Foliar Roughness and Water Content Impact on Escherichia coli Attachment in Baby Leafy Greens

Simple Summary Vegetables may host human pathogens, such as Salmonella enterica and Escherichia coli. Therefore, the consumption of salads may raise safety concerns as they are eaten raw and often without prior proper washing. Bringing to light the differences between different salad crops in the susceptibility to human pathogen contamination and understanding the relationship between the susceptibility and leaf traits can contribute to increase food safety. We evaluated the susceptibility to E. coli attachment of 30 different baby leaves (leafy vegetables harvested at an early stage of growth and consumed as salads), finding inter- and intraspecific variation. In romaine lettuce, the most involved in food disease outbreaks in the U.S., we found genotypes (‘Maraichere’) less susceptible to contamination than others (‘Bionda degli Ortolani’). Among the 30 baby leaves, ‘Bionda degli Ortolani’, rocket and Swiss chard were the most susceptible, with wild rocket, wild lettuce and lamb’s lettuce the least. Further analysis of these six baby leaves revealed that leaf roughness and water content were positively correlated with attachment level. In rougher leaves, we observed a lower effectiveness of UV treatment in reducing attachment. Abstract Understanding the relation between the susceptibility of different leafy greens to human pathogen contamination and leaf traits can contribute to increase the food safety of the fresh vegetable industry. The aim of this research was to evaluate the susceptibility to E. coli ATCC 35218 attachment in 30 accessions of baby leaves, and to identify leaf traits potentially involved in the contamination. The accessions were surface inoculated with a bacterial suspension containing 1 × 107 cells/mL and the attachment was measured 1.5 h after inoculation. Significant differences in attachment were detected between the accessions for p ≤ 0.05. The three most and the three least susceptible accessions were selected and characterized for leaf micro-morphological traits (stomata density and size, surface roughness) and water content. Scanning electron microscopy was used to analyse the stomatal parameters. Roughness was measured by an innovative portable 3D digital microscope. No significant correlation between the attachment of E. coli ATCC 35218 and stomatal parameters was detected, while the attachment was positively correlated with roughness and water content. The E. coli ATCC 35218 population in surface-inoculated leaves was also measured after a UV treatment, which was found to be less effective in reducing bacterial contamination in the rougher leaves. This result suggested that roughness offers UV protection, further highlighting its impact on the microbiological safety of baby leafy greens.


Introduction
Baby leaves are leafy vegetables harvested at an early stage of growth (up to the eighth true leaf) and consumed as salads [1]. They can be commercialized as unprocessed After sowing, seeds were kept for 48 h in the dark at 20 • C for promoting germination. Seedlings were hydroponically grown in a floating system in a growth chamber at 21 ± 2 • C (day) and 14 ± 2 • C (night) with a photoperiod of 16 h under fluorescent lighting units OSRAM L36W/77 (36 W, 120 cm in length, 26 mm in diameter) up to the baby-leaf stage (five/six weeks after sowing, depending on the species). A full-strength Hoagland's nutrient solution N 15.0 mM, p 0.10 mM, K 6.0 mM, Ca 5.0 mM, Mg 2.0 mM, Fe 50.0 µM, B 46.2 µM, Mn 9.2 µM, Zn 0.78 µM, Cu 0.32 µM, Mo 0.12 µM) was used. Baby leaves were cut at the base of the petiole, and immediately used for the bacterial attachment assay. The average fresh weight, dry weight, leaf number, and leaf area of the baby leaf plants are reported in Table S1. Fresh weight (FW) and dry weight (DW) were measured before and after oven desiccation at 80 • C for 48 h or until a constant weight, and water content was calculated as [(FW-DW)/FW]*100. Leaf area was determined by a planimeter LI-3000 (LI-COR, Lincoln, NE, USA). Colour was measured with the Handy Colorimeter NR-3000 (Nippon Denshoku Kogyo C., LTD.) and expressed as a*, b* and L* values (Table S2).

Bacterial Strain
Escherichia coli ATCC 35218 was used for the inoculation assays. This strain (classified as biosafety level 1) has specific virulence genes associated with different E. coli pathotypes related to human and animal infections [26].

Surface Inoculation of the Baby Leaves
The bacterial inocula was started from a −20 • C glycerol stock. Briefly, one mL from an overnight inoculum in Lysogeny Broth (LB, Oxoid, UK) was washed and diluted in sterile Physiological Solution (PS, NaCl 0.85% w/v in H 2 O) to clean the cells from LB medium [27]. Disks 1.5 cm in diameter were cut from the leaves of the 30 accessions used in this study and put in Petri plates. Fifty µL of the working bacterial suspension (containing approximately 1 × 10 7 cells/mL) were placed onto the centre of the leaf disk on the adaxial side. Leaves were incubated for 1.5 h at 25 • C in static. After incubation, the disks were washed three times in 30 mL of sterile PS in Petri plates to remove unattached bacterial cells. The leaf disks were placed on the adaxial side on the first Petri plate and gently washed by manually rotating the plates clockwise 30 times; then, the same procedure was repeated on the second and the third plate. Rinsed disks were grinded with a mini-pestle (into 1.5 mL tubes) in 0.5 mL of PS. After grinding, 20 µL were plated onto selective and differential media MacConkey Agar (Oxoid, UK) and incubated at 37 • C overnight. Colony-forming units (CFU) were counted and log base 10 (log) transformed. Two different seeding sets and the third true leaf of 5 plants at the baby leaf stage from the same seeding set were used for each accession (10 leaves in total/accession).
Among the 30 accessions, we selected the three most and three least contaminated.

Analysis of Leaf Micro-Morphological Traits
Stomatal parameters and roughness were determined in the six selected accessions.

Stomata
Scanning electron microscopy (SEM) was used to analyse the stomata density and size (length, width and stomatal rim area) of the adaxial leaf surface. Before SEM observations, fresh leaf samples were coated by a thin layer of gold (~10 nm) to reduce shrinkage while ensuring the preservation of cell structures as close to the natural state as possible [28]. Eight observations (the third true leaf of 4 plants at the baby leaf stage and 2 sections per leaf) were performed in each accession. Photographs at 1 × 300 and 1 × 600 were used for counting stomata and measuring stomata sizes, respectively. Stomata were counted on the images (view field ranging from 923 to 929 µm 2 ), and then the number per surface unit (mm 2 ) was calculated by a simple proportion. ImageJ software [29] was used to measure stomata size; the measure was expressed in µm by means of a proportion considering the bar of known length present in each image.

Roughness
A portable prototype of 3D digital microscope was used to measure the leaf surface roughness [30]. Basically, its setup includes three elements: an optoelectronic imaging group (a CCD digital camera and a fixed focus optical element), a calibrated stage for translating this group along the optical axis, and a smart polymeric lightning system [31]. The 3D reconstruction of the surface under inspection was obtained by elaborating a sequence of defocused images (Video S1) and implementing a suitable algorithm based on shape from a focusing technique [32]. Surface roughness parameters were extracted from 3D reconstruction, according to ISO standards [33]. In particular, we focused the analysis of plant surface on the average roughness (Ra) and root mean square roughness (Rq).
A suitable optical element was used in order to achieve a field of view of about 7 × 5.2 mm and a vertical resolution of about 10 µm. It is worth noting that these kinds of measurements are completely non-contact, as well as non-destructive; this overcomes the disadvantages of using a contact stylus profilometer with soft biological samples. Five different analyses in two leaves (10 replicates) were performed for each accession.

UV Treatment
The leaves of the six selected accessions were surface-inoculated as described above (5 leaf disks per accession) and then exposed for 5 min under a UV lamp (UVC ≥ 90% with 108.4 µW/cm 2 from 0.5 m) placed at a distance of 10 cm. After exposure, CFU were counted and log transformed as mentioned above. As a control, three separate drops of 50 µL of 10 7 E. coli cell/mL inoculum were put onto a smooth plastic surface and exposed to the above described UV treatment. The effectiveness of the treatment on the control was demonstrated by complete bacterial inactivation after plating onto MacConkey Agar.

Statistical Analysis
Data were subjected to the Shapiro-Wilk test for normality and Levene's test for homogeneity of variances. Analysis of variance (ANOVA) was performed using the CoStat software (version 6.45, Monterey, CA, USA). Significance was set at p ≤ 0.05 (Tukey test). The Pearson's correlation test (p ≤ 0.05) was used to determine the relationship between the E. coli ATCC 35218 attachment and all the considered leaf traits.

E. coli Attachment in the Baby Leaves
Variability in E. coli ATCC 35218 attachment was observed among the 30 studied baby leaf accessions, even within the same species ( Table 2). The romaine lettuce 'Bionda degli Ortolani' showed the highest contamination, with significant differences compared to wild chicory (local accession), romaine lettuce 'Maraichere', lettuce 'Pamela', sorrel, wild rocket 'Yeti', wild lettuce and lamb's lettuce 'Trophy F1 (p < 0.001). The latter resulted in being less susceptible than all the other accessions. Considering the data by species, differences were found not only in Lactuca sativa, but also in Diplotaxis tenuifolia; conversely, no differences were observed within Taraxacum campylodes, Brassica juncea, Brassica rapa, Cichorium intybus and Beta vulgaris. Among the lettuces, 'Bionda degli Ortolani' showed a contamination level of 3.36 ± 0.36 log CFU/cm 2 , significantly different from 'Maraichere' (2.65 ± 0.34 log CFU/cm 2 ) and 'Pamela' (2.45 ± 0.20 log CFU/cm 2 ). In Diplotaxis tenuifolia, 'Yeti' was less susceptible than the Ingegnoli accession.
Based on the results of the attachment assays, we selected six accessions (romaine lettuce 'Bionda degli Ortolani', Swiss chard, and rocket as the most susceptible, and wild rocket 'Yeti', wild lettuce, and lamb's lettuce 'Trophy F1 as the least susceptible; Figure 1). Further analyses aimed to identify traits possibly involved in the bacterial retention. rocket 'Yeti', wild lettuce, and lamb's lettuce 'Trophy F1′ as the least susceptible; Fig  1). Further analyses aimed to identify traits possibly involved in the bacterial retentio

Leaf Micro-Morphological Traits and Water Content
The SEM images ( Figure 2) highlighted significant differences in stomata density size (length, width and rim area) between the six selected baby leaf accessions (Table  In wild lettuce, these parameters were not measured as this species did not have stom on the adaxial leaf surface ( Figure 2E). Some of the observed differences were consis with the contamination data: e.g., the absence of stomata in wild lettuce, the highest lowest stomata density in rocket (92.20 ± 28.40) and lamb's lettuce 'Trophy F1′ (57.1 5.47), respectively, and the highest stomatal rim area in Swiss chard (45.58 ± 6.78 μ However, no significant correlation of E. coli attachment ATCC 35218 versus any of stomatal parameters was detected (data not shown).
Three-dimensional digital microscopy revealed significant differences (p < 0. between the accessions for both the roughness parameters, with romaine lettuce 'Bio

Leaf Micro-Morphological Traits and Water Content
The SEM images (Figure 2) highlighted significant differences in stomata density and size (length, width and rim area) between the six selected baby leaf accessions (Table 3). In wild lettuce, these parameters were not measured as this species did not have stomata on the adaxial leaf surface ( Figure 2E). Some of the observed differences were consistent with the contamination data: e.g., the absence of stomata in wild lettuce, the highest and lowest stomata density in rocket (92.20 ± 28.40) and lamb's lettuce 'Trophy F1 (57.14 ± 5.47), respectively, and the highest stomatal rim area in Swiss chard (45.58 ± 6.78 µm 2 ). However, no significant correlation of E. coli attachment ATCC 35218 versus any of the stomatal parameters was detected (data not shown).
Three-dimensional digital microscopy revealed significant differences (p < 0.001) between the accessions for both the roughness parameters, with romaine lettuce 'Bionda degli Ortolani ' Figure 3. A rough estimate of the Ra parameter can be obtained considering the differences between the maximum and minimum height values, which are higher in 'Bionda degli Ortolani' ( Figure 3A) and lower in lamb's lettuce 'Trophy F1 ( Figure 3F). The Pearson's test revealed a significant correlation between contamination level and both Ra and Rq values ( Figure 4A,B).
'Trophy F1′ (Ra = 24.74 ± 10.92; Rq = 31.38 ± 14.24) ( Table 3). From a visual point of view, the differences in leaf roughness can be observed in the 3D reconstructions shown in Figure 3. A rough estimate of the Ra parameter can be obtained considering the differences between the maximum and minimum height values, which are higher in 'Bionda degli Ortolani' ( Figure 3A) and lower in lamb's lettuce 'Trophy F1′ ( Figure 3F). The Pearson's test revealed a significant correlation between contamination level and both Ra and Rq values ( Figure 4A,B).
Moreover, significant differences (p < 0.001) between the selected accessions were observed in leaf water content ( Table 3). The accession most susceptible to contamination ('Bionda degli Ortolani') showed the highest leaf water content (95.36 ± 0.84%) and differed from all the others, but especially from the least susceptible accessions, i.e., wild lettuce (89.79 ± 1.13%) and lamb's lettuce 'Trophy F1′ (90.22 ± 2.30%). The correlation between E. coli ATCC 35218 attachment and water content was significant ( Figure 4C).  Moreover, significant differences (p < 0.001) between the selected accessions were observed in leaf water content ( Table 3). The accession most susceptible to contamination ('Bionda degli Ortolani') showed the highest leaf water content (95.36 ± 0.84%) and differed from all the others, but especially from the least susceptible accessions, i.e., wild lettuce (89.79 ± 1.13%) and lamb's lettuce 'Trophy F1 (90.22 ± 2.30%). The correlation between E. coli ATCC 35218 attachment and water content was significant ( Figure 4C).

UV Experiment
In the six selected accessions, E. coli ATCC 35218 survival was measured in the surfaceinoculated leaves after UV treatment. UV rays significantly reduced E. coli ATCC 35218 contamination in all the accessions ( Figure 5). Nevertheless, a lower reduction in the attachment was achieved in romaine lettuce 'Bionda degli Ortolani' (−8%) and rocket (−17%) compared to the accessions showing lower roughness (wild rocket 'Yeti', wild lettuce and lamb's lettuce 'Trophy F1 ; −47% on average) ( Figure 5). Swiss chard differed from 'Bionda degli Ortolani' and wild rocket 'Yeti'.

UV experiment
In the six selected accessions, E. coli ATCC 35218 survival was measured in the surface-inoculated leaves after UV treatment. UV rays significantly reduced E. coli ATCC 35218 contamination in all the accessions ( Figure 5). Nevertheless, a lower reduction in the attachment was achieved in romaine lettuce 'Bionda degli Ortolani' (−8%) and rocket (−17%) compared to the accessions showing lower roughness (wild rocket 'Yeti', wild lettuce and lamb's lettuce 'Trophy F1′; −47% on average) ( Figure 5). Swiss chard differed from 'Bionda degli Ortolani' and wild rocket 'Yeti'.

Discussion
Intraspecific variability in the susceptibility to human pathogen contaminatio been observed by many authors in several vegetable crops, among which lettuce w of the most studied [34]. In particular, Jacob and Melotto [23] found differences attachment and persistence of S. enterica and E. coli between nine cultivars of this sp Our study confirmed differences in E. coli attachment within lettuces, even be accessions belonging to the same type (Table 2). Interestingly, 'Bionda degli Ortolan 'Maraichere', whose susceptibility resulted to be different, are both romaine let Romaine lettuce has been involved in many food disease outbreaks in the U.S. a particular, in E. coli infections [35][36][37][38]. These alerts are generically referred to this l type, considered as one of the riskiest leafy vegetable in terms of foodborne dis However, since our data showed that romaine lettuce contamination varies by cult reduction in the risks associated with this lettuce could be achieved through v

Discussion
Intraspecific variability in the susceptibility to human pathogen contamination has been observed by many authors in several vegetable crops, among which lettuce was one of the most studied [34]. In particular, Jacob and Melotto [23] found differences in the attachment and persistence of S. enterica and E. coli between nine cultivars of this species. Our study confirmed differences in E. coli attachment within lettuces, even between accessions belonging to the same type (Table 2). Interestingly, 'Bionda degli Ortolani' and 'Maraichere', whose susceptibility resulted to be different, are both romaine lettuces. Romaine lettuce has been involved in many food disease outbreaks in the U.S. and, in particular, in E. coli infections [35][36][37][38]. These alerts are generically referred to this lettuce type, considered as one of the riskiest leafy vegetable in terms of foodborne diseases. However, since our data showed that romaine lettuce contamination varies by cultivar, a reduction in the risks associated with this lettuce could be achieved through varietal choice. Also, for Diplotaxis tenuifolia, we observed intraspecific variability ( Table 2) which, to our knowledge, had not yet been found in this species. No differences in E. coli ATCC 35218 attachment were detected within the Taraxacum campylodes, Brassica juncea, Brassica rapa, Cichorium intybus and Beta vulgaris accessions. Conversely, the same accessions of C. intybus used in this study but compared at the microgreen stage (immature greens harvested when cotyledons the first pair of true leaves are more or less developed) showed a different susceptibility to S. enterica and E. coli contamination [39]. Actually, the stage of the leaf may affect the contamination, as noticed by Hunter [40].
It is known that the leaf micro-morphological traits influence the attachment and persistence of human pathogens on vegetables. Many studies have shown that stomata are involved in the contamination of leafy greens with S. enterica and E. coli [11,23,41,42]. Bacteria were found on guard cells, within and underneath the stomata cavity, and on the crevices in proximity to them [19]. In our study, the six selected baby leaf accessions differed in stomata (Table 3). Nevertheless, the Pearson's test revealed that the bacterial retention (1.5 h after inoculation) and stomatal parameters were not correlated. This finding agrees with the results reported by Jacob and Melotto [23], who did not find a correlation between stomata aperture width, pore area, and density and E. coli or S. Typhimurium attachment onto leaves of different lettuce genotypes. On the other hand, the same authors found that the bacterial persistence 10 days after surface inoculation was correlated with the stomatal aperture width and pore area [23]. Also, Macarisin [11] observed that stomata density affected the persistence of E. coli O157:H7 in different spinach cultivars. Therefore, as suggested by Jacob and Melotto [23], for lettuce, stomata seem to have a role in bacterial penetration into the leaves, while the attachment might be influenced by other properties of the leaf surface. Traits such as trichomes, hydrophobicity, vein areas, cuticular waxes, surface irregularities and surface proteins and sugars can be associated with different attachment levels [40,43,44]. Palma-Salgado et al. [24] found that leaf surface roughness was positively correlated to E. coli adhesion in different leafy greens. Also, in spinach, roughness was an important factor determining the differential attachment and persistence of E. coli on leaves [11]. Our results support these previous findings. In fact, a significant correlation between the contamination level and roughness parameters (Ra and Rq values) was observed (Figure 4).
Leaf roughness can also reduce the effectiveness of sanitizing treatments for vegetables. Doan et al. [14] and Yi et al. [45] reported that the variation in leaf topography accounted for differences in survival of E. coli in chlorine-treated leafy greens of different species and cultivars. Surface roughness revealed to be a factor protecting E. coli cells from treatment with chlorine also in fruits and seeds [46,47]. UV radiation can be used as an alternative to chemicals for the sanitization of food products [48]. Different studies demonstrated the effectiveness of UV rays to control foodborne pathogens in fresh-cut vegetables, including lettuce [49,50]. In general, the role of surface topography on the effectiveness of UV treatment for bacterial inactivation was demonstrated by Woodling and Moraro in stainlesssteel surfaces [51]. In our study, UV treatment resulted in less E. coli ATCC 35218 attachment reduction in rougher accessions than in those showing less roughness ( Figure 5). Therefore, our results suggest that leaf roughness may offer protective niches to bacteria from UV rays.
Finally, the positive correlation between E. coli ATCC 35218 attachment and leaf water content that we found could suggest that the internal water status may impact the bacterial contamination of the leaves. Interestingly, in Yadav et al. [25], leaf water content was positively correlated with bacterial colonization of the phyllosphere in eight Mediterranean plant species. S. enterica colonization was enhanced by water-congestion in green tomato fruits [52]. The mechanism by which tissue water amount influences bacterial contamination in plants could be related to physical and/or chemical factors associated to tissue turgidity.

Conclusions
The screening of 30 baby leaf accessions for the susceptibility to E. coli ATCC 35218 contamination confirmed that variability exists among different leafy crop species, types of the same species and cultivars. The differences within the same species suggest that the choice of the cultivar can be a means to reduce the risk of foodborne diseases linked to the consumption of salads. Even within var. longifolia of lettuce (romaine type), which is the most involved in food disease outbreaks in the U.S., and for this reason is considered one of the riskiest salads, it seems possible to find genotypes (e.g., cv. 'Maraichere') less susceptible to contamination than others (e.g., cv. 'Bionda degli Ortolani'). These findings could find a practical application in orienting growers and the vegetable industry towards safer crops.
Our study also showed that leaf roughness and water content impact on E. coli ATCC 35218 attachment in baby greens. The use of an innovative portable 3D digital microscope highlighted the differences in the leaf roughness of six selected accessions. Among the measured leaf micro-morphological traits, roughness was the only one to be positively correlated with the E. coli ATCC 35218 contamination level. Roughness also somehow offered UV protection to bacteria. In fact, when E. coli ATCC 35218 retention in surfaceinoculated leaves was measured after UV treatment, bacterial contamination was reduced to a minor extent in the rougher leaves. This finding further underlies the impact of leaf roughness on the safety of vegetables, suggesting that sanitization treatment through UV rays should be modulated accounting for this parameter. The correlation between E. coli ATCC 35218 attachment and leaf water content was also significant, suggesting that the internal water status can impact the bacterial contamination of the leaves.
Looking forward, new findings on leaf characteristics associated with lower susceptibility to attachment and proliferation of human pathogens could be used in selection breeding in order to obtain safer cultivars and thus increase food safety in the vegetable industry. In particular, our results identified roughness as a trait to be considered in breeding programs.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/biology12010102/s1, Table S1: Number of leaves (NL), leaf fresh weight (FW) and dry weight (DW), leaf area (LA); and leaf water content (LWC) of the 30 baby leaf accessions (average ± SD); Table S2: Colour (a*, b*, L*) of the 30 baby leaf accessions (average ± SD); Video S1: 3D reconstruction of the surface under inspection obtained elaborating a sequence of defocused images implementing a suitable algorithm based on shape from the focusing technique.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.