Antisense down-regulation of the strawberry β-galactosidase gene FaβGal4 increases cell wall galactose levels and reduces fruit softening

Highlight We identified a strawberry β-galactosidase gene involved in fruit ripening. Its silencing increases cell wall galactose levels and fruit firmness, suggesting galactose metabolism has a key role in strawberry softening.

Introduction the short postharvest shelf life of this fruit and limiting its storage and postharvest transport. During ripening, fruit cell walls are extensively modified, and these changes are the most important factor leading to fruit softening. In general, cell wall modifications that accompany softening involve the solubilization of pectin polymers, the depolymerization of pectins and matrix glycans, and the loss of neutral sugars from pectin side chains (Brummell, 2006;Goulao and Oliveira, 2008;Mercado et al., 2011). All these processes often occur concurrently during fruit ripening, although the extension of these modifications greatly depends on the kind of fruit (Brummell, 2006;Mercado et al., 2011). Strawberry softening is characterized by a moderate increase in pectin solubilization-i.e. an increase in the amount of pectins loosely bound to the cell wall-and depolymerization (Posé et al., 2011). Functional studies of genes encoding pectinase enzymes, such as polygalacturonase  or pectate lyase (Jiménez-Bermúdez et al., 2002;Youssef et al., 2009;Youssef et al., 2013), support a key role of pectin disassembly in strawberry softening. Transgenic fruits with low expression levels of these genes were significantly firmer than control fruits and displayed a reduction in solubilization and depolymerization of polyuronides (Santiago-Doménech et al., 2008;Posé et al., 2013;Posé et al., 2015). The molecular mechanism underlying pectin disassembly is unclear (Paniagua et al., 2014). Many fruits, including strawberry, show a loss of neutral sugars, mainly arabinose and galactose, during ripening (Gross and Sams, 1984;Redgwell et al., 1997). The removal of these carbohydrates from rhamnogalacturonan I (RG-I) polyuronides by β-galactosidases (β-Gal) and α-arabinofuranosidases has been suggested as one of the possible reasons for the increase of soluble pectins (Koh and Melton, 2002;Brummell, 2006). β-galactosidases (EC 3.2.1.23) are glycosyl hydrolases characterized by their ability to hydrolyse terminal, non-reducing β-D-galactosyl residues from numerous β-D-galactoside substrates (Pérez-Almeida and Carpita, 2006;Tateishi, 2008). In most fruits, β-galactosidases are encoded by small gene families with different patterns of expression that could play distinct roles in fruit development (Smith and Gross, 2000;Mwaniki et al., 2005;Tateishi et al., 2007;Othman et al. 2011). Functional analyses of β-galactosidases genes using transgenic plants have only been carried out in the climacteric model fruit tomato (Solanum lycopersicum) with contrasting results. Antisense TBG4 tomato fruits displayed reduced TBG4 mRNA levels and free cell wall galactose only at the onset of ripening, but softening of ripe fruits decreased by 40% (Smith et al., 2002). By contrast, neither the silencing of TBG1 (de Silva and Verhoeyen, 1998) nor TBG3 (Carey et al., 2001) modified tomato fruit firmness. In TBG6 antisense tomato plants, fruit firmness was also similar to that of control plants, but these fruits showed structural alterations in the cuticle that increased fruit cracking (Moctezuma et al., 2003).
In strawberry (Fragaria × ananassa), a non-climacteric fruit, β-galactosidase activity increases during fruit development and remains high in ripe fruit (Trainotti et al., 2001;Figueroa et al., 2010). At the molecular level, Trainotti et al. (2001) isolated three full-length cDNAs encoding β-galactosidase genes FaβGal1 to FaβGal3. Although all of them could be detected in fruit and vegetative tissues, only FaβGal1 showed an expression pattern that could be related to the fruit ripening process, with the other two genes expressed mainly in green, immature fruits. Transcriptomic studies performed in our research group have identified a large group of genes whose expression increases throughout strawberry fruit ripening. One of these genes, FaβGal4 (accession number KR189030), displays significant sequence homology with putative β-galactosidase from higher plants. The main goal of this study was the functional characterization of this gene. For this purpose, transgenic strawberry plants carrying an antisense sequence of FaβGal4 were generated and the effects of FaβGal4 down-regulation in fruit firmness and cell wall structure were analysed.
Cloning and sequence analysis of full-length cDNA of FaβGal4 The full-length cDNA corresponding to the FaβGal4 gene was isolated from a Fragaria × ananassa R stage fruit cDNA library (Medina-Escobar et al., 1997). The deduced amino acid sequence and the phylogenetic tree construction were performed using the Lasergene software package (DNASTAR).

Auxin and nordihydroguaiaretic acid treatments
Achenes of two sets of 50 middle-sized green fruits (G2) each, still attached to the plant, were carefully removed using the tip of a scalpel blade. One set of deachened G2 fruits was treated with the synthetic auxin 1-naphthalenacetic acid (NAA) in lanolin paste (1 mL) with 1 mM NAA in 1% (w/v) DMSO. The other set of G2 deachened fruits (control group) were treated with the same paste, but without NAA. Both treatments were applied over the whole fruit surface. All fruits were harvested 5 days after treatment, immediately frozen in liquid nitrogen, and stored at −80°C. During the course of the assays, the fruits reached the G3-W developmental stages.
Nordihydroguaiaretic acid (NDGA) is an ideal inhibitor of the 9-cis-epoxycarotenoid dioxygenase enzyme and was used to block abscisic acid (ABA) biosynthesis (Creelman et al., 1992). The lowest NDGA concentration that completely blocks ABA accumulation is 100 μM, as determined by preliminary tests in tomato fruit (Zhang et al., 2009). Strawberry fruits (Fragaria × ananassa, cv. Elsanta) were used at the mature G-W stages for the purposes of this study. All fruits (ten fruits per treatment) were carefully injected with 1-2 mL of NDGA (100μM) sterile solution or sterile water (control fruits) using a hypodermic syringe. Three replications were conducted for each treatment. The samples were harvested after 8 days of treatment, when the fruits reached the R developmental stage, then frozen in liquid nitrogen and stored at −80°C. Fruits treated with NDGA were white, whereas control fruits were red. These samples were used to determine relative expression of FaβGal4.

RNA isolation
Total RNA was isolated from independent pools of strawberry fruits at different growth and ripening stages and from vegetative tissues, in accordance with Asif et al. (2000). Achenes were always removed from fruit and only receptacle RNA was extracted and purified. RNA obtained was treated with RNase-free DNase I (Invitrogen) and purified through the RNeasy Mini kit (Qiagen). RNA concentration and purity were evaluated using a NanodropTM spectrophotometer ND-1000 (Thermo Scientific) and by 1% agarose gel electrophoresis.

Expression analysis by quantitative real-time PCR
Gene expression analysis of FaβGal4 was performed by quantitative real-time (qRT)-PCR through an iCycler (BioRad) device, as previously described (Benítez-Burraco et al., 2003). FaβGal4 gene primer sequences for quantitative amplification were 5′-CAG CCA CCC ACT CCT CTA TAA CCA GTT -3′ and 5′-GCG AAG CAG TAA AAT ACG AAG CAA AGC-3′. Each reaction was performed at least in triplicate and the corresponding cycle threshold (C t ) values were normalized using the C t value corresponding to an interspacer 26S-18S strawberry RNA gene (housekeeping gene) (Benítez-Burraco et al., 2003;Molina-Hidalgo et al., 2013). All of these values were subsequently used to determine the relative increase or decrease in FaβGal4 expression in the samples in comparison to that of the control gene in accordance with Pedersen (2001). Interspacer 26S-18S (primers: 5′-ACC GTT GAT TCG CAC AAT TGG TCA TCG-3′ and 5′-TAC TGC GGG TCG GCA ATC GGA CG-3′) was selected as the control gene owing to its constitutive expression throughout all of the different experimental conditions tested. The efficiency of each particular qRT-PCR and the melting curves of the products were also analysed to ensure the existence of a single amplification peak corresponding to a unique molecular species. The expression levels of the different β-galactosidase genes in control and transgenic antisense FaβGal4 fruits were measured by qRT-PCR as described above, using the following primers: Binary vector for antisense FaβGal4 silencing and generation of transgenic strawberry plants A 300-bp non-conserved region of the FaβGal4 gene in antisense orientation was cloned into the pK7WG2 binary vector using Gateway technology (Invitrogen, Darmstadt, Germany) for antisense silencing of FaβGal4. The forward primer 5′-AGA GGA GAT GCT CGG TCT CGG TAT C-3′ and reverse primer 5′-TGG CAT AGC GCT TAA ATA GTT CAT TCA GTT-3′ were used. The resulting fragment was cloned into pCR8/GW/TOPO (Invitrogen) and then transferred to the Gateway pK7WG2 vector by way of a specific recombination of both plasmids using LR clonase (Invitrogen). The resulting plasmid (pK7WG2-FaβGal) was tested through sequencing and restriction analyses prior to strawberry plant transformation. The plasmid was introduced into Agrobacterium tumefaciens strain AGL1 by electroporation.
Leaf discs of strawberry plants, cv. Chandler, micropropagated in vitro were used as explants for Agrobacterium-mediated transformation experiments, as described in Barceló et al. (1998). Explants were inoculated with a diluted culture of A. tumefaciens and selected in 25 mg L −1 kanamycin. After 7-8 months of selection, kanamycin-resistant shoots were acclimated and transferred to the greenhouse. Transgenic mother plants were propagated by runners and the daughter plants were used for phenotypic analysis. The presence of the transgenes in these putative transgenic lines was confirmed by PCR amplification of a 220-bp fragment belonging to the nptII gene.

Phenotypic analysis of transgenic plants
Transgenic plants were evaluated during three consecutive growing seasons, using non-transformed plants, cv. Chandler, as control. Plants were grown in a greenhouse under natural temperature and light conditions and fruit was collected from March to July. During the first year, nine independent antisense FaβGal4 lines were analysed. Eight plants per line and a minimum of 10 ripe fruits per line were evaluated. During the second and third year, two selected lines showing higher fruit firmness than control were evaluated. Thirty plants per line and 50-100 ripe fruits per line were analysed in both years. Fruits were harvested at the stage of full ripeness, when the fruit surface was completely red, and the weight, size, colour, soluble solids, and firmness were recorded. Colour was measured using a colorimeter (Minolta Chroma Metre CR-400, Osaka, Japan). The instrument was calibrated with a standard white and a standard black reflective plate before use. The L* a* b* colour space parameters (lightness, redness, yellowness) were recorded. Soluble solids were measured using a refractometer (Atago N1), and firmness using a hand penetrometer (Effegi) with a cylindrical needle of 9.62 mm 2 area.
β-galactosidase activity was measured in ripe fruits according to Figueroa et al. (2010). Protein extraction was carried out by grinding 10 g of strawberry fruits under liquid nitrogen into a fine powder. The powder was homogenized using an Ultra-Turrax (Janke & Kunkel) with 20 mL of cold extraction buffer [0.05 M Na-acetate buffer, pH 6, 1% (w/v) polyvinylpolypyrrolidone (PVPP)] containing 1 M NaCl. Homogenates were stirred at 4°C for 3 h and then centrifuged at 11 000 g for 30 min. The supernatant was used to determine β-galactosidase activity by using p-nitrophenyl-β-D-galactopyranoside (Sigma) as the substrate. To measure exogalactanase activity, fruit tissue was extracted as described by Trainotti et al. (2001). Briefly, the powder was homogenized with cold extraction buffer, centrifuged at 11 000 g for 15 min, and the pellet washed with 10 ml of extraction buffer without PVPP. After centrifugation, the pellet was resuspended in 0.05 M Na-acetate buffer, pH 6, containing 1 M NaCl and stirred for 5 h at 4°C. After extraction, the insoluble material was removed by centrifugation and the supernatant was dialyzed against 0.05 M Na-acetate buffer, pH 4.5. A lupin galactan pretreated with α-L-arabinofuranosidase (Megazyme) was used as the substrate for exo-galactanase activity as described by Carey et al. (1995). The galactosidase released in the assay was estimated by measuring the release of reducing sugars using 2-cyanoacetamide (Gross, 1982). Protein extractions were performed in triplicate.

Cell wall analysis
The cell walls were extracted from frozen ripe fruits following the method of Redgwell et al. (1992) with some modifications as previously described by Santiago-Doménech et al. (2008). Briefly, 20 g of fruit were homogenized in a 40 ml mix of phenol, acetic acid, and water (PAW, 2:1:1, w:v:v). The homogenate was centrifuged at 4000 g for 15 min and the supernatant filtered through Miracloth (PAW fraction). The pellet was treated overnight with 90% DMSO to solubilize the starch. The extract was then centrifuged at 4000 g and the pellet washed twice with distilled water. The water fraction was discarded and the pellet, the cell wall material (CWM), was lyophilized and weighed. After extraction, the CWM was sequentially fractionated following the procedure of Santiago-Doménech et al. (2008). Briefly, 300 mg of the CWM was sequentially extracted overnight with deionized water, 0.05 M trans-1,2-diaminocyclohexane-N,N,N′N′-tetraacetic acid (CDTA) in 0.05 M acetate buffer (pH 6), 0.1 M Na 2 CO 3 containing 0.02 M NaBH 4 , 1 M KOH containing 0.02 M NaBH 4 , and 4 M KOH containing 0.02 M NaBH 4 . All fractions were extensively dialyzed against pure water. Then, the extracts were concentrated with a rotary evaporator and freezedried. Three independent fractionations per CWM sample were performed.
The uronic acid (UA) content was measured using the carbazole method (Filisetti-Cozzi and Carpita, 1991) using galacturonic acid as the standard. For neutral sugar analysis, the CWM and the different cell wall fractions were hydrolysed with 72% (w/w) sulphuric acid. After neutralization with ammonia, carbohydrates were derivatized to alditol acetates and analysed by gas chromatography with flame ionization detection (Blakeney et al., 1983). Gas chromatography analysis was carried out in triplicate.
Infrared spectra were recorded on a Jasco FT/IR-4100 (Spain) spectrometer coupled to an attenuated total reflectance (ATR) accessory (MIRacle ATR, PIKE Technologies, USA). Spectra Manager v.2 software (Jasco, Spain) was used to correct for both ATR effect and atmospheric contributions from carbon dioxide and water vapour across the full spectral range.
Statistical analysis SPSS software was used for all the statistical analyses. Levene's test for homogeneity of variances was performed prior to ANOVA, and Dunnett's multiple comparison test or Tamhane's T2 test was used for mean separation. A Student's t test or Mann-Whitney U test was used for pair comparisons.

Sequence analysis of FaβGal4 gene and protein
A putative novel β-galactosidase gene, FaβGal4, was identified in the transcriptomic analysis of green versus red-ripened strawberry receptacles using a 35k microarray oligonucleotide-based platform. A BlastN search showed 99% similarity between the cDNA sequence of FaβGal4 and Fragaria vesca β-galactosidase 16 mRNA (XM_004292664.2). The FaβGal4 gene encodes a polypeptide of 909 amino acids that contains a glycosyl hydrolase family 35 domain and a galactose lectin binding domain. Phylogenetic tree analysis of the FaβGal4 protein with other β-galactosidase fruit proteins from higher plants revealed the highest sequence identity with the FaβGal2 protein from strawberry (84.5%) and low identity with TomβGal1 (45.7%) and TomβGal3 (47%), both from tomato, PaβGal (44.3%) from avocado (Persea americana), and MdβGal (47.2%) from apple (Malus domestica) (Fig. 1).

Gene expression studies
Transcript analyses in fruits of the cultivated Fragaria × ananassa, cv. Camarosa, showed low expression of FaβGal4 in the early stages of fruit development. There was a progressive increase in expression during the ripening process, which starts at the white stage of receptacle ripening, and the highest level of expression was reached in red fruit receptacles ( Fig. 2A). An expression analysis of the other β-galactosidase genes previously described in strawberry, FaβGal1-3, showed that only FaβGal4 was clearly related to the ripening process (see Supplementary Fig. 1 at JXB online).
FaβGal4 transcript levels were very low in both achenes, independent of fruit ripening stage (Fig. 2B), and vegetative tissues ( Fig. 2C) but increased dramatically in the ripe fruit receptacle, indicating the putative specificity of this gene in the fruit receptacle.

Hormonal regulation of FaβGal4
It has been proposed that the ABA to auxin content ratio in strawberry fruit receptacles constitutes a signal that triggers the fruit ripening process (Perkins-Veazie, 1995). Thus, we investigated whether FaβGal4 expression was under the control of these two hormones. Because it is known that auxins present in strawberry receptacles are synthesized by the achenes, a comparative gene expression analysis was carried out between control and de-achened green fruits at the G2 stage that were or were not externally treated with NAA. A clear increase in the amount of FaβGal4 transcript was detected in de-achened fruits (Fig. 3A). This increase did not occur, however, when de-achened fruits were treated with a lanolin paste containing NAA. These results clearly indicate that the expression of this gene was negatively regulated by auxins.
In addition, when the receptacle ABA content was depleted by adding NDGA, a significant reduction in the amount of FaβGal4 and other cell wall enzyme transcripts, e.g. FaRGLyase1, was observed when compared against control fruits (Fig. 3B). This indicates that FaβGal4 expression could be activated by ABA.

Phenotypic analysis of antisense FaβGal4 plants
Nine independent kanamycin-resistant shoots carrying the antisense sequence of the FaβGal4 gene were recovered, yielding an average transformation rate of 2%. The transgenic nature of the plants was confirmed by PCR amplification of a 220-bp fragment belonging to the nptII gene (results not shown). Transgenic FaβGal4 plants showed a vegetative growth pattern similar to that of the control plants. However, during the first year of analysis, fruit yield was reduced by 60-80% in most of the lines compared with the control line. Furthermore, the proportion of malformed and small fruits in these lines was higher than in control. To estimate fruit quality, control and transgenic fruits were harvested at the stage of full ripening, and the weight, length, width, colour, soluble solids, and firmness were recorded. Results obtained . Quantification was based on Ct values as described in Materials and Methods. The increase in the mRNA value was relative to the G1-Ct value, which was assigned an arbitrary value equal to unity. Mean values ± SD of five independent experiments are shown. In all figures, statistical significance with respect to the reference sample (G1 achenes in B and G1 fruit receptacle in A and C) was determined by the Student's t-test. **P-value ˂ 0.01 and ***P-value ˂ 0.001.
during the first year of analysis are summarized in Table 1. Mean fruit weight and size were significantly lower in most transgenic lines when compared with control fruits, with the exception of lines β-Gal21 and β-Gal27. Interestingly, fruit colour, estimated by the L* a* b* colour space parameters, was significantly different in most transgenic lines. Transgenic fruits showed an increase on L* (lightness) and b* (yellowness), while a* (redness) was significantly increased only in lines β-Gal28 and β-Gal37. Soluble solids increased significantly in three of the transgenic lines, β-Gal25, β-Gal28, and β-Gal37. With respect to fruit firmness, lines β-Gal28 and β-Gal37 yielded fruits significantly firmer than control, with an increase in firmness of 47% and 31%, respectively. Based on this observation, these two lines were selected for further studies.
Fruit quality parameters in selected transgenic lines were evaluated during two additional growing seasons, using each years' daughter plants derived from vegetative propagation by runners. Results obtained in the two growing seasons were similar to those observed during the first year of analysis (Table 2). Transgenic fruits were smaller than control and showed higher soluble solid contents, especially those from line β-Gal37. Colour was also slightly modified in both transgenic lines. With respect to fruit firmness, both transgenic lines displayed significantly higher firmness values than control in the two years of analysis (Table 2): the average increase of firmness was 28%. Fruit yield was significantly reduced in both transgenic lines, with mean values of 179.2 ± 31.2, 63.7 ± 10.9, and 102.2 ± 29.2 g of fruit per plant in control, non-transformed plants, and β-Gal28 and β-Gal37 transgenic lines, respectively. This decrease was due to a reduction in fruit weight but also a decrease in the number of fruits produced.

FaβGal4 gene expression and enzyme activity in transgenic fruits
The expression level of FaβGal4, as well as the other three β-galactosidase genes described in strawberry fruit (FaβGal1-3), was analysed in red fruits from control and selected β-Gal28 and β-Gal37 transgenic lines by qRT-PCR (Fig. 4). A significant down-regulation of FaβGal4 was observed in ripe fruits from the two transgenic lines, with a similar level of silencing in the transgenic genotypes, close to 70%. Despite the FaβGal4 sequence sharing a low similarity with the rest of the β-galactosidase genes analysed, both transgenic lines also showed a 70% decrease in the amount of FaβGal1 mRNA levels (Fig. 4). Furthermore, β-Gal28 transgenic fruits showed an additional silencing of FaβGal3. By contrast, mRNA levels of FaβGal2 were not modified in transgenic lines.
β-galactosidase enzymes are characterized by their ability to hydrolyse terminal, non-reducing β-galactosyl residues from numerous substrates. In this research, the term β-galactosidase activity refers to an enzyme that hydrolyses a β-galactosyl residue linked to a variety of aglycones, e.g. nitrophenyl-β-D-galactopyranoside (NPG), whereas exo-galactanase refers to an enzyme that is specific for the non-reducing end of galactan (Smith et al., 2002). Both activities were measured in triplicate in protein extracts from ripe fruits of the two anti-FaβGal4 selected lines. Total β-galactosidase activity was similar in control and transgenic fruits, 0.6 ± 0.1 µmol NPG·g FW −1 ·h −1 in control versus 0.7 ± 0.1 and 0.9 ± 0.2 µmol NPG·g FW −1 ·h −1 in fruits from β-Gal28 and β-Gal37, respectively. Exo-galactanase activity, measured against a lupin galactan, was low, and no differences between control and transgenic lines were observed (1.1 ± 0.5 µg·g FW -1 ·h -1 in control versus 0.8 ± 0.1 and 2.3 ± 1.3 µg·g FW −1 ·h −1 in fruits from β-Gal28 and β-Gal37, respectively). The increase in mRNA values after achene removal and auxin treatment was relative to G2 fruit (control), which was assigned an arbitrary value equal to unity. G2-Achenes: G2 fruit receptacle without achenes; G2-Achenes+NAA: G2 fruit receptacle without achenes treated with 1mM NAA; Control: full-sized green fruit receptacle (G2-Control). (B) Analysis by qRT-PCR of FaβGal4 and FaRGLyase1 expression in strawberry fruits in response to NDGA. NDGA treatment performed for 8 days to G3-stage fruits. Control: G3 fruits injected with water; FaRGLyase1: rhamnogalacturonate lyase 1 (CO381780.1); NDGA: G3 fruits injected with NDGA (100 µM). The expression levels of both genes studied in NDGA-treated fruits are expressed as a percentage against their expression levels in control untreated fruits. Statistical significance with respect to the control samples was determined by the Student's t-test. ***P-value ˂ 0.001 and **P-value < 0.01.

Cell wall analysis
The yield of CWM and soluble PAW fraction obtained per fresh weight of fruit was similar in control and transgenic lines, with average values of 0.93 g CWM·100 g FW −1 and 0.09 g PAW·100 g FW −1 . CWM was sequentially fractionated with water, CDTA, Na 2 CO 3 , and KOH (1 M and 4 M) to extract fractions enriched in water soluble pectins, ionically bound pectins, covalently bound pectins, and hemicellulosic polymers, as described by Santiago-Doménech et al. (2008). The yields of the different cell wall fractions obtained were similar to those described by Santiago-Doménech et al. (2008) and Posé et al. (2013) in fruits of the same cv., with no significant differences observed between control and transgenic fruits in any of the fractions obtained (data not shown).
UA content was measured in all fractions, and the results obtained are shown in Fig. 5. In both transgenic lines, the Na 2 CO 3 fraction was enriched in UA when compared with control. A slight increase in UA was also observed in the transgenic water fractions. The amount of UA was also higher in the CDTA fraction from line β-Gal37, but not β-Gal28. The analysis of neutral sugars by gas chromatography revealed significant differences between the transgenic β-Gal37 line and control CWM. Transgenic CWM showed a 31% increase in Gal and Ara content ( Table 3). The amounts of Rha and Fuc were also higher than control, although the differences were not statistically significant. By contrast, the amount of Xyl, Man, and Glc decreased 38.6 ± 2.5* 37.9 ± 3.3 23.1 ± 3.9* 8.9 ± 2.2 3.2 ± 0.7 β-Gal24 7.3 ± 2.1* 30.3 ± 5.6* 22.9 ± 3.4* 39.9 ± 6.3* 39.8 ± 5.5 26.0 ± 5.5* 9.6 ± 2.9 3.5 ± 0.6 β-Gal25 5.2 ± 2.0* 24.4 ± 4.3* 22.6 ± 3.1* 44.9 ± 5.8* 40.7 ± 4.7 28.1 ± 4.2* 13.8 ± 2.6* 3.5 ± 0.9 β-Gal27 11.9 ± 2.5 36.2 ± 4.7 26.9 ± 2.9 33.8 ± 2.7* 36.9 ± 3.2 18.1 ± 3.5* 7.5 ± 2.0* 3.4 ± 0.6 β-Gal28 9.7 ± 3.0* 28.  Fig. 4. Relative FaβGal4 expression, estimated by qRT-PCR, in ripe fruits from control and transgenic β-galactosidase lines. Bars represent mean ± SD of three independent RNA quantifications. Statistical significance with respect to the control line was determined by Dunnett's multiple comparison test. ***P-value ˂ 0.001 and **P-value < 0.01. slightly (Table 3). This general trend was observed in most cell wall fractions with some exceptions (Table 4). The control PAW fraction, which contained apoplastic free polymers solubilized by in vivo processes (Redgwell et al., 1992), displayed the highest proportion of Rha, but this carbohydrate was not detected in transgenic PAW. Additionally, this fraction, as well as that of 4 M KOH, showed a slight increase in Glc when compared with control. Polymers soluble in CDTA from transgenic fruits showed a slight decrease in Fuc, Ara, Xyl, and Glc content. Similarly, Na 2 CO 3 -soluble pectins displayed large decreases in Fuc, Xyl, and Man. Regarding Gal content, all cell wall fractions from transgenic fruits contained a higher proportion of this carbohydrate (Table 4), with the greatest increase in the water and, especially, the 1 M KOH fractions (see Supplementary Fig. 2 at JXB online). Interestingly, the Gal increase in this last fraction was concomitant with large decreases in Xyl and Man and a high increase in Rha (Table 4). The modification of the carbohydrate composition in the 1 M KOH fraction as a result of FaβGal4 silencing was confirmed by Fourier transform infrared spectroscopy (Fig. 6). Control fruits showed a spectral profile in the mid-infrared region at 1200-800 cm −1 , corresponding to the fingerprint region of carbohydrates, with peaks at 1014, 1026, 1045, 1077, and 1095 cm −1 . According to Kačuráková et al. (2000), homogalacturonan pectin profiles have maximum absorption bands at 1100 and 1017 cm −1 , RG-I shows the strongest peaks at about 1070 and 1043 cm −1 , and the main xyloglucan absorption band is at 1041 cm −1 . The 1 M KOH profile from control fruits could therefore be assigned to a mixture of RG-I and xyloglucans. The 1 M KOH profile from transgenic fruits was different, showing a main peak at 1038 cm −1 and a shoulder at 1070 cm −1 . According to Kačuráková et al. (2000), these bands could be assigned to Ara and Gal, respectively, although the absorption infrared bands of other neutral    sugars overlap in this region. This result suggests a change in the solubilization of RG-I present in the KOH fraction.

Discussion
A common feature of the cell wall disassembly process taking place during ripening of fleshy fruits is the loss of galactose from pectin side chains, supposedly due to the action of β-galactosidase enzymes (Smith and Gross, 2000;Brumell, 2006). In strawberry, Trainotti et al. (2001) identified three β-galactosidase genes. All of them were expressed in young expanding leaves, stolons, flowers, and green fruits, but only FaβGal1 showed increased expression during fruit ripening. Our comparative transcriptomic analysis between strawberry receptacles from immature (green) and ripened (red) strawberry fruits showed a sequence encoding a novel putative β-galactosidase gene, FaβGal4, whose expression was strongly up-regulated, suggesting that this gene could play a pivotal role in strawberry fruit cell wall degradation and fruit softening. Bioinformatic analyses showed a low identity between FaβGal4 full-length cDNA, or its orthologue in F. vesca, and the three β-galactosidase genes previously described in strawberry. At the protein level, the predicted FaβGal4 shares the highest identity with FaβGal2. The deduced protein of FaβGal4 contains the consensus sequence pattern of the putative active site characteristics of the glycosyl hydrolase family 35 (Henrissat et al., 1995), and, interestingly, it is identical to the putative active site for the other three β-galactosidase genes described in strawberry (Trainotti et al., 2001). These three genes showed β-galactosidase activity when expressed in vitro (Trainotti et al., 2001) and, therefore, it is likely that the FaβGal4 product also has β-galactosidase activity. However, enzymatic activity assays should be performed in order to definitively address this question.
FaβGal4 is mainly expressed in ripe fruit receptacles and is regulated by auxin and ABA The expression pattern of FaβGal4 during receptacle fruit growth and maturation was clearly ripening-related, with maximum gene expression in red ripe fruits. This expression pattern was similar to those previously reported for other strawberry genes encoding pectin-degrading enzymes, e.g. pectate lyases, polygalacturonases, or rhamnogalacturonate lyase (Medina-Escobar et al., 1997;Redondo-Nevado et al., 2001;Benitez-Burraco et al., 2003;Quesada et al., 2009;Molina-Hidalgo et al., 2013).
During the development of strawberry fruit, the level of auxins produced by achenes and released into the receptacle declines, which induces the ripening process (Perkins-Veazie, 1995). Several studies have shown that the removing the achenes from the fruit surface induces the expression of many strawberry fruit ripening-related genes that encode cell wall enzymes (Medina-Escobar et al., 1997;Redondo-Nevado et al., 2001;Benitez-Burraco et al., 2003). This was also the case for FaβGal4. More recently, Chai et al. (2011) and Jia et al. (2011) provided molecular evidence indicating that ABA is a signal molecule that, at least, can promote the strawberry ripening-related production of anthocyanins. In fact, ABA levels gradually increase concomitant with the ripening process in strawberry fruits Jia et al., 2011). When strawberry fruits were treated at the green stage with NDGA, an inhibitor that blocks the synthesis of ABA, FaβGal4 was significantly down-regulated, as has been observed for other cell wall enzymes (Molina-Hidalgo et al., 2013). Altogether, these results indicate a putative co-regulation of FaβGal4 with genes encoding other pectin-degrading enzymes, and support a physiological role related with the enzymatic degradation of the cell wall in ripe fruits leading to fruit softening.
Antisense down-regulation of FaβGal4 reduces strawberry fruit softening Transgenic strawberry plants carrying a 300-bp antisense sequence of FaβGal4 under the control of the constitutive promoter CaMV35S were generated to get insight into the role of this gene on fruit softening and cell wall disassembly. In general, fruit yield and fruit weight were reduced in most transgenic lines, as has been observed in other transgenic lines with pectinases down-regulated, e.g. pectate lyase transgenic plants (Jiménez-Bermúdez et al., 2002;Youssef et al., 2009). Average fruit size and weight generally decrease when micropropagated strawberry plants are used directly for fruit production, but this side effect of the in vitro tissue culture phase disappears when the progeny of micropropagated plants is evaluated (Cameron et al., 1989;López-Aranda et al., 1994). In FaβGal4 transgenic plants, the reduction of fruit weight was stably maintained after three rounds of runner propagation in the greenhouse, suggesting that this effect was due to FaβGal4 silencing rather than the in vitro micropropagation of plants. In strawberry, fruit size depends mainly on the number of achenes per fruit, the flower position, and the receptacle sensitivity to auxin (Perkins-Veazie, 1995). In the case of the two transgenic lines selected, a marked reduction of the average number of achenes per fruit was observed, with mean values of 240.1 in control plants, 105.6 in β-Gal28, and 69.5 in β-Gal37. However, the mean number of achenes per gram of fruit was only statistically different to the control in the case of β-Gal28 fruits. These results suggest that the decrease in fruit size was mainly due to deficient fertilization in transgenic plants as a result of FaβGal4 silencing. Early and late male gametophyte development genes with high homology to β-galactosidases have been identified in pollen from tobacco, Arabidopsis, and rice (Tursun et al., 2003;Hrubá et al., 2005). The products of these genes may participate in cell wall loosening during young pollen expansion after microspore mitosis and/or may be associated with pollen germination and pollen tube penetration through the style (Hrubá et al., 2005).
Two out of the nine transgenic lines obtained produced fruit significantly firmer than control at the ripe stage. The firmer fruit phenotype was stably maintained during three growing seasons. Transgenic fruits of selected lines were on average 30% firmer than control, an increase in firmness similar to that obtained when down-regulating a pectate lyase (Jiménez-Bermúdez et al., 2002) or a polygalacturonase gene . The increase in fruit firmness cannot be ascribed to the reduction in fresh weight, because no correlation between fruit weight and firmness was observed among the different transgenic lines obtained. Transcriptomic analysis of transgenic fruit from the selected lines showed a significant reduction, 70%, of FaβGal4 mRNA levels when compared with control fruit. Although FaβGal4 did not share significant sequence homology with the other three β-galactosidase genes previously described in strawberry, the constitutive expression of an antisense FaβGal4 sequence also reduced FaβGal1 mRNA amounts in the two selected lines, as well as FaβGal3 in fruits from β-Gal28. By contrast, FaβGal2 expression was not modified in transgenic fruits. These results indicate an intricate regulation of β-galactosidase genes during the ripening of strawberry fruit. Additionally, despite the high reduction in the expression of β-galactosidase genes, neither β-galactosidase activity nor exo-galactanase activity were reduced in transgenic ripe fruit. Previous studies in transgenic tomato showed a complex relationship between the expression of the different β-galactosidase genes involved in fruit ripening and β-galactosidase activity. Antisense TBG4 tomato plants showed a 40% reduction in fruit softening although TBG4 silencing was only observed at the mature green stage (Smith et al., 2002). The expression of TBG3 was also altered in those antisense tomato fruit, down-regulating or up-regulating its expression depending on the ripening stage. Moctezuma et al. (2003) found an altered expression of TBG4 and TBG5 in tomato fruits with TBG6 downregulated by antisense transformation. Contrary to the report of Smith et al. (2002), the suppression of TBG6 was observed during the whole ripening process. Although transgenic TBG6 fruits had no differences in softening, increased fruit cracking and cuticle thickness, and a reduced locular space were observed (Moctezuma et al., 2003). As regards β-galactosidase activity, neither Smith et al. (2002) nor Carey et al. (2001) found a decrease in total β-galactosidase activity in transgenic antisense TBG4 or TBG1 fruit, respectively. However, the silencing of TBG4 reduced exo-galactanase activity, measured against a lupin galactan, at the breaker plus 3 days developmental stage, returning to wild-type control levels at breaker plus 7 days stage (Smith et al., 2002). The scenario was more complex when modifying the expression levels of TBG6 (Moctezuma et al., 2003). In these antisense fruits, β-galactosidase activity was significantly higher than in control at early stages of development, returning to wild-type levels at the later stages. However, exo-galactanase activity was reduced at the breaker stage (Moctezuma et al., 2003). These results could be explained by the large number of β-galactosidase isoforms with different hydrolysing abilities against the native cell wall polysaccharides and synthetic products that are expressed in fruit (Tateishi, 2008). The in vitro expression of FaβGal4 needs to be investigated to clarify the actual substrate of this protein.
Antisense FaβGal4 down-regulation increases cell wall galactose levels and reduces pectin solubilization According to Redgwell et al. (1997), Gal levels decrease 18% in ripe strawberry fruit when compared with green immature fruit. The silencing of FaβGal4 increased Gal by 30% in transgenic ripe fruit, but also induced a significant increase in the Ara content. After cell wall fractionation, it was observed that all fractions contained higher amounts of Gal than control, with the largest increases present in the H 2 O, CDTA, and especially the 1 M KOH fractions. KOH-soluble fractions comprise mainly hemicellulosic polymers, but also contain pectins (Koh and Melton, 2002;Santiago-Doménech et al., 2008;Posé et al., 2013). Interestingly, Redgwell et al. (1997) found that the highest Gal loss, 82.3% of total cell wall Gal, occurred in the KOH fraction and the CWM residue, while only 17.7% was lost from the CDTA/Na 2 CO 3 polyuronide-enriched fractions. This trend was observed in most of the fruit analysed, including species where an increase of β-galactosidase activity during fruit ripening had previously been reported, e.g. tomato, kiwifruit, and avocado (Redgwell et al., 1997). The high Gal content in the 1 M KOH fraction from antisense FaβGal4 fruits suggests that Gal loss takes place mostly in arabinogalactan pectins loosely bound to xyloglucans by covalent cross-links and/or physically entangled within the matrix glycan. It is noteworthy that this is the first report of the modification of cell wall Gal levels in fruit by suppression of a β-galactosidase gene. Total cell wall Gal content was not modified in antisense TBG4 tomato fruit, although decreased free galactose levels were observed in transgenic fruit at mature green stage, returning to wild-type levels during ripening (Smith et al., 2002).
The relationships among loss of neutral sugars, pectin solubilization, and fruit firmness are far from clear. Some fruits such as apple and nashi pear showed a marked loss of Gal but pectin solubilization was slight (Redgwell et al., 1997). Conversely, plum did not show any Ara/Gal loss but showed extensive pectin depolymerization (Redgwell et al., 1997). In FaβGal4-silenced fruits, the increase of cell wall levels of Gal was paralleled by a reduction in polyuronide solubilization, reflected in the higher amount of Na 2 CO 3 -soluble pectins. Additionally, a preliminary analysis of these pectins by atomic force microscopy showed larger polymers in transgenic samples than in control samples, suggesting a reduced depolymerization during ripening (unpublished results). Previous studies in transgenic strawberry plants with pectate lyase or polygalacturonase genes silenced demonstrated that the increased firmness in these transgenic fruits was a result of reduced pectin solubilization and depolymerization (Santiago-Doménech et al., 2008;Posé et al., 2013;Posé et al., 2015). In both cases, cell walls from transgenic fruit displayed higher amounts of ionically and covalently bound pectins, those soluble in CDTA and Na 2 CO 3 , respectively, than wild type. Whether the reduction in fruit softening in transgenic FaβGal4 fruits is directly due to the increase of cell wall Gal content or by an indirect effect on pectin solubilization needs to be elucidated. The reduced loss of Gal from hairy regions of pectins could limit the access of other cell wall-modifying proteins, such as polygalacturonase or pectate lyase, to their substrate, therefore limiting pectin solubilization and, likely, depolymerization. As an alternative hypothesis, galactosyl residues could have a direct effect on the mechanical properties of the cell wall. In pea (Pisum sativum) cotyledons, the increased pectin galactan content at late developmental stages correlated with an increase in firmness (McCartney et al., 2000). The Arabidopsis mutant MUR3 lacks a xyloglucan-specific galactosyl transferase and its cell walls exhibit reduced xyloglucan galactosidation (Peña et al., 2004). This modification markedly reduces cell wall strength during hypocotyl growth, although this effect could be related to a modulation of xyloglucan endo-transglycosidase/hydrolase activity (Peña et al., 2004). The ectopic expression of a fungal endo-β-galactanase in potato tuber reduces RG-I linear β-1,4-galactan and also modifies the physical properties of the tissue, making it more brittle when subjected to compression (Ulvskov et al., 2005). A putative role of FaβGal4 on the release of galactosyl-containing oligosaccharides that trigger the ripening process also cannot be discarded. Gross (1985) found that mature green tomato fruit infiltrated with free Gal produced ethylene and ripened earlier than control fruits. More recently, Lahaye et al. (2013) found indirect evidence for the role of Gal on fruit ripening; they observed positive correlations between some firmness quantitative trait loci and cell wall Gal content, galactosylated pectins, and galactosylated xyloglucans in tomato. This hypothesis could explain the pleiotropic effect of FaβGal4 silencing on fruit colour and on the expression of the other β-galactosidase genes.

Conclusions
The results obtained support a key role for FaβGal4 in the softening of strawberry fruit. Transgenic fruits with FaβGal4 silenced showed a 30% increase in fruit firmness at the ripe stage. At the cell wall level, these transgenic fruits contained more Gal in all cell wall fractions than wild type, but also displayed lower polyuronide solubilization. Although the removal of Gal from cell walls is a general feature of the ripening process in many fruits, its relationship with pectin solubilization and fruit firmness is controversial. Our results shed light on these processes and clearly indicate a close connection between Gal levels and polyuronide solubilization. The reduction of Gal loss from galactosyl-containing side chains in the wall might result in decreased porosity, obstructing the access of other cell wall enzymes to their substrates, leading to reduced softening. Alternatively, the increased level of galactosyl residues could have a direct effect on the mechanical properties of the cell wall. This work also indicates that FaβGal4 is a good candidate for the biotechnological or conventional improvement of strawberry texture.

Supplementary material
Supplementary data are available at JXB online Supplementary Figure 1: Relative expression of the different β-galactosidase genes during fruit receptacle development, estimated by qRT-PCR. Bars represent mean ± SD of three independent RNA quantifications Supplementary Figure 2: Increase of Gal content, expressed as a percentage of control fruits, in CWM, PAW, and the different cell wall fractions isolated from transgenic β-Gal37 ripe fruits.