Tomato Endurance to Tuta Absoluta Is Mediated by Reinforcement of Leaf Barriers and Induction of Chemical Reaction 


 Background: The leaf miner, Tuta absoluta, is a destructive pest of tomato that causes important losses worldwide. Breeding of resistant/tolerant tomato cultivars could be an effective strategy for T. absoluta management but, despite the economic importance of tomato, very limited information is available about its response to this treat. To elucidate the defense mechanisms to herbivore feeding a comparative analysis was performed between a tolerant and susceptible cultivated tomato at both morphological and transcriptome level to highlight constitutive leaf barriers, molecular and biochemical mechanisms to counter the effect of T. absoluta attack. Results: The tolerant genotype showed an enhanced constitutive barrier due to the higher density of trichomes and increased inducible reactions upon mild infestation thanks to the activation/repression of key transcription factors regulating genes involved in cuticle formation and cell wall strength as well as of antinutritive enzymes, and genes involved in the production of chemical toxins and bioactive secondary metabolites. Conclusions: Overall, our findings suggest that tomato resilience to the leaf miner is achieved by a combined strategy between constitutive and induced defense system. A well-orchestrated modulation of plant transcription regulation ensures a trade-off between defense needs and fitness costs. Our finding can be further exploited for developing T. absoluta tolerant cultivars, acting as important component of integrated pest management strategy for more sustainable production.


Background
The tomato (Solanum lycopersicum L.) is attacked by a wealth of insect herbivores, which cause signi cant crop losses and generate the need to de ne effective pest management programs, both under eld and glasshouse conditions [1]. Among these insect pests, the tomato pinworm, Tuta absoluta (Meyrick) (Lepidoptera: Gelechiidae), is one of the main threats worldwide [2], which is native from South America and, in the last decade, has invaded Europe, Africa and Asia [3].
Any part of tomato plants can be attacked by T. absoluta, at any developmental stage. Adults usually lay eggs preferentially on leaves and, to a lesser extent, on stems and fruits. Newly hatched larvae penetrate into the plant tissues, on which they feed concealed in mines and galleries [4]. Both yield and fruit quality can be signi cantly reduced by insect direct damage and by secondary pathogens, which may enter the plant tissues through the feeding wounds [5]. Since the early 1980s, chemical control was of pivotal importance in most integrated pest management (IPM) programs [6], causing several problems. Indeed, this has promoted a rapid development of insecticide resistance and number of negative effects on environment, non-target species and bene cial insects, impacting both food safety and economic sustainability of the production process [7,8]. Therefore, it would be highly desirable to develop resistant/tolerant tomato varieties to reduce the use of chemical insecticides and pests damage.
Tomato was extensively used as a model plant for resistance/tolerance studies and important progresses were made through both genetic and biotechnological approaches [9,10]. It is known that tomato plants have developed numerous defense mechanisms that are effective against insect pests, such as physical barriers (i.e. glandular trichomes) [11], synthesis of speci c metabolites [12], expression of defense proteins such as proteinase inhibitors [13] and polyphenol oxidase [14]. However, despite the economic importance of tomato, little is known about its molecular interaction with T. absoluta. To date, only limited information about resistance traits to this leaf miner is available for wild tomato species, such as S. habrochaites and S. pennellii [5,15,16,17,18] Therefore, lling this research gap is crucial to obtain new tomato varieties resistant/tolerant to T. absoluta.
In this study, we performed a comparative analysis between a tolerant and a susceptible tomato line to unravel the plant tolerance mechanisms both at phenotypic and transcriptomic level. We examined morphological leaf traits, such as trichome types and distribution, and assessed the transcriptional reprogramming of these two tomato genotypes in response to T. absoluta feeding. The gathered information is a valuable basis on which to develop future breeding plans aiming to enhance the plant defense barriers against this pest.

Methods
Plant material and semi-eld infestation trial A cherry type tomato line putatively tolerant/partially resistant to T. absoluta, BR221 (from now on named as 'T') and a susceptible one, PS650 (from now on named as 'S') were used. An infestation experiment with T. absoluta adults was set up in a ying nylon tunnel (500 cm x 90 cm x 90 cm) located in a glasshouse. The tunnel was divided in two cages by a plastic mesh screen. For both tomato genotypes, 20 plants, 4 weeks old, were placed in each cage, using a randomized complete block design. The plants of one cage were exposed to 100 females and 100 males of T. absoluta, while the second cage served as control. Plants were inspected for egg laying and infestation level at 10-12 day interval up to 40 days after adult release, when the overall plant damage was assessed (type of lea et damage, the number of injuries/mines/cm 2 of lea ets, the percentage of infested fruits), using the scoring method proposed by Resende et al. [15]. On that occasion, leaves with and without mines from each plant were singly collected and immediately frozen in liquid nitrogen for subsequent molecular analyses. The data obtained were processed using the software Statgraphics PlusR (Statgraphics Technologies Inc., The Plains, VA, USA). Data satisfying conditions of normality and homoscedasticity, or after appropriate transformation, if needed, were analyzed by a two-tailed Student's test and the means were separated at the 0.001 level of signi cance.

Microscopy analysis
Leaves were randomly harvested from T and S control tomato plants. Five fully expanded leaves were selected from each plant and ve measurements were made on each sample. The different types of trichomes present on the two tomato genotypes, respectively belonging to classes II-VII, were identi ed and counted according to Channarayappa et al.,19]. Observations were made with a metallurgical microscope (Leitz Wetzlar Ortholux Microscope) equipped with an Ultropak® objective 6.5x, with a visual eld area of 3.1 mm 2 . The data obtained were analyzed by the Student's t-test (P < 0.001).

RNA extraction and sequencing
Leaves collected from single plant belonging to each treatment (Tni= Tolerant non infested; Ti=Tolerant infested; Sni= Susceptible non infested; Si= Susceptible infested), after 40 days, were pooled in three independent biological replicates, each consisting of leaves deriving from ve-six plants, for reducing sampling error. Total RNA extraction was performed on 100 mg of frozen tomato leaves, collected from infested plants but not directly damaged by T. absoluta, using Spectrum™ Plant Total RNA Kit (Sigma Aldrich), according to manufacturer's protocol. A treatment with On-Coloumn DNaseI Digestion Set (Sigma Aldrich) was carried out on total RNA to remove the contaminating genomic DNA. RNA puri ed samples were quanti ed by NanoDrop ND-1000 Spectrophotometer (Nano-Drop Technologies, Wilmington, DE, USA), RNA integrity was rst checked by horizontal electrophoresis on a 1% (w/v) agarose gel with GelRed Nucleic Acid Stain 10,000X (Biothium) by UV light (UV Gel Doc BIORAD) and con rmed by Bioanalyzer (Agilent Technologies). Samples with required standards of quality were rst converted to cDNA libraries and then sequenced on Illumina HiSeq1500 platform by paired-end 100 base pair (bp) sequencing using a strand-speci c library. The quality of the sequences generated were evaluated by the FastQC software [20]. High quality reads, with a minimum length of 25 bp and a quality score of 35 were used for further analysis. High quality reads were aligned against the Solanum lycopersicum reference genome sequence (S_lycopersicum_chromosomes 2.50) with TopHat (version 2.0.11). Uniquely mapping reads were used as input for Feature Counts (Subread package, version 1.4.4, [21]) to calculate gene expression values (read counts).
Differentially expressed genes (DEGs) identi cation was performed using the edgeR package [22] considering all the genes passing the HTSFilter step.

Transcriptome data analysis
The annotation of biological information related to the identi ed differential expressed loci was performed using ITAG2.4 protein functional annotation le. The R package topGO [23] was used to carry out a Gene Ontology Enrichment Analysis with a P-value cut-off of 0.05. DEGs assignment to speci c metabolic pathways was performed using MapMan 3.0.0 tool [24] and Plant MetGenMAP package [25]. The Sol Genomics database [26] was queried to gather additional information on annotated genes, while SolCyc database [27] was used to obtain detailed information on pathways and biochemical reactions involved in the tomato-T. absoluta interaction.

Inspection of Tuta absoluta infestation
At all sampling times the parameters scored showed highly signi cant differences between the T and S lines (P < 0.0001, t-test, Table 1). At 10 days after adult release (10 DAAR), the difference in number of eggs/plant laid on the two tomato lines was highly signi cant (t=-6.16; n=20; P < 0.001) with the average value registered on T genotype about 4 times lower than on S (Table 1). At 20 DAAR, the highly signi cant difference in the average number of mines/plant recorded in T and S remained in the same range (t=-5.05; n=20; P < 0.001). The plant damage assessed 20 days later (i.e. 40 DAAR) followed the same trend of signi cant differences between the two genotypes, both in terms of number and type of mines. In particular, an average of 0.2 mines/cm 2 and 0.71 mines/cm 2 of lea ets were counted on T and S, respectively (t=-9.26; n=20; P < 0.001). Moreover, while in the T genotype the large majority of mines were single and non-coalescent, for the S genotype this type of mine was recorded only on 52.8% of analyzed samples (t=-7.94; n=20; P < 0.001). The difference in the level of fruit infestation per plant was even more pronounced, with a 6.5% for T genotype and 33% for S genotype (t=-7.75; n=20; P < 0.001).

Tomato transcriptional response
The changes of gene expression in tomato leaves as affected by T. absoluta feeding on T and S genotypes were determined. RNAsequencing produced an average of 24 million reads per sample (Additional File 1: Table S1). Approximatively 19,000 and 21,500 transcripts, for T and S respectively (Additional File 1: Fig. S1a,b), were mapped to the tomato reference genome (SL.2.50) and used for further analysis. Differentially expressed genes (DEGs) were computed for each genotype (T and S) comparing T infested vs T non-infested (Ti vs Tni) and S infested vs S non-infested (Si vs Sni) conditions. The comparison Ti vs Tni showed a marked gene expression change in T genotype after T. absoluta challenge (11486 DEGs, Fig. 2a, Additional le 2: Dataset S1-S2), while for the S genotype (Si vs Sni) a total of 6,793 DEGs were obtained (Fig. 2a, Additional le 2: Dataset S3-S4). Out of the total DEGs, 2,845, and 696 genes resulted speci cally upregulated in T and S, respectively, while 3074 in T genotype and 530 in S genotype were speci cally down-regulated (Fig. 2b). More than 5,000 genes (3,039 up-regulated and 2,431 down-regulated) resulted differential expressed in both genotypes.
The perturbations of tomato pathways induced by T. absoluta feeding were depicted by integrating Gene Ontology enrichment analysis and DEGs metabolic mapping. Key metabolic pathways, modulated during the herbivore-plant interaction, in the two tomato genotypes (T and S), in infested vs non-infested conditions, were mainly related to cell wall, sterol and steroid metabolism, leaf development and photosynthesis (Fig. 2c).

Plant development and leaf structures morphogenesis and differentiation
The T genotype showed an enrichment of up-regulated genes involved in leaf development, structural meristem formation and photosynthesis (Fig. 2c). The GO term "photosynthesis, light harvesting in photosystem I", included many Chlorophyll a-b binding proteins.
The GO terms "leaf development" and "leaf morphogenesis" enclosed an abundance of receptor like kinases (RLKs) such as Solyc08g061560, encoding a putative orthologue of ERECTA, Solyc08g014030, a gene closely related to the Arabidopsis gene encoding SHORTROOT (SHR) [28], which, together with a GRAS-domain TF, SCARECROW (SCR), regulates the duration of cell proliferation in leaves [29], the Solyc03g112750 protein tornado-TRN1 and the Solyc04g081590 CLAVATA 1, involved in meristem organization. Transcription factors (TFs) involved in meristem formation and trichome differentiation, exclusively present in T, including PHANTASTICA (Solyc09g010840) belonging to MYB family, Solyc02g092370 (SlSHRa) belonging to GRAS family, Solyc08g066500 (SlHB8) an HD-ZIP III transcription factor, and a subunit of the CAF (Solyc11g008670, FAS1) were also identi ed (Table 3). An enrichment of genes involved in cell wall synthesis, degradation and assembly as well in cell wall structural proteins was observed in T ( Fig. 3a,b), including glycosyl transferases (GTs) families involved in cellulose (GT2), hemicelluloses (GT47), pectin (GT8) and xylan (GT43) synthesis (Fig. 3c) as well as glycoside hydrolases (GHs) belonging to GH9, GH5 and GH31 families, lyases (PL1), esterases (CE1) and expansins (Fig. 3c). Moreover, a higher number of cell wall associated genes, such as Fasciclin-like arabinogalactan (FLA) proteins (13 up-regulated in T and 5 up-regulated in S), involved in cell wall modi cation and assembly, and Trichome Birefringence-Like (TBL) genes (21 up-regulated in T and 9 up regulated in S) were also detected in the T genotype (Fig. 3c).
Perception of damage, signal transduction and defense response A plethora of up-regulated RLKs resulted enriched in T during T. absoluta challenge (Fig. 4a,b). Most of them belonged to families involved in defense response and growth. In particular, SERK1 Solyc04g072570 (LRR II), several NUCLEAR SHUTTLE PROTEIN-INTERACTING KINASE NIK members and LRR-XIII members (including Solyc03g007050-ERECTA-like 1 and Solyc08g061560-ERECTA) as well as speci c extensin proteins were identi ed.
Divergences in the signal transduction between genotypes were also detected in the activation of mitogen-activated protein kinases (MAPKs) and calcium-dependent protein kinases (CDPKs) (Fig. 4c,d). In T was observed the up-regulation of SlMPK3, well known to be induced by wounding, and SlMAPK5, homologous to A. thaliana MPK4 (AT4G01370.1), a positive regulator of JA mediated signaling (Fig.  4c). Interestingly, in T also the up-regulation of SlCDPK18 (equivalent to LeCDPK1) and SlCDPK4/LeCDPK2 (Fig. 4d) was noted.

Defensive secondary metabolites synthesis
The sterol metabolism resulted strongly activated in the T genotype, as indicated by the up-regulation of a sterol side chain reductase enzyme (SSR2, Solyc02g069490), catalyzing the conversion of cycloartenol (the precursor for phytosterols) to cycloartanol, the rst committed step in cholesterogenesis, that, in contrast, in S was down-regulated (Fig. 5a). In addition, the core steps of sterol biosynthesis, starting from cycloartenol, resulted clearly activated in T, as shown by the speci c up-regulation of Obtusifoliol 14-alpha demethylase (O14DM), a delta14-sterol reductase (D14SR) as well as of three Sterol 4-alpha-methyl-oxidases (SMO, Solyc01g091320, Solyc06g005750 and Solyc08g079570) and an Acid phosphatase-like (Solyc04g072190). Interestingly, among the SMO, the gene Solyc06g005750 was upregulated in T and down-regulated in S (Fig. 5a). The sterol transport seemed to be activated in the T genotype, as indicated by upregulation of two ABC(G) transporters, Solyc12g100180 and Solyc12g100190. The production of sterol glycoalkaloids (SGA) in the T genotype was promoted by up-regulation of genes involved in glycoalkaloids biosynthesis (GAME), clustering on chromosome 7 (Fig. 5b).
DEGs were identi ed in the pathway of glucosinolate biosynthesis in both genotypes, while a desulfoglucosinolate sulfotransferase, Solyc03g114800, was exclusively up-regulated in T. Furthermore, two 3-isopropylmalate dehydratases (Solyc03g005730 and Solyc09g090900), involved not only in leucine synthesis but also in the production of the glucosinolate methylthioalkylmalate, resulted upregulated only in T.
Transcriptional changes in the terpenoid biosynthetic pathway of T plants were relatively higher, both in terms of DEGs number and of their log 2 FCs. Interestingly, a cluster of genes located on chromosome 8, including the terpene synthases (TPS) TPS20, TPS21 and TPS18, a Dimethylallylcistransferase CPT1/NDPS1 and the cytochrome P450-oxidoreductase CYP71BN1, were speci cally up-regulated in T. In addition, in T the up-regulation of TPS5 was observed, while in S it was observed for TPS46. Both genotypes up-regulated TPS7, but with a higher induction in T (Fig. 5c).
The biosynthesis of volatile benzenoid esters, mediated by two benzoyl transferases (SlAAT3, Solyc07g049660 and SlAAT5, Solyc05g015800) producing benzylbenzoate, resulted exclusively activated in T (log 2 FC=2,51 and log 2 FC=2,48). Synthesis of the toxic compound quinone 1,4-dihydroxy-2-naphthoate was promoted in T, as indicated by the up-regulation of all genes involved in the biosynthetic pathway, while only one gene was activated in S. T also showed a higher induction of genes required for the synthesis of hydroxycinnamic acid tyramine amides (Fig. 5d) and of genes involved in the production of a modi ed fatty acid, clustering on chromosome 12 (Fig. 5e).

Transcription Factors
The transcriptional regulation in response to T. absoluta attack mediated by TFs in T and S was examined in-depth, since T genotype showed a very high number of speci cally differentially expressed TF, including 145 up-regulated and 192 down-regulated. Among the TF with a contrasting expression, ve were up-regulated in T and down-regulated in S. On the contrary three were up-regulated in S and downregulated in T (Fig. 6a). The classi cation of speci c TF showed that they were distributed among different classes with many copies belonging to bHLH and MYB family (Fig. 6b).
Transcription factors involved in the regulation of secondary metabolites production (i.e. terpenes), trichome formation and jasmonic acid signaling, and cutin/wax metabolism were identi ed (Table 3). T speci cally up-regulated two transcription activators, SlMYC1 and EXPRESSION OF TERPENOIDS 1 (EOT1), controlling the terpene biosynthesis in glandular trichomes and the development of type VI glandular trichome. GLABRA2 (GL2, Solyc03g120620), implicated in cell differentiation of various epidermal components, including trichomes, was also up-regulated in T and down-regulated in S. By contrast, TRIPTYCHON (SlTRY, Solyc01g095640), a negative regulator of trichome formation, was up-regulated in S while the bHLH TF (SlbHLH150, Solyc09g065100), important for marginal trichome development, was exclusively induced in T.
Jasmonic acid 1, involved in the activation of threonine deaminase (TD), was up-regulated in T (Solyc05g007180, log 2 FC=2,32). By contrast, MTB1 (bHLH113, Solyc01g096050), that is negative regulator of JA signaling, was down-regulated in T. MYB30, a key regulator of both the protective hypersensitive response (HR) and wax biosynthesis, was up-regulated in T whilst MYB41, that mediates the negative regulation of cutin biosynthesis in response to stress, was repressed in T and induced in S. Furthermore, the negative regulator of cuticle development CFL1 (Solyc01g009770, log 2 FC= -1.01) resulted down-regulated in T.

Reinforcement of external structural barriers
The rst line of defense against herbivores and pathogens is often provided by physical barriers, such as trichome density and quality, cuticle tickness, and cell wall strength [30]. A higher density of glandular trichomes, (type VI and type IV) as well as non glandular trichomes (type II and V), trait of resistance to herbivory [31,32], was observed in T. It is conceivable that difference in frequency and quality of trichomes on the surface of S compared to T could affect the Tuta feeding preference. Glandular trichomes secrete avonoids, poisonous terpenoids and alkaloids [30,33] that act as deterrent for oviposition and feeding of insect pests [34], and serve as mechanoreceptors of herbivores [35]. In particular, type VI trichomes accumulate JA, an important sensor for detecting insect movement on the leaf surface [35].
In T genotype, T. absoluta attack could trigger the production of trichomes on newly emerging leaves due to the up-regulation of the gene SlMYC1 involved in type VI trichome formation and in the synthesis of mono and sesquiterpenes [36]. Indeed, SlMYC1 knocking down induced the production of smaller type VI glandular trichomes at lower densities and its knocking out led to their absence [37]. In addition, SlbHLH150 (Solyc09g065100), a transcription factor belonging to helix-loop-helix gene family [38], and GLABRA2 (GL2), encoding a homeobox leucine zipper protein, promoting leaf development and trichome formation [39,40,41] were also up-regulated in the T genotype.
By contrast, the up-regulation in S of Solyc01g095640, a negative of regulator of trichome formation suggested that trichome morphogenesis process was reduced.
We hypothesize that the cuticle structure of T genotypes could be reinforced in response to T. absoluta attack thanks to the up-regulation of genes involved in cutin monomers formation and assembly, such as HOTHEAD (HTH), Glc-methanol-choline oxidoreductase family proteins [42] and MYB30, a crucial regulator of both wax biosynthesis and Systemic Acquired Resistance (SAR) [43]. Instead, a negative regulation of cutin biosynthesis was observed in S as indicated by the up-regulation of the genes MYB41 and CFL1 [44,45]. T cell wall remodeling promoted reinforcement and the maintainance of cell function thanks to the induction of glycosyl transferases (GTs), glycoside hydrolases (GHs), polysaccharide lyases (PLs) and carbohydrate esterases (CEs). Acetylation of cell wall polymers enhanced the interaction of the hemicelluloses with other wall polymers [46], to obtain a more rigid wall conformation [47]. Moreover, the activation of a higher number of fasciclin like arabinogalactan proteins, involved in cell-cell communication during cell wall remodeling, supported the idea that changes in cell wall conformation occurred. The dirigent-like (DIR) proteins activated in T may be involved both in dynamic reorganization of the cell wall and in the production of defense compounds such as lignin and lignans [48], disruptors of insect endocrine system [49]. DIR-like protein-encoding genes during the interaction of moss Physcomitrella patens with Colletotrichum gleosporioides or Pectobacterium carotovorum promoted cell wall modi cations, including increased incorporation of phenolic compounds [50]. In addition, the signi cant overexpression of the peroxidase gene TPX1 suggested that it served as storage pool of monolignols for ligni cation for enhancing the rigidity and decreasing the digestibility of the cell wall [51]. The accumulation of hydroxycinnamic acid tyramine amides as well as of phenolic compounds (E-feruloyltyramine and E-p-coumaroyltyramine) in T genotype may limit the T. absoluta infestation as well as the digestibility of the cell wall [52,53].

Molecular defense response activation
The extensive reshaping of gene expression in T genotype induced by T. absoluta feeding activated a signaling cascade that culminated into a complex re-arrangement of primary and secondary metabolism leading to a systemic plant defense response. The cell surface perception resulted enhanced in T by the induction of cell surface receptor-kinase ERECTA (ER, Solyc03g007050), BAK1 and SERK1, involved in plant growth, developmental and immune processes regulation [54]. Members of the ER pathway interact with BAK1 in the regulation of innate immune response, plant growth and cell wall composition [55,56,57,58]. A prompter signaling in T could be mediated by the up-regulation of the calmodulin and calcium-sensing genes, involved in the activation of calcium cascade, and of MAPKs and wounding signaling [60]. These signal transduction pathways lead to the activation of well-characterized proteins involved in ROS detoxi cation such as PPO (Solyc08g074630-Solyc08g074680), proteinase and metallocarboxypeptidase inhibitors (Solyc09g084470, Solyc07g007260, Solyc07g007250, Solyc06g061230 and defensin proteins (Solyc07g007760, Solyc07g007750) that prevent pest damage [61]. Defense signaling acted synergistically to activate a nely tuned response to increase tolerance to T. absoluta without compromise plant tness.

Inducible chemical defense response ampli cation
Inducible defenses against herbivores include the synthesis of a wide range of species-speci c toxic plant secondary metabolites (e.g., phenylpropanoids, avonoids, anthocyanins, alkaloids, terpenoids, glucosinolates), and anti-nutritive enzymes and proteins (e.g., proteinase inhibitors (PIs), amino acid catabolizing enzymes, polyphenol oxidases, and peroxidases), mostly controlled by jasmonic acid [62,63,64]. T showed a much more pronounced activation of the gene arsenal involved in primary metabolism with defensive functions against insects [65], characterized by the presence of a higher number of up-regulated genes or of genes with a higher FC, such as the JA-induced defenses markers, including antinutritive and enzyme proteins TD, LAP, PPO and PI and defensins.
The divergent synthesis and composition in alkaloids, terpenes and glucosinolates between the two lines underlined the signi cant difference found in trichome density and type [66]. T genotype could accumulate steroidal glycoalkaloids (SGAs), as emerged from the activation of genes involved in GLYCOALKALOID METABOLISM (GAME). In particular, GAME25 gene promoted higher levels of α-tomatine and acetoxytomatine that act synergistically against pathogens to exert cytotoxic and anti-nutritional function [67]. Moreover, genes involved in cholesterol production, that serves as the precursor for the biosynthesis of SGAs as well as of two orthologs of Petunia hybridaPDR2, involved in the control of steroidal compounds accumulation in leaves and trichomes [68], were highly induced in T. This is consistent with the fact that tomato resistance to phytophagus Spodoptera litura caterpillars was weakened by reduced SGA accumulation [69].
T. absoluta feeding could increase the production of some monoterpenes and other minor products in the T genotype [70] thanks to the activation of terpene synthases including the trichome-localized TPS5/MTS1, that preferentially produces linalool [71], TPS21 that, together with CYP71BN1, leads to the formation of lycosantanolol [72], and with CPT1/NDPS1 is involved in phellandrene synthesis [73]. By contrast, the up-regulation of TPS46/GLS in S lead to the production of the diterpene geranyllinalool [74]. Finally, although both genotypes up-regulated TPS7, the stronger induction in T suggests a higher production of pinene, myrcene, sabinene and linalool [75]. A different level of glucosinolate biosynthesis in T was supposed by the up-regulation of a desulfoglucosinolate sulfotransferase (Solyc03g114800) promoting the accumulation of 3-methylthiopropyl-glucosinolate at expense of other types of glucosinolates. In addition, the formation of benzylbenzoate, a major volatile compound with ovicidal properties produced after leaf disruption and wounding [76] was exclusively promoted in T by up-regulation of two benzoyl transferases (Solyc07g049660; Solyc05g015800). T genotype also showed a higher number of up-regulated genes involved in the production of 1,4-dihydroxy-2-naphthoate (DHNA), toxic precursor toward insects [77]. A major involvement of up-regulated clustered genes, required for the production of the highly functionalized oxylipin falcarindiol [78] in T leaves, was also observed.

Conclusions
Collectively, the gathered experimental data suggest that both constitutive barriers and inducible reactions are enhanced in T genotype in response to T. absoluta infestation. Leaf miner development was impaired by multiple constraints through a consistent remodulation of the main cellular processes and secondary metabolism synthesis. Recovery from infestation occurs through photosynthesis and general cellular process promotion. A well-orchestrated modulation of transcription regulation ensured a trade-off between defense needs and tness costs. A coordinated activation/repression of transcription factors involved in the production of antinutritive enzyme, chemical toxins and bioactive secondary metabolites drove the adjustment of plant structures and chemical compound synthesis to counter T. absoluta feeding.

Declarations
Ethics approval and consent to participate Not applicable Consent for publication Not applicable Availability of data and materials The datasets supporting the conclusions of this article are included within the article (and its additional les). The Illumina sequence data generated during the current study are available in NCBI's Gene Expression Omnibus [79] and are accessible through GEO Series accession number GSE159085 (http://www.ncbi.nlm.nih.gov/geo)

Competing interests
The authors declare that they have no competing interests.

Funding
This work was supported by Prinpom project funded by PSR Regione Campania Authors' contributions DD was centrally involved in gene annotation, data interpretation and in manuscript writing. DM was involved in transcriptomic analysis, data interpretation and paper drafting. AR was involved in infestation analysis. APG contributed to infestation data analysis and interpretation. ADN performed microscopy analysis and morphological data interpretation. LF critically revised the manuscript. FP revised the manuscript draft providing useful suggestions. MRE conceived the study and was mainly involved in data interpretation and in manuscript writing. All authors read and approved the nal manuscript.   T  S  T  S  T  S  T  S  T  S  T  S   Leaf  Trichome  II  III  IV  V  VI  VII   Adaxial  N° (cm 2 )