We used fruits from two Citrus sinensis cultivars: a mid-season orange cultivar (cv. Washington Navel) that usually undergoes pre-harvest abscission and a late-season orange cultivar (cv. Ricalate Navel) with delayed abscission. Maturing fruits were harvested after color change from adult trees grown in a homogeneous experimental orchard under normal cultural practices at the Institut Valencià d'Investigacions Agràries (IVIA). Fruits were separated from the tree leaving 2 cm peduncles to isolate the AZ-C for further analyses. For abscission kinetics studies and tissue collection, Washington Navel fruits were incubated for 0, 24, 48, and 96 h in the presence or absence of ethylene (10 μL/L) in sealed 10 l containers at 22°C with a 16 h light period under fluorescent lighting. Ricalate Navel fruits were incubated for 0, 24, 48, 96, and 192 h in the presence of 1-aminocyclopropane-l-carboxylic acid (ACC; 0.1 mM) or water under the same temperature and light conditions. In this case, a 3 mL Pasteur pipette containing the ACC solution or water was fitted to the fruit peduncles.

Phloroglucinol staining for lignin in fresh cut tissue portions (0.5 cm³) containing the AZ-C after 0, 24, and 48 h of ethylene or ACC treatment was performed according to Tadeo and Primo-Millo (1990). Samples were cut longitudinally to allow AZ-C staining and for further image acquisition. A saturated solution of phloroglucinol (Sigma-Aldrich) in 20% HCl was directly applied to samples. Observation was carried out with an Olympus SZ61 stereomicroscope (Olympus GmbH).

Longitudinal sections as well as the proximal (peduncle) and distal (fruit) fracture plane of the ethylene-promoted AZ-C were observed using cryo-SEM. To examine longitudinal sections of the AZ-C, 1 cm portions of tissue were manually dissected with a razor blade. In the second case, the peduncle was forcibly separated from the fruit. Specimen mounting and AZ-C observation were carried out as previously described in Agustí et al. (2009). At least three samples containing the AZ-C after 24, 48, and 96 h of ethylene treatment were observed.

Tissue containing the AZ-C after 0, 24, and 48 h of ACC treatment was manually dissected using a razor blade in 0.5 cm³ portions. These samples were fixed overnight at 4°C in a 4% (w/v) paraformaldehyde-PBS solution. After fixation, samples were washed with PBS, dehydrated in a graded ethanol series and embedded in LR White (Electron Microscopy Sciences). Longitudinal sections of the calyx button area (1 μm thickness) were cut with a Leica RM2165 microtome and placed on glass slides. Slides were further stained with PAS (Sigma-Aldrich) and mounted with DPX Mountant (Fluka). Observations were performed on a Leica DMLA microscope (Leica Microsystems) and images were processed with Leica ASMLD Version 4.0 software.

Portions of tissue containing the AZ-C (0.5 cm³) were dissected from fruits after 0, 12, and 24 h of ethylene treatment for the transcriptomics assay, and after 0, 12, 24, and 36 h of ACC treatment for lignin intermediates quantification. Preparation of cryosections and microdissection were performed as previously described in Agustí et al. (2009). Cells from the AZ-C and the adjacent FR were selected from 30 to 40 cryosections and collected separately.

Cryosections of 14 μm of tissue containing the AZ-C after 48 h of ethylene treatment were processed as described in Agustí et al. (2009) and mounted on CryoJane® adhesive coated slides (Instrumedics) following the manufacturer's instructions. Slides were stored at −80°C until phloroglucinol staining. Staining for lignin was performed using a saturated solution of phloroglucinol (Sigma-Aldrich) in 20% HCl. Observation was carried out with an Olympus SZ61 stereomicroscope (Olympus GmbH).

Three independent biological replicates were collected for each cell type at 0, 12, and 24 h after ethylene treatment. For each independent sample, total RNA from ~40,000 pooled cells was extracted using the RNeasy Micro Kit (Qiagen) following the manufacturer's instructions. The RNA purity was assessed by measurements of OD260/OD280. Two RNA amplification rounds were performed utilizing the TargetAmp™ 2-Round Aminoallyl-aRNA Amplification Kit (EPICENTER) according to the manufacturer's instructions to synthesize the antisense cRNA. The quality of the amplified RNA was evaluated by OD260/OD280 measurements and agarose gel electrophoresis. Each sample was labeled with Cy5 and co-hybridized with Cy3-labeled antisense cRNA from a reference sample containing a mixture of equal amounts of RNA from all experimental samples (0, 6, 12, 24, 48, and 96 h of ethylene/air treatment). RNA labeling, microarray hybridization, and slide washes were performed as previously described in Cercos et al. (2006). Hybridized microarrays scanning, hybridization data acquisition, and microarray normalization and analysis were carried out as described in Agustí et al. (2009). A cDNA microarray including 21.081 putative genes of citrus was utilized (Martinez-Godoy et al., 2008). Gene expression differences were considered significant under a P-value lower than 0.05 and an M contrast cutoff value of +0.5 or −0.5. In this work, a time course experiment was designed for each cell type (AZ-C and FR), therefore, the expression level corresponds to M = log₂ [AZ-C_t/AZ-C₀] or M = log₂ [FR_t/FR₀]. The raw microarray data as well as the protocols used to produce the data and the normalized data were deposited in the ArrayExpress database under the accession number E-MTAB-4538. Functional classification of the selected genes was performed using MIPS (Munich Information Center for Protein Sequences, http://www.helmholtzmuenchen.de/en/mips/) categorization. Amplified RNA was used for the validation of microarray hybridization data by semi-quantitative RT-PCR (sqRT-PCR) analysis (Figure S1).

sqRT-PCR analysis was carried out using the SuperScript II Reverse Transcriptase kit (Invitrogen, Carlsbad) following the manufacturer's instructions. After first-strand cDNA synthesis, PCR reactions were performed using the Biotools Taq DNA Polymerase (BIOTOOLS, B&M Labs). Size and intensity of expected bands were checked by 1% agarose gel electrophoresis. Citrus UBC gene (Ubiquitin-conjugating enzyme) was used as a reference to evaluate the amounts of mRNA in each sample. Primer sequences are available in Table S1.

RNA in situ hybridization with digoxigenin-labeled probes was performed as described (Gomez et al., 2011). Portions of tissue containing the AZ-C (0.5 cm³) were dissected from fruits after 24 h of ethylene treatment and immediately fixed at 4°C overnight in FAE (25% formaldehyde, 5% acetic acid, 50% ethanol), dehydrated, embedded in paraffin wax and sectioned to 8 μm. For CitCEL6 and CitPG20, RNA antisense and sense probes were generated with SP6 and T7 RNA polymerases, using as substrate a 1518 bp fragment of the CitCEL6 cDNA (1–1518 from ATG) or a 1110 bp fragment of the CitPG20 cDNA (217–1326 from ATG), amplified by PCR and cloned into the pGEM-T Easy vector (Promega).

Coumaric acid, caffeic acid, and ferulic acid were analyzed by UPLC coupled to tandem mass spectrometry (UPLC-MS/MS) as described by Argamasilla et al. (2014). Three independent samples containing ~40,000 pooled AZ-C cells were isolated by LM for each time point of ACC treatment (0, 12, 24, and 36 h) and collected in 60 μL of water.

Members of the different gene families associated with cell wall remodeling in citrus (based on CAZy classification; Carbohydrate-Active Enzymes; Cantarel et al., 2009; http://www.cazy.org/) were identified by TBLASTN search in the Citrus clementina haploid genome (version 0.9) database web browser (http://www.phytozome.net/search.php) using the consensus sequence for the catalytic domain of each family. Prediction of characteristic domains and conserved motifs was carried out through SMART (http://smart.embl-heidelberg.de/; Schultz et al., 1998; Letunic et al., 2009), PSORT (http://psort.hgc.jp/form.html), InterProScan (http://www.ebi.ac.uk/Tools/pfa/iprscan/) and big-PI Plant Predictor (http://mendel.imp.ac.at/gpi/plant_server.html) servers. Phylogenetic trees are based on multiple alignments using the profile alignment function of ClustalW (www.ch.embnet.org/software/ClustalW-XXL.htm) and were generated with MEGA7 (Kumar et al., 2016) using the neighbor-joining algorithm with 1000 bootstrap replicates. Poisson correction for multiple substitutions was used and only values higher than 50% were considered.

The primary monoclonal antibodies (mAbs) used in this study and provided by Prof. Paul Knox (Centre for Plant Sciences, Faculty of Biological Sciences, University of Leeds, UK) were LM5 [anti-(1,4)-β-D-galactan; Jones et al., 1997], LM6 [anti-(1,5)-α-L-arabinan; Willats et al., 1998] and JIM5 [anti-partially methylesterified/de-esterified homogalacturonan; (Knox et al., 1990)]. The secondary antibody was fluorescein isothiocyanate (FITC)-conjugated anti-rat IgG (Sigma-Aldrich). Immunolocalization of pectic polysaccharides was performed on semi-thin sections of tissue containing the AZ-C (0.5 cm³) from fruits after 0, 24, and 48 h of ACC treatment. Immunolocalization of pectic polyssacharides, light microscopy and image acquisition were perfomed as described in Coimbra et al. (2007).

We performed a kinetics assay of citrus fruit abscission in response to abscission-accelerating treatments to determine the optimal sampling for the transcriptomic analysis. To that end, we carried out a comparison between orange (C. sinensis) fruits incubated with ethylene or its immediate metabolic precursor 1-aminocyclopropane-1-carboxylic acid (ACC) and fruits incubated with air or water (controls). We used maturing fruits from a mid-season orange cultivar (cv. Washington Navel) that usually undergoes pre-harvest abscission and from a late-season orange cultivar (cv. Ricalate Navel) with delayed abscission. We observed a faster decrease of fruit detachment force (FDF) in both Washington Navel and Ricalate Navel fruits treated with ethylene/ACC in comparison to air-/water-treated control fruits (Figure 1A). At 48 h after treatment, the FDF in fruits treated with ethylene or ACC was around 4 kgf. However, control fruits of Washington Navel and Ricalate Navel only showed values of FDF around 4 kgf at 96 and 192 h, respectively, a response that matches their pre-harvest abscission behavior. Thus, ethylene and ACC accelerated the abscission process in both varieties tested. This result strongly suggests that the natural delay in the schedule of FDF decline in fruits of the late-season variety Ricalate Navel in comparison with those of the mid-season variety Washington Navel was not related to any impairment in the response of well-developed tissues to ethylene (Zacarias et al., 1993). Based on these findings, we used induced AZ-C samples from both varieties for further analyses.

Phloroglucinol staining in receptacles of both Washington Navel and Ricalate Navel fruits revealed lignin deposition at the central core of the AZ-C, between the axial vascular bundles, 24 h after ethylene and ACC treatments (Figures 1B,C). Forty-eight hours after the treatments, lignin deposition spread out along the AZ-C, perfectly drawing the separation line between the calyx button and the FR. Accordingly, timing for lignin deposition positively correlated with abscission kinetics.

We used scanning electron microscopy (SEM) to examine changes in the cellular morphology of the AZ-C from Washington Navel fruits treated with ethylene (Figure 2). We observed the first cellular signs of activation of abscission by ethylene in the central core of the AZ-C at 48 h of treatment (Figures 2A,B). At that time, AZ-C samples could be split into two groups, one showing early stages of cell separation and the other one showing late events of cell separation. In the former group, the AZ-C was clearly distinguishable (Figure 2A), with accumulation of an amorphous material probably derived from the partial dissolution of the middle lamella and cell wall of the AZ-C cells. In the latter group of samples, cell separation was observed in the central core of the AZ-C (Figure 2B). A greater accumulation of amorphous material was observed, suggesting that cell wall and middle lamella degradation was complete after 48 h at the central region. At 96 h after ethylene treatment, cell separation extended from the central core to the periphery of the AZ-C (Figure 2D) and differential cell expansion was observed at proximal (calyx button) and distal (fruit) sides (Figures 2C,E). At the proximal side, parenchymatic pith cells underwent expansion (Figure 2C) while, at the distal side, expansion occurred in the cells of the axial vascular bundles (Figure 2E).

Periodic acid-Schiff (PAS) staining was used to characterize anatomically the AZ-C after ACC treatment (Figure 3). This method detects insoluble carbohydrates and was used to distinguish the cells belonging to the AZ-C since these accumulate starch grains (Wilson and Hendershott, 1968; Huberman et al., 1983; Shiraishi and Yanagisawa, 1988; Goren, 1993). In addition to the starch-rich cell area (SA) previously identified by Wilson and Hendershott (1968) at the distal side of the AZ-C (FR side), we identified another cell area located at the proximal side of the AZ-C (pith side) and composed by recently divided cells based on the observation of thinner cell walls formed between cells (Divided Cells Area, DCA; Figure 3D). Then, the AZ-C was constituted by 10–15 cell layers distributed in cellular areas with two different cell morphologies and organellar composition (i.e., cells from the SA contain amyloplasts). The analysis of the AZ-C after 48 h of treatment suggests that cell wall degradation and cell degeneration occurred mainly at the layers of the SA in the fracture plane, adjacent to the mesocarp cells of the FR known as the albedo (see Figure 9). However, cell expansion occurred at the DCA (Figure 3F). These results together with kinetics (Figure 1A), phloroglucinol staining (Figures 1B,C) and SEM (Figure 2) data suggested that the activation of the fruit AZ-C by ethylene/ACC occurred early after treatment, probably prior to 24 h. The events related to cell wall loosening might start at 24 h, while cell separation seemed to begin at 48 h and to be completed at 96 h after treatment.

For gene expression analysis, we used the 20 K citrus cDNA microarray (Martinez-Godoy et al., 2008). We isolated cells from the central core of the AZ-C as well as from the FR located beneath the AZ-C through LM (Figure S2) to perform a time-course experiment (0, 12, and 24 h-ethylene) and compared data from each analysis. Results showed that ethylene differentially regulated 2280 genes exclusively in the AZ-C cells, 1742 genes exclusively in the FR cells, and 2001 genes were regulated by ethylene in both the AZ-C and the FR cells (Table S1).

All differentially regulated genes were grouped into functional categories according to the Munich Information Center for Protein Sequences (MIPS). The categories sugar, glucoside, polyol, and carboxylate metabolism and polysaccharide metabolism showed a higher percentage of regulation in the AZ-C compared with the FR (Table S2).

The set of genes discussed below was selected because of its prominent representation in the AZ-C or its particular biological interest. These gene families included those associated with cell wall metabolism and monolignol biosynthesis and polymerization (Table S3).

A high number of genes encoding cell wall modification enzymes were differentially regulated by ethylene exclusively in the fruit AZ-C. Our data suggested that this strong activation of cell wall metabolism occurred through both degradation and biosynthesis (Table 1) as we previously showed in the laminar abscission zone (LAZ) during leaf abscission (Agustí et al., 2008, 2009, 2012). Genes encoding enzymes that hydrolyze the cell wall and middle lamella would be responsible of the cell wall disassembly observed mainly in the SA at the distal side of the AZ-C (Figure 3F), which would lead to the effective organ separation. On the other hand, genes involved in the biosynthesis of new cell wall components would be related to the cell expansion observed mainly in the cell layers of the DCA located at the proximal side of the fracture plane (Figures 2C, 3F). The cell expansion might be associated with the generation of protective layers at the regions of the receptacle that remain exposed to the environment after abscission. Indeed, the recently divided cells at the proximal side of the AZ-C did not show cell degeneration (Figure 3F), suggesting that these cells might acquire such a function during the last step of organ abscission (Roberts et al., 2002; Estornell et al., 2013).

Cell wall-remodeling enzymes are key players in abscission, but also in many other biological processes. Thus, the identification of abscission specific genes within these gene families acquires special relevance. However, this task becomes complicated due to the large size and high functional diversification of such gene families. To identify potential abscission specific cell wall-remodeling enzymes as well as reinforce the gene expression data, a phylogenetic analysis of gene families related to cell wall modification was carried out. For this purpose, genes encoding cell wall-remodeling proteins in the sequence of the C. clementina haploid genome (http://www.phytozome.net; Wu et al., 2014) were retrieved and compared to those of orange (C. sinensis; http://citrus.hzau.edu.cn/orange/) to identify all members of these citrus gene families (Table S3). We considered the phylogenetic relationships between cell wall-remodeling proteins of citrus and Arabidopsis along with proteins previously described as associated with abscission or dehiscence in other plant species.

Having identified a set of cell wall-related genes that were up-regulated during fruit abscission and to test whether such gene expression changes corresponded to enzymatic activity, we next attempted to identify alterations in pectic polysaccharides arrangements at the AZ-C cell walls. To examine such changes, we carried out an immunolocalization of pectic polysaccharides assay during ACC-promoted abscission. At 0 h, the epitopes (1 → 4)-β-D-galactan, (1 → 5)-α-L-arabinan, and partially methylesterified/de-esterified homogalacturonan (HG) recognized by the mAbs LM5 and LM6 and JIM5, respectively, were homogeneously distributed along cell walls of the AZ-C (Figures 7A–C). An increase in the (1 → 4)-β-D-galactan (LM5) labeling intensity in the starch-rich cell area of the AZ-C was observed at 24 h after ACC treatment (Figure 7E) whereas (1 → 5)-α-L-arabinan (LM6) and partially methylesterified/de-esterified HGs (JIM5) labeling highlighted cell wall changes in the divided cell area (Figures 7F,G). At a more advanced stage of the process (48 h), the labeling intensity detected for the three mAbs was higher at the cell layers closer to the separation line (Figures 7H–J), probably due to the accumulation of residual compounds of these pectic polysaccharides from the dissolution of the middle lamella and cell walls. Furthermore, an amorphous material with high fluorescence was observed at the regions where cell separation was completed (Figures 7K–M), in accordance with findings observed in Figure 2B and also with the first anatomical and histological description of citrus fruit abscission reported by Wilson and Hendershott (1968).

Changes in the pectic polysaccharides distribution in the AZ-C cell walls enabled us to correlate evidences of enzymatic activity with gene expression results. In particular, based on immunodetection of partially methylesterified/de-esterified HG and expression results, we propose that PMEs CitPME24 and CitPME41 and PAE CitPAE4 may act on de-esterification of HGs in the AZ-C cell walls (Table 1, Figure S3). The activity of PMEs and PAEs is thought to facilitate the subsequent action of pectin hydrolases (Chen and Mart, 1996). Thus, the PGs CitPG43, CitPG16, CitPG20, CitPG41, and CitPG42, and the PLs CitPL5, and CitPL19 may potentially hydrolyze the HGs highly accessible after CitPME24, CitPME41, and CitPAE4 activity (Table 1, Figures 5, 6, Figures S3–S5). Finally, the only α-L-arabinofuranosidase identified in citrus (CitASD1) did not show significant changes in gene expression based on the statistical cutoff mentioned in materials and methods (Figure S4). In Arabidopsis, it has been reported that AtBXL1 and AtBXL3 acted as bifunctional α-L-arabinofuranosidase/β-D-xylosidases (Minic et al., 2004; Arsovski et al., 2009). Therefore, changes observed in (1, 5)-α-L-arabinans at the AZ-C might be due to the dual activity of β-XYLs such as CitBXL11 and CitBXL19, which were in the same clade as AtBXL1 and AtBXL3 (Figure S4).

Significant expressed genes belonging to different gene families involved the monolignol biosynthesis pathway were up-regulated by ethylene exclusively in the AZ-C (Table 2, Figure 8, Figure S7). These included a phenylalanine ammonia-lyase (CitPAL5), a p-coumarate 3-hydroxylase (CitC3H1), a hydroxycinnamoyl-CoA shikimate/quinate hydroxycinnamoyl transferase (CitHCT2), a 4-coumarate-CoA ligase-like protein (Cit4CL7), a cinnamoyl-CoA reductase-like protein (CitCCR1), and a cinnamyl alcohol dehydrogenase (CitCAD3). In addition, three genes encoding proteins involved in the oxidative coupling of monolignols and belonging to the CASPARIAN STRIP MEMBRANE DOMAIN PROTEIN family (CitCASPL1D1a, CitCASPL2B2, and CitCASPL4A4) were also up-regulated by ethylene exclusively in the AZ-C (Table 2, Figure 8, Figure S7). These results correlated with the observation of lignin deposition in the AZ-C after 24 and 48 h of ethylene/ACC treatment (Figures 1B,C) and the increase in the level of lignin intermediates detected by UPLC-MS/MS in AZ-C cells (Figure 8B). A significant increase of coumaric acid was observed at 12 h after ACC treatment. In addition, levels of caffeic acid and ferulic acid, compounds that are synthesized from coumaric acid, increased at 12 h and were maintained up to 36 h. Taken together, these data mainly reflected the activation of the H lignin pathway which results from the incorporation of p-hydroxyphenyl (H) units into the lignin polymer (Figure 8C). However, G lignin (lignin with guaiacyl units) biosynthesis would be also possible in activated AZ-C cells since an increase in CitC3H1 and CitHCT2 expression and caffeic and ferulic acids levels also occurred in AZ-C cells despite any member of the CCoAOMTs gene family were up-regulated (Figure 8). Regarding S lignin biosynthesis, four caffeic acid O-methyltransferases (CitCOMT2, CitCOMT3, CitCOMT6, and CitCOMT12) were down-regulated exclusively in AZ-C cells (Table 2, Figure 8C). CitCOMT3 was closely related in sequence to AtCOMT1 (Figure S7), a member of the Arabidopsis COMT gene family with 5-hydroxyconiferaldehyde O-methyltransferase activity that has been implicated directly in S lignin synthesis (Nakatsubo et al., 2008). These results suggest that the S lignin pathway might be inactive in AZ-C cells during fruit abscission. In woody dicotyledons, such as citrus, lignin is polymerized from mostly G and S lignin subunits (Sarkanen and Ludwig, 1971). However, lignin composition can differ among cell types (Nakashima et al., 2008; Ruel et al., 2009) and expression data suggested that cell walls of AZ-C cells might be mainly enriched in H lignin and probably also in G lignin subunits.

The role of lignin deposition has been associated with the generation of protective layers at the tissues remaining in the plant during the last step of the abscission process (Addicot, 1982; Agustí et al., 2008; Van Nocker, 2009). However, it has been suggested that lignification could also facilitate the mechanical cell wall breakage during cell separation processes (Sexton, 1979; Liljegren et al., 2000). In the AZ-C, lignin deposition only occurred at the distal side of the AZ-C (Figure 8A). This differential deposition of lignin strongly suggests that this polymer mainly acts by generating a tension in the fracture plane to facilitate cell wall breakage during citrus fruit abscission rather than forming protective layers.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.