Introduction

The phyllosphere is defined as the aerial part of plant leaves circumscribed by the epidermal cell wall (Hirano and Christen 2000; Borges and Lopes 2008). The phyllosphere is governed by diverse biotic (plant species, plant phenological stages, etc.) and abiotic factors [ultraviolet radiation, temperature, dehydration, humidity, etc.], which can radically change within hours, days, and even seasons, thus the phyllosphere has been categorized as an extreme environment itself (Yang et al. 2001; Lindow and Brandl 2003). The microorganisms that colonize this habitat are known as epiphytes, and bacteria are considered to be the dominant microbial group (Bulgarelli et al. 2013).

In this context, diverse approaches [phospholipid fatty acid, denaturing gradient gel electrophoresis (DGGE), community-level physiological profile, etc.] have reported the occurrence of Firmicutes (Bacillus) and Proteobacteria (Pseudomonas) phyla as the dominant bacterial groups in the phyllosphere of vegetables, such as spinach, celery, rape, broccoli, and cauliflower (Zhang et al. 2010). By using high-throughput DNA sequencing (454-pyrosequencing), members of phyla Proteobacteria (Pseudomonas, Massilia, and Pantoea), Firmicutes (Bacillus), Bacteroidetes (Flavobacterium), and Actinobacteria (Arthrobacter) were found to be the most represented in the phyllosphere of field-grown lettuce plants (Rastogi et al. 2012). Nevertheless, our knowledge of the role of and ecological relationship between epiphytic bacterial populations and their plant host is still very limited since the majority of research studies have focused on the rhizosphere followed by the endosphere of plants (Berg et al. 2014). Raja et al. (2008) described that an epiphyte of the Methylobacterium strain utilizes methanol produced by cell metabolism of the host plant as an energy source, whereas the plant is favored by phytohormones (auxins) produced by the strain, which stimulate plant growth. Phyllosphere bacteria have also been found to fulfill nitrogen requirements in host plants that have lost contact with the ground, such as the perennial flowering plant Tillandsia (Bromeliaceae) (Brighigna et al. 1992). Cyanobacteria and gammaproteobacteria have also been suggested to be the most active diazotrophs in the phyllosphere of the tropical rainforest (Carludovica drudei, Grias cauliflora, and Costus laevis) (Fürnkranz et al. 2008).

Currently, Antarctic bacteria are of interest as producers of biotechnological bioactive compounds (enzymes, antibiotics, pigments, etc.) (Rojas et al. 2009); however, bacteria colonizing the phyllosphere of vascular Antarctic plants have not been investigated in detail so far. In this context, ice-binding proteins (IBPs) produced by bacteria living in cold environments could be potentially applied to diverse biotechnological fields, such as medicine (cryopreservation of cells and cryosurgery), food industry (improving ice-cream and frozen food quality), and agriculture (inhibition of ice nucleation and frost damage in plants) (Cid et al. 2016). IBPs, such as ice-nucleating proteins (INPs) and antifreeze proteins (AFPs), regulate the formation and growth of ice crystals (Davies 2014), and have been found in diverse microorganisms that survive and proliferate under freezing temperatures (Kawahara et al. 2004). AFPs act by binding to ice crystals to bring about thermal hysteresis (TH) and ice recrystallization inhibition (IRI). In the first case, TH is a noncolligative effect defined by a difference of the equilibrium freezing and melting points of a solution (Raymond and DeVries 1977), whereas IRI prevents generation of large ice crystals by boundary migration of smaller ice crystals (Yu et al. 2010). Few studies have purified and characterized AFPs from bacteria. Sun et al. (1995) isolated a Pseudomonas putida GR12-2 from the rhizosphere of Canadian High Arctic plants, which secretes an extracellular AFP. Inoculation of spring and winter canola with these AFP-producing strains resulted in an increase in root elongation length at 5 °C (Sun et al. 1995). Recent studies have identified AFPs secreted by epiphytic bacteria from Antarctic moss (Raymond 2015; Davies 2016). However, AFP-producing bacteria in the phyllosphere and their contribution to adaptation and/or protection of vascular plants against freezing temperatures have not been investigated so far.

In the light of these findings, the objective of this study is to explore the bacterial community structure associated with the phyllosphere of Deschampsia antarctica, and to look for IRI activity in protein crude extracts from culturable bacteria isolated from its phyllosphere.

Materials and methods

Plant collection

The selection of sampling sites and sample numbers were subject to (1) logistics provided by the Chilean Antarctic Institute (INACH), (2) abundance of plants, and (3) weather conditions. Eight whole plants were collected during February 2014 in the South Shetland Islands as follows: The first three plants (plants 1, 2, and 3) were taken from a rocky area at Gemelos Hill (62°11′45.73″S; 58°59′39.95″W) at height of 40 m above sea level. The plants were separated by no more than 2 m from each other. No more plants were taken from this site given the low amount of plants present at the sampling site. Plants 4, 5, and 6 were taken from Collins Glacier (62°10′9.00″S; 58°51′21.06″W) in areas with different characteristics. Plant 4 was taken from a bird breeding area, 15 m above sea level. Plant 5 was taken from an area highly populated by mosses, 30 m above sea level. Plant 6 was taken from a rocky place, 45 m above sea level. Only one plant (plant 7) was taken from Ardley Peninsula (62°12′36.78″S; 58°56′59.25″W), 10 m above sea level. No more plants were taken from this sandy area also populated by mosses, given the low amount of plants found at this site. Another plant (plant 8) was taken from Hanna Point (62°39′15.64″S; 60°36′51.82″W). This is an area highly covered by D. antarctica plants, but only one plant was taken given the bad weather conditions when samples were collected. A map showing these sampling sites is included in Fig. 1. We would like to emphasize that the number of samples and locations were defined at the time according to available resources (time, plant coverage, etc.) and changing climate conditions.

Fig. 1
figure 1

Map of Antarctic Peninsula. In detail, sampling sites of D. antarctica plants (red circles) on South Shetland Islands. Plants 1, 2, and 3 were taken from Gemelos Hill (62°11′45.73″S; 58°59′39.95″W). Plants 4, 5, and 6 were taken from Collins Glacier (62°10′9.00″S; 58°51′21.06″W). Plant 7 was taken from Ardley Peninsula (62°12′36.78″S; 58°56′59.25″W). Plant 8 was taken from Hanna Point (62°39′15.64″S; 60°36′51.82″W). (Color figure online)

Full size image

The plants were transported in coolers and processed at the laboratory of the Prof. Julio Escudero Scientific Base of the Chilean Antarctic Institute (INACH), King George Island.

Bacterial community structures

Bacterial community composition was assessed by DGGE as described by Kawai et al. (2002). First, 1 g of D. antarctica leaves was cut (aerial parts), gently washed, then vortexed for 10 min in 10 ml sterile saline solution (0.85 % NaCl). Leaves were removed, and the recovered liquid was centrifuged at 15,700×g for 10 min to collect detached bacterial cells. To obtain 1 g of leaf, it was necessary to use almost all of the leaves on a single plant. Therefore, the DGGE banding profile was obtained from total DNA extract from a composited sample of several leaves of every single plant sampled. Bacterial cells were suspended in 50 µl sterile distilled water, and this suspension was subsequently frozen in liquid nitrogen and thawed at room temperature three times to break the cells. The samples were then centrifuged at 15,700×g for 40 min, and the supernatant (~40 µL) was used as template DNA in the PCR reaction.

The bacterial 16S rRNA genes (regions V6–V8) were amplified by touchdown PCR with primer set EUBf933–GC/EUBr1387 as described by Iwamoto et al. (2000). The conditions for the PCR reaction were as follows: hot start was performed at 95 °C for 10 min, the annealing was initially set at 65 °C and then decreased by 0.5 °C every cycle and held at 55 °C for 1 min, followed by extension at 72 °C for 3 min. Ten additional annealing cycles were then carried out at 55 °C, followed by denaturation at 94 °C for 1 min and extension at 72 °C for 3 min. The final extension step was done for 7 min at 72 °C. The DGGE runs were performed using the DCode™ universal mutation detection system (Bio-Rad Laboratories Inc., USA). Twenty-microliter aliquots of PCR product were loaded onto a 6 % (w/v) polyacrylamide gel with a 40–65 % denaturing gradient (7 M urea and 40 % formamide). The electrophoresis was run for 12 h at 100 V. The gel was stained with SYBR Gold (Invitrogen™, Thermo Fisher Scientific Inc., USA) for 30 min and photographed on a UV transilluminator (GelDoc-It®TS2 Imager, UVP). Clustering of DGGE banding profiles using a dendrogram was carried out with Phoretix 1D Pro gel analysis software (TotalLab Ltd., UK; http://totallab.com/). Based on the matrix obtained from the Phoretix 1D analysis, differences between bacterial communities were calculated by similarity profile analysis (SIMPROF test) with Bray–Curtis similarity index, 5 % significance level, and <0.1 stress values (Clarke 1993; Clarke et al. 2008), and visualized by nonmetric multidimensional scaling (NMDS) analysis using Primer 6 software (Primer-E Ltd., UK; http://www.primer-e.com/).

A total of 19 representative bands from DGGE gels were carefully excised, purified, reamplified by touchdown PCR, and run again to avoid inclusion of more than one band, to be sent to Macrogen, Inc. (Korea) for sequencing (Fig. 2a). The sequences obtained in this study were compared with those present in the GenBank database (National Center for Biotechnology Information; NCBI) by using BLAST tools (http://blast.ncbi.nlm.nih.gov/Blast.cgi) to identify bacterial groups. More bands were sequenced, however only those having higher similarity % (≥90) are included in Table 1.

Fig. 2
figure 2

a Denaturing gradient gel electrophoresis (DGGE) banding profile of bacterial communities present in eight plants of the D. antarctica phyllosphere. Arrows indicate excised and sequenced bands. Analysis of similarity percentages of dendrogram b and nonmetric multidimensional scaling c of DGGE profiles (16S rRNA gene) from bacterial communities from eight plants of the D. antarctica phyllosphere

Full size image
Table 1 Phylogenetic assignment of representative DGGE bands
Full size table

Isolation of culturable bacteria

Isolation of bacteria from the D. antarctica phyllosphere was carried out by a plating method using four culture media: NM-1 [0.5 g l−1 d-glucose, 0.5 g l−1 polypeptone, 0.5 g l−1 sodium glutamate, 0.5 g l−1 yeast extract, 0.44 g l−1 KH2PO4, 0.1 g l−1 (NH4)2SO4, 0.1 g l−1 MgSO4·7H2O, 15 g l−1 agar, and 1 ml vitamin solution containing 1 g l−1 nicotinamide, 1 g l−1 thiamine hydrochloride, 0.05 g l−1 biotin, 0.5 g l−1 4-aminobenzoic acid, 0.01 g l−1 vitamin B12, 0.5 g l−1 d-pantothenic acid hemicalcium salt, 0.5 g l−1 pyridoxamine dihydrochloride, 0.5 g l−1 folic acid (Nakamura et al. 1995)], R2A (Oxoid Ltd., Thermo Fisher Scientific Inc., UK), Pseudomonas (Oxoid Ltd.) and Actinomyces (Oxoid Ltd.) agar medium. One gram of D. antarctica leaves in 45 ml sterile saline solution (0.85 % NaCl) was vortexed for 10 min to detach the adhered bacteria. Then, dilutions of bacterial suspensions were spread on culture media agar plates in triplicate. The plates were incubated at 13 °C until colony-forming units (cfu) were observed on agar. The value of cfu per cm2 was also calculated by scanning the leaf’s surface area with ImageJ free software (http://imagej.nih.gov/ij/). After 1 week of incubation, 267 single colonies showing diverse phenotypes (color, brightness, form, elevation, and margin; Smibert and Krieg 1994) were transferred to fresh media, purified by streaking on agar, and stored in 3:7 glycerol:LB (10 g l−1 d-glucose, 5 g l−1 yeast extract, 10 g l−1 tryptone, 5 g l−1 NaCl) broth at −20 °C for further analysis.

Selection of isolates by genotyping

The enterobacterial repetitive intergenic consensus polymerase chain reaction (ERIC-PCR) technique was used to differentiate colonies with similar phenotype but different genotype, following the protocol described by Houf et al. (2002). Briefly, crude DNA extracts from samples were obtained after boiling bacterial cell suspension followed by a quick spin down to discard cell debris. Supernatant at concentration of 50 ng DNA µl−1 was used in the PCR mixture. DNA fragments were amplified by PCR using the primer set ERIC motifs 1R (5′-ATG TAA GCT CCT GGG GAT TCA C-3′) and 2 (5′-AAG TAA GTG ACT GGG GTG AGC G-3′). The PCR conditions were as follows: hot start at 94 °C for 5 min, followed by 40 cycles of denaturing at temperature of 94 °C for 1 min and annealing at temperature of 25 °C for 1 min, and extension at temperature of 72 °C for 2 min. The final extension step was set at 72 °C for 7 min. PCR products were then run on a 2 % agarose gel at 100 V for 1 h and stained with ethidium bromide. Electrophoretic gels were photographed, and the images were analyzed using Phoretix 1D Pro gel analysis software. The isolates that showed different banding profiles were considered genetically different and used for further analysis.

Ice recrystallization inhibition (IRI) detection

Detection of IRI was based on the behavior of ice at subzero temperatures. Ice recrystallization and the consequent ice crystal growth is visually different with higher transparency, whereas in IRI, ice crystals remain at small size, being visually more turbid (Gilbert et al. 2004; Budke et al. 2009). Given these differences in turbidity, ice recrystallization inhibition effect was measured in this work by a spectrophotometric method. The details of IRI detection theory are given in Fig. 4a. Absorbance differences were determined using a microtiter plate spectrophotometer (Multiskan GO, Thermo Fisher Scientific, Inc., USA). First, the most sensitive absorbance to detect IRI under our conditions was determined by wavelength (λ) scanning from 300 to 650 nm using type III AFP (A/F Protein, Inc., USA) and bovine serum albumin (BSA) as positive and negative protein control, respectively. Detection of IRI was assayed with both pure proteins at final concentration of 0.04 mg ml−1 in 30 % sucrose solution according to the protocol described by Gilbert et al. (2004). The microtiter plate was frozen at −80 °C for 15 min and incubated for 2 days at −6 °C before reading the absorbance in a spectrophotometer. Microtiter plate wells with 30 % sucrose solution were used as blanks. The detection of IRI was assayed with different total protein concentrations ranging from 0.01 to 2 mg ml−1.

Screening for IRI in isolates

Based on our results of IRI detection with different wavelengths and protein concentrations, the screening of IRI detection by spectrophotometry was standardized at λ of 500 nm using crude extracts with protein concentrations from 0.01 to 2 mg ml−1. Antarctic bacteria isolates were grown in LB broth at 15 °C for 1 week, then cold acclimatized at 4 °C for 1 week. The cultures were then centrifuged at 5000×g for 10 min, and soluble proteins from the pellets were extracted by using B-PER bacterial protein extraction reagent (Thermo Fisher Scientific, Inc., USA) according to the manufacturer’s instructions. The protein concentrations in crude extracts were determined using the Bradford protein assay (Bio-Rad Laboratories, Inc., USA), and IRI detection was carried out as described above. Type III AFP and BSA were used in each microtiter plate reading to verify that IRI detection was working correctly. Crude protein extracts from Escherichia coli JM109 were used as negative control for IRI screening. E. coli JM109 was chosen because its genome information in the GenBank database did not include any genetic traits associated with IBPs. This was also confirmed in our laboratory, where fresh LB cultures were exposed to −20 °C for 24 h. Before and after freezing, serial dilutions (from 10−1 to 10−7) of E. coli JM109 culture plated on LB agar plates did not grow. Isolates from the D. antarctica phyllosphere were also included in this assay to compare strain growth after freezing exposure.

Samples with statistically higher absorbance (see below) compared with E. coli JM109 were considered as active isolates for IRI. These IRI-active isolates were repeatedly screened to confirm IRI activity, and those isolates which always showed higher absorbance compared with E. coli JM109 were selected for additional assays for IRI detection without cold acclimation.

Statistical analysis

Absorbance data were analyzed by Dunnett multiple comparisons (Dunnett 1955). Means and standard errors were calculated in quadruplicate, and difference at p ≤ 0.05 was considered significant. Statistical analyses were conducted using IBM SPSS 21 software.

Results

Bacterial community structure

The results obtained from the DGGE analysis and visualized with a dendrogram showed that the bacterial community in the D. antarctica phyllosphere was mainly distributed in two clusters (Fig. 2b). One of the clusters showed higher similarity (>57 %) among plants collected at Gemelos Hill (plants 1, 2, and 3), sharing some bands with plant 6 collected at Collins Glacier, whereas the second cluster with lower similarity (>40 %) was obtained from plants taken from Collins Glacier (plant 4), Ardley Peninsula (plant 7), and Hanna Point (plant 8). The most dissimilar bacterial community structure was observed for plant 5, collected from Collins Glacier. Differences between bacterial communities were also visualized by nonmetric multidimensional scaling (NMDS) analysis (Fig. 2c), where only 40 % similarity was found among plants collected from different sampling sites (plants 1, 2, 3, 6, and 7) and only 20 % similarity for all samples studied.

Despite this, significant differences between bacterial communities were observed among collected plants: sequencing of representative dominant bands on DGGE gels revealed that most members belonged to the Pseudomonadales order, followed by less abundant bands represented by the Rhizobiales order (Table 1). Coincidently, the closest relative sequences deposited in the GenBank database are associated with bacteria from Arctic and Antarctic ecosystems, while others were also reported to be related to plant rhizosphere habitats. The sequences obtained in this study were deposited in GenBank database under accession numbers from KU645377 to KU645395.

Isolation and genotyping of culturable bacteria

As shown in Table 2, bacterial counts in the D. antarctica phyllosphere revealed bacterial loads ranging from 3.6 × 106 to 6.7 × 106 cfu g−1 of leaf, equivalent to 3.6 × 104 to 6.7 × 104 cfu cm−2. A total of 265 isolates were obtained with high genetic variability (from 76 to 91 %) as revealed by ERIC-PCR genotyping (Table 2; Fig. 3). It is noteworthy that the same genotypes were found in different plants; however, this occurred in bacterial isolated from plants sampled at the same site, such as plants 4, 5, and 6, all obtained from Collins Glacier (Fig. 3).

Table 2 Bacterial counts (cfu g−1 of leaf), ERIC-PCR genotyping, and ice recrystallization inhibition (IRI) activity of culturable bacteria isolated from the Deschampsia antarctica phyllosphere
Full size table
Fig. 3
figure 3

Dendrogram of ERIC-PCR genotyping of culturable bacteria from the D. antarctica phyllosphere. a Isolates grown in Pseudomonas agar (PA) plates and b grown in NM1 culture medium

Full size image

Detection and screening for IRI in isolates

During standardization of the protocol for the detection of IRI activity, our results showed differences in absorbance between the positive (type III AFP) and negative (BSA) controls at the different tested wavelengths (from 300 to 650 nm) (Fig. 4b). These differences in absorbance between controls were observed at 500 nm using different protein concentrations (from 0.01 to 2.0 mg ml−1) (Fig. 4c). The screening for IRI activity was carried out at 500 nm, given that this was the wavelength in the middle of the plateau region, as presented in Fig. 4b.

Fig. 4
figure 4

Standardization of detection of ice recrystallization inhibition (IRI) in culturable bacteria of the Deschampsia antarctica phyllosphere. a Background theory, b absorbance at different wavelengths (from 300 to 650 nm), c and d absorbance at 500 nm at different BSA and AFP protein concentrations (from 0.01 to 2.0 mg ml−1). AFP: antifreeze protein in 30 % sucrose solution; BSA: bovine serum albumin in 30 % sucrose solution; Blank: 30 % sucrose solution

Full size image

Compared with crude protein extract from E. coli JM109, the screening for IRI activity showed significantly higher absorbance (p ≤ 0.05) for protein extracts obtained from 32 isolates; therefore, 21 % of isolates were considered as IRI active. The screening was repeated to confirm IRI activity in positive isolates. These data are presented in Fig. 5a and summarized in Table 2. However, when IRI activity was screened without cold acclimation, to explore whether IRI was induced by low temperature or was a constitutive phenotype of selected isolates, only five (3 %) isolates (strains 28 NM1, 43 NM1, 19 PA, 33 PA, and 20 AA) showed significantly higher absorbance (p ≤ 0.05) compared with E. coli JM109 protein extract (Fig. 5b).

Fig. 5
figure 5

Ice recrystallization inhibition in the phyllosphere: a Screening of ice recrystallization inhibition (IRI) activity in crude protein extract (1 mg ml−1) from bacteria isolated from the D. antarctica phyllosphere using Pseudomonas and Actinomyces agar plates. b Detection of IRI activity in crude protein extract (0.5 mg ml−1) from bacteria grown without cold temperature acclimatization. Asterisks denote significant (p ≤ 0.05, Dunnett’s test) differences in comparison with crude protein extract of E. coli JM109 (negative control)

Full size image

Discussion

The profiles obtained by DGGE (16S rRNA gene) analysis revealed differences in bacterial communities in the D. antarctica phyllosphere, which varied among the different plants and collection sites. Variations of bacterial communities in the phyllosphere between different vegetables (spinach, celery, rape, broccoli, and cauliflower) have previously been described by Yang et al. (2001). High variability was found in bacterial community composition in forest, among Pinus trees, as described by pyrosequencing (16S rRNA gene) (Redford et al. 2010). High diversity of bacterial composition was also detected in the phyllosphere of the leaf canopy of a tropical Atlantic forest (Lambais et al. 2006). One of the more intriguing findings of this study is that the microbial communities varied substantially between different locations, but were more similar between plants at the Gemelos Hill location, sharing higher similitudes with plant 6 sampled from a rocky place at Collins Glacier. Some theories postulate that environments select their colonizers, meaning that local conditions regulate the abundance, composition, and activity of the inhabitants (Perfumo and Marchant 2010). In a controlled greenhouse experiment, the spatial distribution of Arabidopsis thaliana plants was found to affect the dispersion of bacteria on plants, showing similar bacterial community structure in closer plants (Maignien et al. 2014). Our results agree with this finding; however, more studies are required for confirmation. In this sense, considering that the majority of the Antarctic Continent is covered by ice, less than 1 % of the land, mostly in the Antarctic Peninsula and coastal areas, is available for plant colonization (Alberdi et al. 2002). Oceanic winds, rain, hailstones, and snow could be the most important mechanisms for dispersion of microorganisms in this area; For example, Santl-Temkiv et al. (2013) demonstrated that hailstone samples were dominated mostly by plant-surface bacteria, thus those authors hypothesized that adaptations to life in the phyllosphere favor survival of airborne bacteria, affecting long-distance transport and the spatial distribution of bacteria on Earth. In other ecosystems such as desert, dispersion of bacteria is a natural phenomenon conducted by bacteria attached to dust soil particles, contributing to the diversity of downwind ecosystems (Yamaguchi et al. 2012). Another mechanism for bacterial transport is local fauna such as skuas and gulls, to whose nests this plant has often been found to be attached (Gerighausen et al. 2003). In addition, differences in bacterial communities in the phyllosphere could also be dependent on the location and morphology of the leaves, phenological status of plants, and amount of carbon-containing exudates released by leaf plants and used by bacteria as a nutrient source (Wilson and Lindow 1994; Yang et al. 2001).

Despite differences in bacterial community composition in the phyllosphere, the sequencing of dominant DGGE bands revealed the presence of members of Pseudomonadales (Pseudomonas and Psychrobacter) and Rhizobiales (Agrobacterium and Aurantimonas) orders. Pseudomonadales is a large taxonomic bacterial group that is metabolically versatile and collectively exhibits a high diversity of activities such as nutrient cycling, degradation of xenobiotic organic compounds, and growth promotion and pathogen protection of plants (Redford et al. 2010; Loeschcke and Thies 2015). By using molecular approaches (DGGE and 454-pyrosequencing), Pseudomonadales have been reported as a dominant inhabitant in the phyllosphere of different plants such as spinach, celery, rape, broccoli, cauliflower, and Arabidopsis thaliana (Bodenhausen et al. 2013). Pseudomonadales are widely described as a common inhabitant of soil; therefore, we cannot discard the soil as a source of phyllosphere colonizers (mentioned above). Rhizobiales is also a metabolically versatile soil bacterial group characterized by conversion of atmospheric nitrogen to ammonia, to the benefit of the host plant. Rhizobiales are commonly found in the phyllosphere of rice (Oryza spp.) and perennial herbaceous plants (Typha angustifolia) by metaproteogenomic analysis and 16S rDNA gene sequencing (Li et al. 2011; Knief et al. 2012; Banik et al. 2015). Both bacterial groups are also frequently isolated from Antarctic and Arctic soils (Goh and Tan 2012; Park and Kim 2015). It is noteworthy that the sequencing of dominant DGGE bands also revealed the presence of Psychrobacter. Psychrobacter arcticus is a cold-adapted bacteria used as a model for psychrophilic “cold-shock” proteins, proposed as key for life at subzero temperatures (Kuhn 2012).

In samples from the D. antarctica phyllosphere, our counts ranged from 3.6 × 106 to 6.7 × 106 cfu g−1 of leaf, equivalent to 3.6 × 104 to 6.7 × 104 cfu cm−2. The bacterial loads found in the D. antarctica phyllosphere seem to be similar to some plants from other climate zones. Variable bacterial loads have been reported for different plant species, such as olive plants [from 1.7 × 104 to 3.4 × 105 cfu cm−2 (Ercolani 1991), Mediterranean perennial species (from 1.3 × 104 to 1.4 × 107 cfu g−1 (Yadav et al. 2004)], evergreen subtropical forest trees [from 7.41 × 102 to 3.02 × 105 cfu g−1 (Yadav et al. 2013)], and common beans [from 106 to 107 cfu cm−2 (Hirano and Upper 1989)]. Other authors have suggested that differences in bacterial counts could be explained mostly by the distance of leaves to the soil, as a direct source of epiphytic bacteria, and environmental factors, such as relative humidity, especially in water-saturated soil conditions (Kinkel et al. 2000).

Similar to DGGE analysis, ERIC-PCR genotyping also suggests high genetic variability (from 76 to 91 %) among the culturable bacteria from D. antarctica. In addition, this result helped us to differentiate colonies with similar phenotype but different genotype. Interestingly, the same genotypes determined by ERIC-PCR were found in plants sampled from the same location, but not among plants from different sampling sites. The latter result could be explained by the fact that vegetative reproduction is the most frequent and reliable means of dispersion in D. antarctica plants (Smith 2003). Other authors have postulated that the occurrence of bacteria with the same genotype in different plants could be attributed to migration of bacteria from neighboring plants and/or plant debris transported by wind, insects, and other animal sources (Whipps et al. 2008; Vokou et al. 2012). In addition, numerous phyllosphere bacteria have been associated with seeds, with the ability to grow on seedlings and leaves as the plant grows (Hirano and Christen 2000), thus seeds could also be considered vectors for specific bacterial genotype transference in this type of ecosystem; however this statement has not been evaluated in extreme environments such as Antarctica.

With regards to screening for IRI activity, a total of 32 (21 %) crude protein extracts from phyllosphere isolates showed significantly higher absorbance compared with protein extract from E. coli JM109. Ice recrystallization inhibition has been shown to be a functional and cost-effective mechanism for preliminary detection of AFP activity within a solution (Capicciotti et al. 2013). Moreover, the percentage obtained of active IRI isolates (21 %) is higher than that previously described by Gilbert et al. (2004) (10 %) for putative AFP-producing isolates from an Antarctic lake. However, the presence of IRI activity in culturable bacteria does not mean that these are the dominant group in the bacterial community structures in the phyllosphere of D. antarctica, and such activity does not ensure that bacterial IRI activity is expressed on leaf surfaces. In addition, as reported in other studies, the IRI effect has mostly been related to freezing tolerance, as it maintains ice crystals at a small, likely harmless size (Do et al. 2014). A different theory points out that proteins secreted into the surrounding environment have IRI activity that favors formation of water channels (water pockets) through which bacteria could obtain nutrients and divide (Raymond 2011; Raymond and Kim 2012). In this context, it has been proposed that the repeated occurrence of the DUF3494 sequence (coding a protein of unknown function) in the genome of psychrophilic organisms (bacteria, archaea, and eukaryotes) is correlated with IRI activity, and that it can be transferred between prokaryotes and eukaryotes by horizontal gene transfer (Raymond 2014). The discovery of epiphytic bacteria coding the DUF3494 domain in metagenomic analysis conducted in Antarctic nonvascular plants (mosses) suggests that this activity might favor a commensal relationship between the moss and epiphytic bacteria, where the latter receives sustenance and the former freezing protection, nevertheless this statement has not been empirically proven (Raymond 2015).

Detection of IRI activity in crude protein extracts from bacteria grown without cold acclimatization was also observed (Fig. 5b). The constitutive expression of AFP by bacteria could provide an immediate benefit during sudden harsh local conditions, such as freezing; however, constitutive protein expression represents a permanent metabolic cost for cells, as discussed by Geisel (2011).

Finally, further studies are required to elucidate factors influencing the diversity and abundance of bacteria in the phyllosphere of Antarctic vascular plants and their activity under freezing temperature, such as production of ice-binding proteins. This information could be pivotal for our understanding on Antarctic ecosystems and the physiological response of Antarctic vascular plants to tolerate freezing temperature and climate change (Cavieres et al. 2016).