hydrological Sciences, pp. 157–166.

Coléou, C., Xu, K., Lesaffre, B., and Broska, J. (1999) Cap-illary rise in snow. Hydrol Process 13: 1721–1732.

Edwards, A.C., Scalenghe, R., and Freppaz, M. (2007)Changes in the seasonal snow cover of alpine regions andits effect on soil processes: a review. Quatern Int 162:172–181.

Evanno, G., Regnaut, S., and Goudet, J. (2005) Detecting thenumber of clusters of individuals using the softwareSTRUCTURE: a simulation study. Mol Ecol 14: 2611–2620.

Falush, D., Stephens, M., and Pritchard, J.K. (2003) Infer-ence of population structure using multilocus genotypedata: linked loci and correlated allele frequencies. Genetics164: 1567–1587.

Gavazov, K.S. (2010) Dynamics of alpine plant litter decom-position in a changing climate. Plant Soil 337: 19–32.

Gross, D.C. (1985) Regulation of syringomycin synthesis inPseudomonas syringae pv. syringae and defined condi-tions for its production. J Appl Bacteriol 58: 167–174.

Gross, D.C., and Devay, J.E. (1977) Role of syringomycin inholcus spot of maize and systemic necrosis of cowpeacaused by Pseudomonas syringae. Physiol Plant Pathol11: 1–11.

Hirano, S.S., and Upper, C.D. (2000) Bacteria in the leaf



ecosystem with emphasis on Pseudomonas syringae – apathogen, ice nucleus, and epiphyte. Microbiol Mol BiolRev 64: 624–653.

Hodson, A., Anesio, A.M., Tranter, M., Fountain, A., Osborn,M., Priscu, J., et al. (2008) Glacial ecosystems. EcolMonogr 78: 41–67.

Jombart, T. (2008) adegenet: a R package for the multivariateanalysis of genetic markers. Bioinformatics 24: 1403–1405.

King, E.O., Ward, M.K., and Raney, D.E. (1954) Two simplemedia for the demonstration of pyocyanin and fluorescin.J Lab Clin Med 44: 301–307.

Lehning, M., Volksch, I., Gustafsson, D., Nguyen, T.A., Stahli,M., and Zappa, M. (2006) ALPINE3D: a detailed model ofmountain surface processes and its application to snowhydrology. Hydrol Process 20: 2111–2128.

Ley, R.E., Williams, M.W., and Schmidt, S.K. (2004) Microbialpopulation dynamics in an extreme environment: control-ling factors in talus soils at 3750 m in the Colorado RockyMountains. Biogeochemistry 68: 313–335.

Lindemann, J., Constantinidou, H.A., Barchet, W.R., andUpper, C.D. (1982) Plants as sources of airborne bacteria,including ice nucleation active bacteria. Appl EnvironMicrobiol 44: 1059–1063.

Lindow, S.E. (1983) The role of bacterial ice nucleationin frost injury to plants. Annu Rev Phytopathol 21: 363–384.

Lipson, D.A., Schmidt, S.K., and Monson, R.K. (1999) Linksbetween microbial population dynamics and nitrogen avail-ability in an alpine ecosystem. Ecology 80: 1623–1631.

Lipson, D.A., Schadt, C.W., and Schmidt, S.K. (2002)Changes in soil microbial community structure and functionin an alpine dry meadow following spring snow melt.Microb Ecol 43: 307–314.

Louws, F.J., Fulbright, D.W., Stephens, C.T., and Debruijn,F.J. (1994) Specific genomic fingerprints of phytopatho-genic Xanthomonas and Pseudomonas pathovars andstrains generetad with repetitive sequences and PCR. ApplEnviron Microbiol 60: 2286–2295.

McCarter, S.M., Jones, J.B., Gitaitis, R.D., and Smitley, D.R.(1983) Survival of Pseudomonas syringae pv tomato inassociation with tomato seed, soil, host tissue, and epi-phytic weed hosts in Georgia. Phytopathology 73: 1393–1398.

McCullagh, P., and Nelder, J.A. (1989) Generalized LinearModels, 2nd, edn. Boca Raton, FL, USA: Chapman &Hall/CRC.

Maki, L.R., Galyan, E.L., Chang-Chien, M.M., and Caldwell,D.R. (1974) Ice nucleation induced by Pseudomonas syrin-gae. Appl Microbiol 28: 456–459.

Manly, B. (1997) Randomization, Bootstrap and Monte CarloMethods in Biology. London, UK: Chapman and Hall/CRC.

Mohan, S.K., and Schaad, N.W. (1987) An improved agarplating assay for detecting Pseudomonas syringae pv.syringae and P. syringae pv. phaseolicola in contaminatedbean seed. Phytopathology 77: 1390–1395.

Mohr, T.J., Liu, H., Yan, S., Morris, C.E., Castillo, J.A., Jelen-ska, J., and Vinatzer, B.A. (2008) Naturally occurring non-pathogenic isolates of the plant pathogen Pseudomonassyringae lack a type III secretion system and effector geneorthologues. J Bacteriol 190: 2858–2870.

Morris, C.E., Glaux, C., Latour, X., Gardan, L., Samson, R.,and Pitrat, M. (2000) The relationship of host range, physi-ology, and genotype to virulence on cantaloupe inPseudomonas syringae from cantaloupe blight epidemicsin France. Phytopathology 90: 636–646.

Morris, C.E., Bardin, M., Berge, O., Frey-Klett, P., Fromin, N.,Girardin, H., et al. (2002) Microbial biodiversity:approaches to experimental design and hypothesis testingin primary scientific literature from 1975 to 1999. MicrobiolMol Biol Rev 66: 592–616.

Morris, C.E., Georgakopoulos, D.G., and Sands, D.C. (2004)Ice nucleation active bacteria and their potential role inprecipitation. J Phys IV 121: 87–103.

Morris, C.E., Sands, D.C., Vinatzer, B.A., Glaux, C., Guil-baud, C., Buffiere, A., et al. (2008) The life history of theplant pathogen Pseudomonas syringae is linked to thewater cycle. ISME J 2: 321–334.

Morris, C.E., Bardin, M., Kinkel, L.L., Moury, B., Nicot, P.C.,and Sands, D.C. (2009) Expanding the paradigms of plantpathogen life history and evolution of parasitic fitnessbeyond agricultural boundaries. PLoS Pathog 5:e1000693.

Morris, C.E., Sands, D.C., Vanneste, J.L., Montarry, J.,Oakley, B., Guilbaud, C., and Glaux, C. (2010) Inferring theevolutionary history of the plant pathogen Pseudomonassyringae from its biogeography in headwaters of riversin North America, Europe, and New Zealand. mBio 1:1–10.

Oki, T., and Kanae, S. (2006) Global hydrological cycles andworld water resources. Science 313: 1068–1072.

Östman, Ö., Drakare, S., Kritzberg, E.S., Langenheder, S.,Logue, J.B., and Lindstrom, E.S. (2010) Regional invari-ance among microbial communities. Ecol Lett 13: 118–127.

van Overbeek, L.S., Nijhuis, E.H.M., Koenraadt, H., Visser,J., and van Kruistum, G. (2010) The role of crop waste andsoil in Pseudomonas syringae pathovar porri infection ofleek (Allium porrum). Appl Soil Ecol 46: 457–463.

Pritchard, J.K., Stephens, M., and Donnelly, P. (2000) Infer-ence of population structure using multilocus genotypedata. Genetics 155: 945–959.

Reynolds, C.M., and Ringelberg, D.B. (2008) Non-indigenousendospore persistence following release in a snow – soilsystem. Cold Reg Sci Tech 52: 146–154.

Riffaud, C.M.H., and Morris, C.E. (2002) Detection ofPseudomonas syringae pv. aptata in irrigation water reten-tion basins by immunofluorescence colony-staining. Eur JPlant Pathol 108: 539–545.

Sarkar, S.F., and Guttman, D.S. (2004) Evolution of the coregenome of Pseudomonas syringae, a highly clonal,endemic plant pathogen. Appl Environ Microbiol 70: 1999–2012.

Schmidt, S.K., and Lipson, D.A. (2004) Microbial growthunder the snow: implications for nutrient and allelochemicalavailability in temperate soils. Plant Soil 259: 1–7.

Schmidt, S.K., Wilson, K.L., Monson, R.K., and Lipson, D.A.(2009) Exponential growth of ‘snow molds’ at sub-zerotemperatures: an explanation for high beneath-snow res-piration rates and Q (10) values. Biogeochemistry 95:13–21.

Schneider, T., Gerrits, B., Gassmann, R., Schmid, E.,Gessner, M.O., Richter, A., et al. (2010) Proteome analysis



of fungal and bacterial involvement in leaf litter decompo-sition. Proteomics 10: 1819–1830.

Snider, C.S., Hsiang, T., Zhao, G.Y., and Griffith, M. (2000)Role of ice nucleation and antifreeze activities in pathogen-esis and growth of snow molds. Phytopathology 90: 354–361.

Tamura, K., Dudley, J., Nei, M., and Kumar, S. (2007)MEGA4: molecular evolutionary genetics analysis (MEGA)software version 4.0. Mol Biol Evol 24: 1596–1599.

The R Development Core Team (2009) R: A Language andEnvironment for Statistical Computing. Vienna, Austria: RFoundation for Statistical Computing.

Vali, G., Christensen, M., Fresh, R.W., Galyan, E.L., Maki,L.R., and Schnell, R.C. (1976) Biogenic ice nuclei. Part II:Bacterial sources. J Atmos Sci 33: 1565–1570.

Versalovic, J., Koeuth, T., and Lupski, J.R. (1991) Distributionof repetitive DNA-sequences in eubacteria and applicationto fingerprinting of bacterial genomes. Nucleic Acids Res19: 6823–6831.

Williams, M.W., Seibold, C., and Chowanski, K. (2009a)Storage and release of solutes from a subalpine seasonalsnowpack: soil and stream water response, Niwot Ridge,Colorado. Biogeochemistry 95: 77–94.

Williams, M.W., Helmig, D., and Blanken, P. (2009b) White ongreen: under-snow microbial processes and trace gasfluxes through snow, Niwot Ridge, Colorado Front Range.Biogeochemistry 95: 1–12.

Xia, X., and Xie, Z. (2001) DAMBE: software package fordata analysis in molecular biology and evolution. J Hered92: 371–373.

Yamamoto, S., Kasai, H., Arnold, D.L., Jackson, R.W., Vivian,A., and Harayama, S. (2000) Phylogeny of the genusPseudomonas: intrageneric structure reconstructed fromthe nucleotide sequences of gyrB and rpoD genes. Micro-biology 146: 2385–2394.

Young, J.M., Luketina, R.C., and Marshall, A.M. (1977)Effects on temperature on growth in vitro of Pseudomonassyringae and Xanthomonas pruni. J Appl Bacteriol 42:345–354.

Zinger, L., Shahnavaz, B., Baptist, F., Geremia, R.A., andCholer, P. (2009) Microbial diversity in alpine tundrasoils correlates with snow cover dynamics. ISME J 3: 850–859.

Supporting information

Additional Supporting Information may be found in the onlineversion of this article:

Fig. S1. Example of BOX-PCR profiles in leaf litter and thefirst layer of snow at one site and two dates. Occurrence ofthe 11 profiles (p1 to p11) found in the two substrates at Colde Vars in March 2010 (A) and April 2010 (B). For bothfigures, profiles found in the leaf litter are underlined by thesolid black line, and those underlined by the hatched linewere in the snow. Those underlined by both solid and hatchedlines were found in both substrates. The lanes labelled Lshow the DNA molecular size marker (1-kb ladder; GibcoBRL, Life Technologies SARL, Cergy Pontoise, France); thesizes are indicated in base pairs.Table S1. Origin off strains from which P. syringae was iso-lated in this study.Table S2. Synthesis of frequency distributions of the prob-ability for each combination to be in a given substrate, site,date or sample.Appendix S1. The permutation test.

Please note: Wiley-Blackwell are not responsible for thecontent or functionality of any supporting materials suppliedby the authors. Any queries (other than missing material)should be directed to the corresponding author for the article.



Documents

Emigration of the plant pathogen Pseudomonas syringae from leaf litter contributes to its population dynamics in alpine snowpack