Changes in spinach phylloepiphytic bacteria communities following minimal processing and refrigerated storage described using pyrosequencing of 16S rRNA amplicons


Monica A. Ponder, Department of Food Science and Technology, Virginia Polytechnic Institute and State University, Blacksburg, VA 24061, USA. E-mail:


Aims:  To investigate the changes in bacterial diversity on fresh spinach phyllosphere associated with storage at refrigeration temperatures.

Methods and Results:  Community structure and population dynamics of spinach phylloepiphytic bacteria associated with packaging and refrigeration of ready-to-eat fresh produce were evaluated using pyrosequencing of 16S rRNA gene amplicons. A diverse community responsive to storage at refrigerated temperatures was detected belonging to over 1000 operational taxonomic units, including many diverse members not previously described on the phyllosphere. Of the approx. 8800 unique sequences examined from fresh spinach leaves, 75% were from previously undescribed taxa. The classified sequences from the fresh spinach phyllosphere were assigned to 11 different phyla with the largest number of reads belonging to Proteobacteria and Firmicutes. Packaging and storage of spinach under refrigerated conditions decreased the richness, diversity and evenness of the bacterial community. Refrigeration at 4 and 10°C and storage resulted in a decrease in number of taxa represented from 11 phyla in fresh spinach to only 5 phyla after 1 day of storage. Sequences belonging to γ-Proteobacteria, particularly Pseudomonas spp. and members of the Enterobacteriaceae, were the most numerous after 15 days of storage at both temperatures. Growth inhibition of the genera Escherichia was achieved at 4°C but not at 10°C storage, thus highlighting the importance of temperature in fresh packaged spinach.

Conclusions:  The application of pyrosequencing to describe composition and diversity of the phyllosphere on spinach leaves provided a broader outlook of the bacterial composition of this community complementing other phyllosphere studies that have used culture- and nonculture-dependent approaches.

Significance and Impact of the Study:  Pyrosequencing allowed a broader description of the bacterial composition and diversity of the spinach leaf surface than previously obtained using culture-based detection and will be a powerful tool to help ensure the future safety and quality of packaged spinach.


Packaging is a common technology utilized to increase shelf life of leafy greens by decreasing plant respiration rate and delaying deterioration associated with growth of spoilage micro-organisms. A combination of packaging and refrigerated storage, reduced growth, but failed to eliminate spoilage-associated microbes, and it is not designed to eliminate human pathogens (Delaquis et al. 2007; Rico et al. 2007). Little is known how packaging and temperature affect the complex microbial communities established on the leaf surface or phyllosphere and whether these changes could affect the establishment or survival of human pathogens on fresh plants. Recent outbreaks of human disease associated with fresh produce have highlighted the vulnerability of fresh market production to colonization by food-borne pathogens such as Escherichia coli O157:H7 and Salmonella enterica. In 2006, a multistate outbreak of E. coli O157:H7 was traced back to consumption of spinach grown in California, packaged and shipped across the United States and Canada (CDC 2006). Exploration of microbial diversity on the phyllosphere of plants to study changes in composition of these communities can provide insights into bacterial interactions that can promote the establishment of foreign bacteria into a community.

Bacteria are considered to be the dominant microbial inhabitants of the phyllosphere with culturable bacteria numbering between 102 to 1012 cells per gram of leaf (Thompson et al. 1993). Bacteria belonging to more than 85 different species in 37 genera have been cultured from the phyllosphere of rye, olive, sugar beet and wheat, yet nonculture-dependent techniques have identified 95–671 bacterial species associated with tree leaves (Thompson et al. 1993; Hirano and Upper 2000;. Yang et al. 2001; Lambais et al. 2006). Vegetables, including leafy greens are also colonized by a variety of micro-organisms. To date, characterization of the microbial communities of minimally processed vegetables has focused on cultivation or microarray-based detection of spoilage-associated bacteria, specifically Pseudomonadaceae and Enterobacteriaceae (Babic et al. 1996; Allende et al. 2004; Ragaert et al. 2007). Although these bacteria are frequently isolated from the phyllosphere of leafy greens, other bacterial members present in lower numbers could influence quality and safety of packaged edible leaves. Bacteria from the spinach phyllosphere are capable of complex interactions with human pathogens on the plant surface (Lopez-Velasco et al. 2010), which might impact their fitness. As the preharvest-associated microbiota remains associated with raw vegetables during transportation, further processing and storage (ICMSF 2005), it is important to examine the effect of packaging and low-temperature storage on epiphytic bacteria as they may offer a strategy to control food-borne pathogens.

Pyrosequencing is a recently developed culture-independent technology in which more than 300 000 sequences can be simultaneously determined without cloning. A highly variable region of 16S rRNA gene is amplified using primers that target adjacent conserved regions, followed by direct sequencing of individual PCR products (Ronaghi and Elahi 2002; Liu et al. 2007). The methodology has been used to describe microbial communities from deep sea environments (Sogin et al. 2006; Huber et al. 2007), intestinal tract (Andersson et al. 2008; Dowd et al. 2008), soils (Roesch et al. 2007; Urich et al. 2008), fermented foods (Humblot and Guyot 2009) and the rhizosphere (Jesus et al. 2010) and have broadened our understanding of the structure and diversity of these complex environments.

We used 454 pyrosequencing of bacterial 16S rRNA gene amplicons to compare the composition and richness of spinach microbiomes obtained from fresh and packaged spinach stored at refrigeration temperatures (4 and 10°C). The purpose of this study was to characterize and evaluate the bacterial composition on fresh spinach leaves subjected to minimal processing to identify potential bacterial members that affect its quality and could influence the survival of pathogenic bacteria and to show the suitability of high throughput metagenomic analysis for describing changes in bacterial composition during storage of packaged spinach.

Materials and methods

Spinach production

Fresh and packaged spinach  Spinach production and the packaging of spinach have been previously described (Lopez-Velasco et al. 2010). Briefly, a flat-leaf cultivar (Monza; Seedway LLC, Hall, NY, USA) was seeded in April, 2008, at the Virginia Tech Kentland Research Farm. A plot of 6 × 3·75 m was equally subdivided into four subplots on raised beds approx. 1·5 × 6 m. Subplots were planted in mid-April, 2008, with four rows per bed. The soil was prepared prior to seeding by ploughing and fertilized using 10-10-10 fertilizer (1 : 1 : 1 N, P, K) (Southern States Cooperative, Richmond, VA, USA). After seeding, the plot was irrigated every other day with well water. Spinach was harvested at baby stage (7–10 cm in length from the petiole base to leaf tip) approx. 30 days after seeding. To randomly select spinach for harvest, a hula-hoop (0·32 m2 of area) was thrown into the field, and spinach within the hoop was collected until approx. 2 kg of spinach leaves were harvested. The selected spinach plants were cut from the stem using scissors, placed in sterile plastic bags and transported to the laboratory on ice where they were processed within two hours.

Fresh spinach was received in the laboratory and immediately washed with sterile water to remove soil particles and dried using a household salad spinner. Ten grams of spinach leaves was transferred aseptically into polyethylene PD 961 bags (14 × 21 cm) (O2 transmission rate = 450 cc per 100 sq. in. and CO2 transmission = 1355 cc per 100 sq. in., Cryovac, Duncan S.C). The bags were sealed using an Ultravac® 225 vacuum packaging machine under a 30% vacuum (Ultravac solutions, Kansas, MO, USA). Packages of spinach were held at either 4 or 10°C for up to 15 days. Samples that were freshly packaged and those stored for 15 days at 4 and 10°C were chosen for pyrosequencing analysis based on major shifts in composition that were observed using DGGE (Denaturant Gradient Gel Electrophoresis) patterns (Lopez-Velasco et al. 2010).

DNA isolation

DNA was isolated from spinach leaves subjected to four different treatments: no treatment (fresh spinach), packaged and stored at 4°C for 1 day, packaged and stored at 4°C for 15 days, and packaged and stored at 10°C for 15 days. The leaves were aseptically transferred into Pulsifier® bags (Microbiology International, Frederick, MD, USA) containing 90 ml 1% (w/v) sterile peptone water (Sigma-Aldrich Co., St Louis, MO, USA) supplemented with 1% (v/v) Tween-90 (PTW) (Fisher Scientific, Atlanta, GA, USA), and samples were pulsified for 5 min to detach epiphytic microbiota. Bacterial cells were collected from the bacterial suspension by centrifugation at 2250 g for 20 min at 4°C washed with 1× PBS and resuspended in 100 μl 1× TE buffer. The cells were lysed by incubation with 300 μg of lysozyme (Fisher Scientific), 10 units of mutanolysin (Fisher Scientific) and 25 mg of achromopeptidase (Sigma-Aldrich) at 37°C for 30 min. Lysates were treated with 25 units of proteinase K (Fisher Scientific) and incubated at 65°C for 30 min. DNA was extracted from the lysed cells of each sample using the ZR soil microbe DNA kit™ (Zymo Research Co., Orange, CA, USA) per manufacturer’s instructions. DNA was isolated in triplicate from three replicates (packages) for 12 samples, resulting in 36 extractions.

The total number of 16S rRNA copies was assessed by real-time PCR using universal primers as described by Fierer et al. 2007. Samples were adjusted to approx. 106 copies of the 16S rRNA gene prior to amplification and further pyrosequencing of 16S rRNA gene.

PCR amplicons library construction

A nucleotide sequence length of 270–300 bp of the V4 region of the 16S rRNA gene was amplified by PCR ( Amplicons were generated using the forward primer 5′-GCCTCCCTCGCGCCATCAGAYTGGGYDTAAAGNG-3′ where the underlined sequence is that of 454 Life Science primer A (Roche, Base, Switzerland), and the italicized sequence is of the broad range V4 region primer. Separate amplifications were performed with three reverse primers as previously described (Jesus et al. 2010).




where the underlined sequence is that of 454 Life Science primer B (Roche), and the italicized sequence for the universal reverse primers ( The amplicon PCR products were generated by amplifying 50 ng of genomic DNA. Each 25 μl PCR contained 1·5 mmol l−1 of MgCl2, 50 mmol l−1 of KCl, 0·2 mmol l−1 of each dinucleotide, 1% of dimethylsulfoxide (DMSO), 25 mmol l−1 of Tris–HCl (pH 8), 1 U/μl−1 of HotStart-IT FideliTaq DNA polymerase (USB 71156, Cleveland, OH, USA), 0·5 μmol l−1 of each primer. The PCR protocol consisted of denaturation at 95°C for 5 min, followed by 30 cycles of 95°C for 30 s, amplification at 50°C for 45 s and elongation at 72°C for 1 min, followed by a final elongation step at 72°C for 7 min. Products of each primer set were cleaned using the Qiagen minElute PCR purification kit (Qiagen, Valencia, CA, USA) as per manufacturer’s instructions. Products of these three primer sets were then pooled in equal concentrations to create amplicon libraries (Kanagawa 2003). Reactions were performed for each replicate within a treatment, and resulting amplicons for each reaction and treatment were pooled creating four libraries. The concentration and quality of the products were assessed using a DNA1000 LabChip on the Bioanalyser 2100 (Agilent, Palo Alto, CA, USA). Only sharp, distinct amplification products with a total yield of >200 ng were used for 454 pyrosequencing.

Four separate libraries were constructed for each of the following conditions: fresh spinach, packaged spinach stored at 4°C for 1 day, packaged spinach stored at 4°C for 15 days, and packaged spinach stored at 10°C for 15 days. The amplicon libraries were bound to beads with two DNA molecules per bead. The beads were taken through the emPCR process using the Amplicon Primer A kit (Roche). After breaking the emulsion, the DNA strands were enriched and prepped for deposition onto the PicoTiter Plate with an eight gasket format (454 Life Sciences) to separate each individual sample. Pyrosequencing was performed on a Genome Sequencer GS FLX system (LR 700 100 cycles; Roche).

Preprocessing of sequence reads

The reads obtained from GS FLX were preprocessed to identify sequencing errors. Sequences were eliminated if they did not contain an exact match with the forward primer at the proximal end, if they were <50 nt before reaching reverse primer on the distal end, and if they did not contain at least 12 bases of the reverse primer before the read ended. Reads were then grouped into clusters containing other sequences that were 99% similar over 99% of their lengths. A unique representative read sequence was selected randomly from each cluster (unique sequence) for further taxonomic classification and operational taxonomic unit (OTU) assignment.

Taxonomic assignment of sequence reads

Metagenomes of fresh and packaged spinach were analysed using the RDP pyrosequencing pipeline ( release 10 (Cole et al. 2009). Each unique sequence was aligned using the RDP pyrosequencing function, Aligner to generate phylogenetically ordered rRNA sequences (44). Aligned data sets were clustered using the default parameters for the RDP Clustering function. The resulting clusters were utilized to calculate the Shannon index, Chao 1 estimator and rarefraction curves using the pyrosequencing analysis tools from RDP at the level of 3% dissimilarity, approximately species level. The RDP sample abundance statistics tool was utilized to calculate the Jaccard’s index to compare the four spinach microbiomes (Chao et al. 2006). Aligned data sets from each microbiome were merged into a single cluster file, which was utilized to construct a distance matrix at 3% dissimilarity, that produce a dendrogram using unweighted pair group method with arithmetic mean (UPGMA). The RDP Classifier, a Bayesian rRNA classifying algorithm (Wang et al. 2007), was used to assign phylogenetic groups, based on sequence similarity. Matches with a RDP confidence estimate below 60% were designated as unclassified bacteria.

Raw sequences of spinach microbiomes were deposited on National Center for Biotechnology Information, utilizing the Sequences Read Archive (SRA) with the accession number SRA012571.11 (


Characteristics of the sequenced data

The average sequence length of the pyrosequencing data was 200 bp (range, 100–281). The total number of reads for each sample in each gasket ranged from 14 000 to 24 000 (Table 1) depending on the sample. Reads that met the quality parameters were trimmed. Sequence reads that belonged to chloroplast sequences (3–15% of reads) were considered as contamination and were not analysed.

Table 1.   Number of reads, unique sequences and trimmed reads for spinach microbiomes
SampleTotal readsSequences trimmedUnique
Fresh spinach22 16230668862
Packaged spinach
Stored 4°C/day 117 75740645911
Stored 4°C/day 1514 68618629467
Stored 10°C/day 1524 294391517 404

The majority of the unique sequences belonged to the domain Bacteria (99% of total sequences). Sequences obtained from the fresh spinach samples and packaged spinach stored at 4°C for 1 day were assigned to the domain Bacteria, but no further classification of the sequences could be made for 75 and 85% of the sequences, respectively. For samples stored at 4 and 10°C for a period of 15 days, the percentage of classified bacteria increased to 60 and 90%, respectively.

Fresh spinach microbiome

Bacterial sequences that were classified from fresh spinach microbiome were assigned to 11 phyla (Fig. 1a,b). The majority of the sequences belonged to Proteobacteria (80%), particularly the classes of α-Proteobacteria and γ-Proteobacteria (Fig. 1a). Bacteria belonging to the phyla Firmicutes, Chlamydiae, Gemmatimonadetes, Verrucomicrobia, TM7, Deinococcus-Thermus, Planctomycetes, Actinobacteria, Acidobacteria and Bacteroidetes were also identified (Fig. 1b). From all the classified sequences, 54% were attributed to a genus, and 46% were assigned to taxa only at the family level. Within the identified phyla, 75 different genera were identified (Table 2). The most represented sequences (each representing more than 1% of the total classified bacteria) belonged to the genera Spartobacteria spp., Deinococcus spp., Gp4, Gp6, Methylobacterium spp., Rhizobium spp., Brevundimonas spp., Acinetobacter spp., Pseudomonas spp., Ralstonia spp., Sphingomonas spp., Duganella spp., Naxibacter spp. and Massilia spp. and the families Propionibacterineae and Micrococcineae. The most numerous sequences were identified as Pseudomonas spp. (9%).

Figure 1.

 Relative phylum/class abundance of sequences that were classified beyond the Bacteria domain. (a) Represents percentage of relative abundance of class for Proteobacteria phylum. (80% of sequences) (b) Phyla classification of the remaining 20% of sequences. Treatments: (1) fresh spinach, (2) spinach stored at 4°C for 1 day, (3) spinach stored at 4°C for 15 days and (4) spinach stored at 10°C for 15 day. (a) (inline image) α-Proteobacteria; (inline image) β-proteobacteria; (inline image) γ-proteobacteria and (inline image) δ-proteobacteria. (b) (inline image) Firmicutes; (inline image) Actinobacteria; (inline image) Bacteroidetes; (inline image) Deinococcus-Thermus; (inline image) Gemmatimonadetes; (inline image) Acidobacteria; (inline image) Verrucomicrobia; (inline image) Planctomycetes; (inline image) TM7 and (inline image) Chlamydiae.

Table 2.   Taxonomic identification of sequences detected in the fresh spinach microbiome*
Taxonomical classification%Taxonomical classification%
  1. *The percentage of sequence reads was calculated based on the total number of reads classified beyond the Bacteria domain using the RDP Classifier tool.

Chlamydiae0·05Ferruginibacter spp.0·05
 Gemmatimonas0·54 Rubellimicrobium spp.0·09
Chloroflexi Rhodopila spp.0·13
 Sphaerobacterales0·09 Roseomonas spp.0·32
   Novosphingobium spp.0·27
Verrucomicrobia Sphingomonas spp.6·43
 Subdivision3 genera incertae sedis0·05 Sphingosinicella spp.0·45
 Xiphinematobacter0·05 Porphyrobacter spp.0·09
 Spartobacteria genera incertae sedis1·21 Aurantimonas spp.0·18
 Unclassified Verrucomicrobia0·05 Mesorhizobium spp.0·05
TM7 Balneimonas spp.0·05
 TM7 genera incertae sedis0·72 Bosea spp.0·09
Deinococcus-Thermus Bradyrhizobium spp.0·05
 Truepera spp.0·18 Labrys spp.0·09
 Deinococcus spp.2·97 Methylobacterium spp.1·21
Bacteria incertae sedis Beijerinckia spp.0·14
 Ktedonobacter spp.0·05 Devosia spp.0·18
Planctomycetes Hyphomicrobium spp.0·09
 Singulisphaera spp.0·75 Rhizobium spp.1·22
Actinobacteria Phenylobacterium spp.0·13
 Iluminobacter spp.0·05 Brevundimonas spp.2·79
 Desulfosporosinus spp.0·05 Unclassified Rhodospirillales0·32
 Kineosporiineae0·05 Unclassified Sphingomonadaceae2·38
 Streptosporangineae0·05 Unclassified Sphingomonadales1·08
 Micromonosporineae0·14 Unclassified Bradyrhizobiaceae0·13
 Pseudonocardineae0·09 Unclassified Methylobacteriaceae0·22
 Frankineae0·36 Unclassified Beijerinckiaceae0·05
 Corynebacterineae0·41 Unclassified Hyphomicrobiaceae0·32
 Propionibacterineae2·61 Unclassified Rhizobiales1·22
 Micrococcineae1·39 Unclassified α-Proteobacteria1·04
 Unclassified Solirubrobacterales0·14γ−Proteobacteria
 Unclassified Rubrobaceridae0·05 Acinetobacter spp.1·53
 Unclassified Actinomycetales0·14 Pseudomonas spp.9·18
 Unclassified Actinobacteria0·18 Steroidobacter spp.0·05
Firmicutes Pseudoxanthomonas spp.0·05
 Exiguobacterium spp.0·05 Stenotrophomonas spp.0·31
 Bacillus spp.0·27 Dokdonella spp.0·05
 Paenibacillus spp.0·05 Luteimonas spp.0·05
 Unclassified Paenibacillaceae0·05 Dyella spp.0·27
 Unclassified Bacillales0·63 Lysobacter spp.0·27
Bacteroidetes Serratia spp.0·13
 Mucilaginibacter spp.0·05 Pantoea spp.0·13
 Pedobacter spp.0·05 Unclassified Pseudomonaceae0·40
 Dyadobacter spp.0·05 Unclassified Xanthomonadaceae1·40
 Hymenobacter spp.0·45 Unclassified Enterobacteriaceae13·7
 Segetibacter spp.0·05 Unclassified γ-Proteobacteria1·03
 Gp130·05 Nitrospira spp.0·05
 Gp10·31 Aquaspirillum spp.0·05
 Gp70·09 Aquabacterium spp.0·05
 Gp170·05 Methylibium spp.0·05
 Gp160·45 Burkholderia spp.0·23
 Gp42·02 Ralstonia spp.1·22
 Gp62·65 Polaromonas spp.0·36
 Gp30·36 Acidovorax spp.0·09
δ-Proteobacteria Variovorax spp.0·14
 Bacteriovorax spp.0·05 Undibacterium spp.0·04
 Geobacter spp.0·05 Duganella spp.1·03
 Nannocystaceae0·05 Janthinobacterium spp.0·04
 Polyangiaceae0·36 Naxibacter spp.3·73
 Cystobacteraceae0·09 Massilia spp.7·38
 Unclassified Desulfomonadales0·05 Unclassified Alcaligenaceae0·09
 Unclassified Rhodobacteraceae0·14 Unclassified Burkholderiaceae0·09
 Unclassified δ-Proteobacteria0·14 Unclassified Comamonadaceae2·30
   Unclassified Oxalobacteraceae8·55
   Unclassified Burkholderiales1·23
   Unclassified β-Proteobacteria0·94
   Unclassified Proteobacteria2·21

Estimates of the OTUs were assessed using rarefraction curves (Fig. 2) and a nonparametric method, Chao 1 (Table 3). Rarefraction curves reached saturation at a distance level of 20% (phylum level – Fig. S1a–d) but not at 3 or 5% of dissimilarity.

Figure 2.

 Rarefraction curves for fresh and packaged spinach indicating the observed number of operational taxonomic units within the 16S rRNA microbiomes derived from spinach leaves. Curves were calculated with RDP using pyrosequencing tools at 3% of dissimilarity. (inline image) Fresh spinach; (inline image) packaged spinach (4°C/day 1); (inline image) packaged spinach (4°C/day 15) and (inline image) packaged spinach (10°C/day 1).

Table 3.   Richness and diversity estimators that predict the number of species in fresh and packaged spinach at 3% of dissimilarity
SampleChao 1 estimator*Richness (rarefraction)*Shannon index (H′)Species evenness (E)†
Lower limitChao 1Upper limitLower limitRichnessUpper limit
  1. *Richness estimated with Chao 1 and rarefraction curves represent values obtained at 3% of dissimilarity.

  2. †Evenness was calculated as E = H′/Hmax where Hmax = ln S being S the total number of species in the sample, estimated with Chao 1.

Fresh spinach194517412204978·78977976·484·21 ± 0·590·56
Packaged spinach
 Stored 4°C/day1981·011401358551·51551552·383·77 ± 0·590·53
 Stored 4°C/day 15840·0942·01083571·47570570·363·90 ± 0·610·56
 Stored 10°C/day 15947·010681233622·78623624·723·51 ± 0·540·50

Changes in bacterial community composition during storage of packaged spinach at refrigeration temperatures

Storage at refrigeration temperatures affected the microbiome composition. Microbiomes of fresh and packaged spinach stored at refrigerated temperatures were compared using Jaccard’s index. The microbiome of fresh spinach were more similar to the microbiomes from packaged spinach stored at 4°C for 1 or 15 days than to the microbiomes from spinach held at 10°C for 15 days, which formed a separate cluster (Fig. S1e). The total number of OTUs predicted at 3% dissimilarity was 977 for fresh spinach and reduced rapidly when stored at 4°C to 551 OTUs after 1 day and 570 OTUs after 15 days (Table 3). A similar decrease, to 623 OTUs, was seen at 10°C after 15 days storage. From all the sequences that were classified, there were 165 different bacteria and/or families, from those only 40 are shared in all the four conditions. Chao 1 estimator, Shannon’s index and rarefraction curves indicated a reduction in richness, evenness and diversity after spinach was packaged and stored at 4°C or at 10°C for 15 days (Table 3 and Fig. 2). An increase in Shannon’s index was calculated for packaged spinach stored at 4°C for 15 days and was associated with decreased evenness.

Reductions in the richness, evenness and diversity of the community are attributed to a decrease in abundance of sequences belonging to all phyla, and an increase in relative abundance of γ-Proteobacteria (Fig. 1a). These reductions were most pronounced for spinach held at 10°C after 15 days of storage (Fig. 1b), particularly the families of Oxalobacteraceae, Rhizobiales, Acinetobacter and Sphingomonadaceae. Despite an overall reduction in relative abundance of sequences belonging to the families Enterobacteriaceae and Micrococcineae, the genera of Pseudomonas spp., Stenotrophomonas spp., Pantoea spp. and Escherichia spp. (Table 4) increased in abundance in the microbiome of spinach stored at 10°C for 15 days. The microbiome of spinach stored at 4°C for 15 days was dominated by sequences belonging to the family Enterobacteriaceae (52% of the total classified bacteria) and the genera Pseudomonas spp. (28% of total classified bacteria), while the relative abundance of most, including Deinococcus spp. and Sphingomonas spp. was reduced (Table 4).

Table 4.   Comparison of spinach microbiomes during storage at 4 and 10°C for up to 15 days*
PhylaFamily and/or genusRelative abundance (%)†
Fresh4°C/day 14°C/day 1510°C/day 15
  1. ND Sequences that were not detected in the microbiomes.

  2. *Representative bacterial genera that showed major shifts in bacteria community composition during storage, relative abundance of the complete microbiomes are in Table S1.

  3. †The percentage of relative abundance was calculated based on the total number of classified beyond the Bacteria domain using the RDP Classifier tool.

  4. ‡Sequence reads classified only at family level.

FirmicutesExiguobacterium spp.0·050·20ND0·21
Unclassified α-Proteobacteria1·040·600·130·02
Sphingomonas spp.6·4410·04ND0·87
Methylobacterium spp.1·220·503·490·18
Rhizobium spp.1·221·000·130·10
Unclassified γ-Proteobacteria1·049·542·141·59
Acinetobacter spp.1·530·400·110·04
Pseudomonas spp.9·1914·7627·6626·25
Stenotrophomonas spp.0·320·200·241·28
Pantoea spp.0·140·100·441·14
Escherichia spp.NDND0·0523·07
Unclassified β-Proteobacteria0·950·600·040·02
Ralstonia spp.1·220·800·040·01
Naxibacter spp.3·741·200·420·04
Massilia spp.7·382·310·223·20


Pyrosequencing of 16S rRNA genes from the spinach phyllosphere identified more than 1000 OTUs, a greater richness than the microbiome of human gut (300 OTUs) but less diverse than soil and marine environments, whose richness is estimated in a range of 3000–6000 OTUs and 9000 OTUs (Huber et al. 2007; Roesch et al. 2007; Acosta-Martinez et al. 2008; Andersson et al. 2008; Gilbert et al. 2009). Recent studies that used pyrosequencing to characterize the phyllospheres of soybean, clover and Arabidopsis thaliana, detected approx. 600 OTUs (Delmotte et al. 2009), while pyrosequencing of several different tree species identified 5476 unique bacterial OTUs, but only 252 OTUs per tree species, none of which were shared. The diversity of the phyllosphere is even greater than described here, or in other studies, as the rarefaction curves did not asymptote indicating not all the lineages have been surveyed (Fig. 2). In this study, the majority (75%) of the sequences were classified only within the bacterial domain, representing novel sequences that have not been placed into a recognized phylum within the RDP database or represent bacterial divisions not yet described (Andersson et al. 2008), indicating an even greater number of rare lineages within the phyllopshere. However, the membership of all phyllosphere communities examined to date was similar suggesting the phyllosphere is colonized by members who have adapted to the harsh conditions including low nutrient availability, temperature fluctuations and exposure to UV light. Variations in the individual members can be attributed to differences in plant features (leaf topography, metabolic profile) and environmental factors that may determine the establishment of epiphytes on plant surfaces (Whipps 2001).

Pyrosequencing of 16S rRNA genes from the spinach phyllosphere identified members belonging to 11 different phyla (Fig. 1a,b), indicating greater diversity of the phyllosphere than recognized using culture-based techniques that mostly identify members of the Firmicutes and Proteobacteria phyla (Thompson et al. 1993). Bacterial taxa such as Enterobacteriaceae and Pseudomonadaceae, which are commonly isolated using culture-based studies, were also identified here in high relative abundance indicating that these are dominant members of the phyllosphere (Ibekwe and Grieve 2004; Ragaert et al. 2007). The most abundant sequences in this study, Sphingomonas spp., Brevundimonas spp., Methylobacterium spp., Acinetobacter spp., Pseudomonas spp., Naxibacter spp. and Massilia spp., have been previously detected through analysis of 16S rRNA gene in the phyllospheres of other plants (Rheims et al. 1996; Yang et al. 2001; Kadivar and Stapleton 2003; Krimm et al. 2005; Jackson et al. 2006; Knief et al. 2010; Yutthammo et al. 2010). Methylobacterium spp., Sphingomonas spp. and Pseudomonas spp. were recently described as the most active members of the phyllosphere of soybean, suggesting these bacteria may also play an important role on the spinach phyllosphere (Delmotte et al. 2009). Deinococcus-Thermus, β-Proteobacteria, Acidobacteria and low relative abundances of sequences belonging to the phyla Chlamydiae, Gemmatimonadetes, Verrucomicrobia, TM7, Planctomycetes, Actinobacteria, Acidobacteria and Bacteroidetes were comparable to that reported in phyllospheres of other plants (Table 2) (Jackson et al. 2006; Yang et al. 2008). Additionally, the spinach microbiome contained novel sequences for genera within the phyla Planctomycetes, Verrucomicrobia and TM7 not previously described as associated with the phyllosphere (Table 2). Members of the Planctomycetes and the division TM7 possess the ability to survive and grow under a wide range of conditions, such as exposure to UV light, which is also encountered on the phyllosphere (Hugenholtz et al. 2001; Kulichevskaya et al. 2007). Verrucomicrobia is a prevalent member of the rhizosphere, and it was recovered from soils, possessing characteristics broadly distributed among rhizosphere members like aerobic, heterotrophic and slow growth (da Rocha et al. 2009). The presence, even in lower abundance on the leaf surface, may reflect greater ecological importance of these poorly studied phyla and may possess important functions.

In this study, pyrosequencing of the 16S rRNA gene amplicons was utilized to identify the effects of typical retail storage conditions (4°C) or temperature abuse conditions (10°C) on the composition and diversity of the spinach phylloepiphytic bacterial community. Previous studies by this research group indicate that length and temperature of storage of packaged spinach were associated with changes to the spinach bacterial community (Lopez-Velasco et al. 2010). Application of DGGE community analysis indicates that the overall species richness decreases with prolonged storage at 4 or 10°C. Pronounced changes in richness were observed for samples that were packaged and held at 4°C for 1 day, 4°C for 15 days and 10°C for 15 days and were thus chosen for further examination using pyrosequencing which allows an in-depth identification of the members responsible for the observed changes. The richness of the spinach microbiome decreased from representing 11 phyla to 5 phyla after storage at 4°C for only 1 day. No unique OTUs were detected on the packaged spinach held at either temperature that was not present on the fresh spinach prior to packaging. However, only 40 OTUs, belonging to the β- and γ-Proteobacteria, Acidobaceria and Actinobacteria, were identified from spinach subjected to the four treatments (Table S1), indicating these may be the most robust members of the community.

A reduction in the abundance of the majority of the families, so that only a few genera were maintained on the packaged leaves after 15 days (Table 4), was associated with the substantial decreases in richness and evenness when spinach was stored for 15 days at 4 or 10°C (Table 3). Sequences within the α- and β-Proteobacteria, Acidobacteria and Deinococcus-Thermus decreased with refrigeration compared with fresh spinach, which suggests that these members are prone to negative selection under low temperatures. These results were most significant after 15 days of storage, suggesting that length, temperature and atmospheric conditions during storage led to the decreased populations of these bacteria. This effect could also be related to increased dominance of Proteobacteria, which can outcompete members of these genera, or owing to the larger relative abundance of sequences belonging to these genera, these phyla were not detected. The refrigerated microbiomes were composed of larger numbers of sequences belonging to genera associated with psychrotrophic growth, including members of the family Enterobacteriaceae, Corynebacterineae, Micrococcineae and the genera Pseudomonas, Methylobacterium and Pantoea (Fig. 1a and Table 4). Previous studies have noted that psychrotrophic bacteria, particularly Pseudomonas sp., Erwinia sp., Rhanella sp., and lactic acid bacteria, tend to dominate the culturable populations on vegetables (Rudi et al. 2002; Randazzo et al. 2009). Many of these bacteria have been reported as responsible for spoilage and deterioration of fresh vegetables (Kraft 1992; Nguyen-the and Carlin 1994; ICMSF 2005).

Conditions associated with temperature abuse were associated with an increase in the potentially pathogenic species including Stenotrophomonas spp., Massilia spp. and E. coli spp. (Chatelut et al. 1995; La Scola et al. 1998). Storage conditions at 10°C, considered as temperature abuse, showed an increase in bacteria from the Enterobacteriaceae family. Of particular concern was the increase in relative abundance for Escherichia spp. (23% of the total classified sequences) when spinach was stored at 10°C but not at 4°C, indicating the importance of temperature and time of storage on the proliferation of these potentially harmful bacteria. Previous studies have concluded that while E. coli O157:H7 can increase 2 log CFU g−1 on packaged spinach held at 8°C, no growth occurs below 8°C though E. coli O157:H7 can persist for up to 12 days of storage at 4°C (Abdul-Raouf et al. 1993; Francis and O’Beirne 2001; Delaquis et al. 2007). This persistence of E. coli O157:H7 may be aided by commensal interactions with the native microflora, whose different carbon source utilization patterns allow co-existence and even increase populations (Cooley et al. 2006, Lopez-Velasco et al. 2010). Sequences belonging to bacteria associated with antagonism of E. coli O157:H7 growth were detected in large numbers in the spinach microbiome. These include members of the genera Enterobacter, Exiguobacterium, Pseudomonas, Kluyvera (Johnston et al. 2009; Lopez-Velasco 2010). This suggests that microbial interactions that occur postprocessing and during storage in produce might play a major role in the survival of pathogenic bacteria (Bhagwat 2006), especially if conditions favour their growth. In vitro antagonism of the growth of E. coli O157:H7 and Salm. enterica has been described for many of the genera identified in the spinach microbiome including Pseudomonas spp. (Schuenzel and Harrison 2002), Bacillus spp. (Liao 2007), Enterobacter spp. (Cooley et al. 2006) and Pantoea spp. (Johnston et al. 2009). Pseudomonas spp. also increased with prolonged storage suggesting this genus may provide new targets for biocontrol of potential human pathogens in packaged produce. Identifying bio-control strains that are adapted to the phyllosphere environment and that persist during postharvest operations may provide a strategy to improve the safety of processed ready-to-eat vegetables.


The application of pyrosequencing to describe composition and diversity of the phyllosphere on spinach leaves provided a broader outlook of the bacterial composition of this community complementing other phyllosphere studies that have used culture- and nonculture-dependent approaches. This study showed that the spinach microbiome is highly diverse and changes dynamically during storage at refrigeration temperatures with the establishment of a dominant population mostly composed of γ-Proteobacteria. In this study, we have shown that abundance and diversity of the spinach microbiota were altered during storage at low temperatures with selection towards psychrotrophic micro-organisms indicating the adaptation of certain members of the community to this food preservation method. Further research is necessary to correlate the composition and changes in diversity with functionality of the phyllosphere community to better understand the phyllosphere microbial ecology of edible plants and identify better strategies to reduce growth of spoilage and/or pathogenic bacteria.


This work was supported in part by Virginia Bioinformatics Institute and Fralin Life Science Institute Exploratory Grant. Fellowship for partial financial support of Gabriela Lopez-Velasco was provided by the National Council for Science and Technology (CONACyT) from Mexico.