Ascochyta blight is one of the main lentil diseases in most of the countries producing this crop. This disease, caused by the fungus Ascochyta lentis (syn A. fabae f. sp. lentis , teleomorph Didymella lentis ), affects all above ground parts of the plant and is characterized by necrotic lesions. On susceptible cultivars under favorable conditions, it can lead to breakage of the stems and severe yield reduction. Seed quality may also be reduced through seed discoloration or retardation of seed development. Chemicals can control Ascochyta blight, but besides the environmental problems and the inputs increase related with the use of fungicides, the development of resistant cultivars is considered a more efficient and sustainable approach ( Davidson & Kimber, 2007 ).

Ascochyta lentis is a host-specific pathogen ( Pee­ver, 2007 ) and it is considered as a necrotrophic fungus, although a short biotrophic period cannot be completely excluded ( Tivoli & Banniza, 2007 ; Sa­ri et al. , 2017). The pathogenicity of this type of fungus and the resistance of the plant could be related to the production of specific fungal toxins and plant receptors or detoxifying molecules. Whatever the me­chanism is, it should include the recognition of the pathogen in order to trigger the defense response. The recognition can be based on interactions between pathogen-associated molecular patterns (PAMPs) and the corresponding plant pattern recognition receptors (PRRs) in the pattern-triggered immunity (PTI), or alternatively it could be based on pathogen effectors and plant receptors, in special nucleotide-binding domain and leucine rich repeat containing (NLR) proteins, in the effector-triggered immunity (ETI). Both ways result in the activation of the plant immune response, mainly by the jasmonic acid (JA) and ethylene (ET) dependent pathways, which act in a synergistic way (JA/ET pathway). The salicylic acid (SA) signaling pathway, and probably the abscisic acid (ABA) and gibberellic acid (GA) signaling pathways, also have a role in the final output ( Mengiste, 2012 ; Abdullah & Akhtar, 2016 ).

Gene expression analyses in lentil revealed that the JA and SA signaling pathways play a central role in determining the resistance/susceptibility to A. len­tis , although the ET, SA, GA and ABA pathways might also be involved ( Mustafa et al. , 2009 ; Ford et al. , 2017 ). Nevertheless, at least for some genotypes, only the JA pathway genes could explain the different reaction against Ascochyta in resistant vs susceptible lentil cultivars ( Sari et al. , 2017 ). These observations generally agree with the results obtained in other legumes after their host-specific Ascochyta infection, as in Pisum ( Fondevilla et al. , 2014 , 2018 ), Lathyrus ( Almeida et al. , 2015 ), Vicia ( Ocaña et al. , 2015 ) or Cicer ( Madrid et al. , 2012 ).

Recently, Khorramdelazad et al. (2018) and Sari et al. (2018) have utilized the RNA-seq approach for the study of transcriptome changes in lentil after the infection by A. lentis . In this way, a set of candidate genes involved in the different stages of the defen­se response has been identified, from the patho­gen recognition to the systemic acquired resistance. Although the response was genotype-specific, showing up the existence of several resistance mechanisms in lentil ( Sari et al. , 2018 ), some common characteristics are found in the resistant/tolerant genotypes, such as an earlier and faster detection and signalling response to the infection, and a higher expression level of structu­ral defense-related genes, in special those coding for proteins involved in the reinforcement of cell walls ( Khorramdelazad et al. , 2018 ).

An alternative technology to analyze gene ex­pression is the Massive Analysis of cDNA Ends (MACE) because it is considered one the most cost-effective methodologies for the studies on changes in gene expression ( Zajac et al. , 2015 ). This technique has proved its usefulness in other plant species and it yielded data on candidate genes further validated by qPCR ( Yakovlev et al. , 2014 ; Zajac et al. , 2015 ; Weiꞵ et al. , 2017 ; Weiꞵ & Winkelmann, 2017 ). Our main objective was to obtain a general overview of the changes in gene expression that occur in other interesting lentil genotypes showing different levels of resistance when they are infected by Ascochyta using the MACE technique. The purpose was to identify some general and genotypic-specific defense responses, in special the common gene expression changes showed by the resistant genotypes, in order to obtain a set of candidate genes presumably involved in the resistance. This group of genes will provide the basis for the development of genic markers useful in breeding programs, and for further studies on mapping and expression changes of the candidate genes in other lentil genotypes.

Three samples of Lens culinaris have been used, two of them belong to the cultivated species L. c. subsp. culinaris (accession 'ILL5588' and cv. 'Lupa') and one to the wild ancestor L. c. subsp. orientalis (accession 'BG 16880'). 'Lupa' is a Spanish cultivar, 'ILL5588' is an accession from the International Center for Agricultural Research in the Dry Areas (ICARDA) lentil collection and 'BG 16880' is an accession from the Spanish Centro Nacional de Recursos Fitogenéticos (CRF). The response of the three genotypes to the A. lentis infection was different: L. c. orientalis showed resistance to the fungus, 'ILL5588' could be considered as moderately resistant, and 'Lupa' was highly susceptible to the pathogen.

Thirty plants of each accession were grown in sterile conditions until inoculation (15 days after germination) with a conidial suspension of the isolate AL 84 of A. lentis (ATCC ^® 46981™) following the method described in Ford et al. (1999) . Other 30 plants per accession were grown in the same conditions and similarly treated, but without the presence of the pathogen, to use them as controls. Twenty four hours after inoculation, the aerial parts of the corresponding plants were pooled, frozen in liquid nitrogen, grinded and stored at -80ºC until total RNA extraction was carried out. According to Roundhill et al. (1995) , it is around 24 hours after infection when the first signs of disturbance in the plant cell cytoplasm in reaction to the hyphae penetration are detected. Also at this time the magnitude of oxidative burst seems to be related with the level of the plant resistance ( Sambasivam et al. , 2017 ) and the differences in expression changes among lentil genotypes are more apparent ( Khorramdelazad et al. , 2018 ; Sari et al. , 2018 ). Total RNA was extracted using the method described in Chang et al. (1993) .

The RNA of each corresponding genotype and treatment were pooled separately to produce six different MACE cDNA libraries. The libraries were constructed using the "MACE-Kit" (GenXPro GmbH, Frankfurt am Main, Germany) as described in Zawada et al. (2014) . In short, cDNA was obtained from ≈5 µg of total RNA, then randomly fragmented and biotinylated 3' ends were captured after binding to a streptavidin matrix. A library ready for high-throughput sequencing was prepared using TrueQuant adapters included in the kit. The library consisted of 50-

1000 bp-long fragments derived from the 3'-end of the cDNAs. The 3' ends of the libraries were sequenced on a HiSeq 2000 (Illumina) with 100 cycles to generate the MACE tags, each tag representing one single transcript molecule. The Illumina-derived sequence reads were processed with GenXPro’s in-house MACE analysis pipeline. Briefly, the elimination of duplicate PCR-derived tags, identified by TrueQuant technology, was carried out. Then, low-quality sequences were deleted, and finally the removal of sequencing adapter primers and poly(A)-tails was carried out using an in-house Python script.

Reads from all samples were assembled using the Trinity RNA-Seq de novo Assembly ( Grabherr et al. , 2011 ). The resulting contigs of the assemblies longer than 100 bp were annotated by BLAST (BLAST 2.2.25+) to different sources, using in all cases the default parameters implemented in the NCBI BLAST tool and a threshold e-value of 10 ^-5 . The contigs were compared by BLASTX, first to all Fabaceae Swissprot and afterwards to all Fabaceae Trembl protein sequences from Uniprot ( http://www.uniprot.org/ ). Contigs that did not have a match to one of these databases were annotated by BLASTN to all Lens mRNA sequences from NCBI. Contigs were hereafter again used as reference for annotation and quantification of the MACE tags. Only uniquely mapped reads were taken into account. In order to compare the different samples, a normalization of the raw counts was performed according to Anders & Huber (2010) .

The p -values for differential gene expression be­tween the samples were calculated using the DEGSeq R/Bioconductor package ( Wang et al. , 2010 ). The values for controlling the False Discovery Rate (FDR) were estimated according to Benjamini & Hochberg (1995) . A change in a contig expression was considered statistically significant when the FDR value was ≤ 0.05.

The functional annotation of the sequences was carried out using the MACE2GO software, available at GenXPro's server at http:\\tools.genxpro.net . The enrichment of the gene ontology (GO) categories was based on the genes that showed differential expression at p ≤ 0.05, and the statistical significance was calculated using the Fisher's exact test. The results of the GO enrichment analysis were summarized and plotted using the REVIGO (Reduce plus Visualize Gene Ontology) software that clusters the terms by their relationship and avoids the redundancy, making easier the results interpretation ( Supek et al. , 2011 ).

In order to test the reliability of the MACE results, RT-qPCR was carried out for 18 genes. Three biological and three technical replicates for each genotype and treatment (control and infected) were obtained in the same conditions as in the MACE experiment and used for this analysis. Total RNA was extracted using the same method as in the MACE experiment, and after a DNase digestion, RNA quality and quanti­ty was evaluated by electrophoresis and NanoDrop spectrophotometry. To be sure of the complete absence of DNA contamination, we followed the method described in Vaghefi et al. (2013) . For each sample, 1 µg of RNA was reverse transcribed using the High-Capacity RNA-o-cDNA™ kit (Applied Biosystems™) following the manufacturer's instructions, and the cDNA obtained was diluted 1:10 to be used in the RT-qPCR reactions. Three technical replicates per sample were prepared and each one contained 7.5 µL

of Power SYBR™ Green PCR Master Mix (Applied Biosystems™), 2 µL of cDNA and 0.4 µM of each primer in a total volume of 15 µL. The RT-qPCR was performed in a StepOnePlus Real-Time PCR instrument (Applied Biosystems™) using the follo­wing program: 95 ºC for 10 min, 40 cycles of 95 ºC for 15 s and 60 ºC for 1 min and finally a melting stage from 60 ºC to 95 ºC with increasing steps of

0.3 ºC. The amplification efficiency for each primer pair was estimated by the software LingRegPCR v2017.1 ( Ruijter et al. , 2009 ). The analysis of gene expression changes was carried out with the Comparative C _T (ΔΔC _T ) method implemented in StepOne™ software v2.3 (Life Technologies).

Most of the primers for RT-qPCR were designed from specific MACE contig sequences using the soft­ware Primer3Plus ( Untergasser et al. , 2007 ). For the reference gene, we used the primer pair LcActin-257 described in Sari et al. (2017) because it has been validated in RT-qPCR experiments dealing with Asco­chyta infection in lentil. The information about the primer sequences, PCR efficiencies and foldchanges can be found in the File S1 [suppl.].

Because one of the causes of the limited production of lentil worldwide is the fungal disease caused by A. lentis , we were interested in the analysis of the differences in response to the infection in susceptible and resistant cultivars.

In this study, the changes in the lentil transcriptome after the A. lentis infection were estimated in new lentil materials by using the MACE technique for the first time in this species. This approach has allowed us to obtain high-quality information, not only on the expressed genes in the plant genotypes, but also on their expression levels.

The significant and positive correlation between the expression changes observed in the MACE and RT-qPCR experiments in the three genotypes supports the reliability of the obtained results in the transcriptome profiling using the MACE technique. Moreover, the resolution of the MACE technique is highlighted by the fact that also sequences from the pathogen were identified, although they were in a small proportion in the total RNA sampled from infected plants.

The length of the MACE contigs, mainly between 100 and 1000 bp, was enough to allow the identification of homologous genes in the databases. It is remarkable that a 48 % of the sequences did not match any characterized sequence, maybe due to the fact that MACE mainly covers the 3' end of a transcript that often only contains the 3' UTR, which makes difficult to get a protein annotation.

Even with a limited functional annotation, these data yielded a number of candidate genes involved in the resistance response based on the type of identified genes and the expression changes in the three genotypes analyzed. The key resistance genes are probably distinct in different genotypes, as illustrated by the different genes used in the lentil breeding programs ( Pérez de la Vega et al. , 2011 ; Sudheesh et al. , 2016 ), although some common mechanisms seemed to have a role, such as the activation of the JA pathway ( Sari et al. , 2018 ).

The way in which A. lentis is recognized by the plant and how this fact triggers the defense response is not yet resolved. Although a change in a gene expression after the infection does not necessarily imply a direct role of that gene in the resistance response, when the changes are similar in several resistant genotypes and different from the changes in susceptible ones, it seems logical to consider the gene as involved in the resistance. Based on this rationale this study has pointed out several genes probably involved in the response between the fungus recognition by the plant until the accumulation of antifungal compounds in plant cells.

Similarly, a change in gene expression in the same direction for all genotypes (resistant or susceptible) can be the signature of those genes involved in general mechanisms of response against pathogens, but not the ultimate responsible for the resistance, or alternatively by-products due to the stress. This set of genes included candidate genes previously described as responsible for the resistance, such as some genes coding for PR proteins ( PR4 and PR10a ). The RT-qPCR data did not show a significant change of expression for these genes with the exception of the up-regulation of PR4 in the susceptible genotype (File S9 [suppl.]).

Although these results suggest that these genes are not main candidates for the differences in resistance response among the three genotypes analyzed in this work, it would be possible that the resistance depended not only in the magnitude of the expression change, but also in the absolute quantity of a gene product inside the cell. Besides, we cannot rule out that the resistance response would be determined by the different alleles borne in the three genotypes or by post-transcriptional modifications no detected in the sequences obtained with the MACE technique.

The recognition of the pathogen is usually mediated by the interaction of a pathogen molecule ( i.e. , chitin) by a plant receptor, usually a wall associated receptor-like kinase ( Mengiste, 2012 ). At least 10 contigs in our work belonged to this kind of receptors, but they were in low copy number and their expression changes were not significant. On the other hand, the recognition can also be mediated by interactions between fungal effectors and R-proteins, many of which belong to the NBS-LRR gene family. We found 42 contigs related to NBS-LRR genes, and most of them did not show significant changes after the infection.

After the pathogen recognition, several hormones pathways, mainly SA, ABA and the synergistic JA and ET, would mediate the defense response. Our data point to a high number of contigs matching genes related with those signalling pathways. For instance, 33 contigs in 'ILL5588' contain the GO term 0009751 (response to salicylic acid) and four out of the 33 were down-regulated and 11 up-regulated after infection, or 142 contigs contain the GO term 0006952 (defense response) among which seven were down-regulated and 34 up-regulated (File S6 [suppl.]).

A general view of the processes that change sig­nificantly after the infection was obtained by sum­marizing the enriched GO terms using the software REVIGO (Fig. 3 and File S7 [suppl.]). In 'ILL5588' (moderately resistant), many up-regulated enriched GO terms were related with the plant defense, some were general terms such as "defense response", "response to biotic stimulus" or "response to stress", but others were more specifically related in the response to Ascochyta (defense response to fungus, response to chitin, response to SA, response to JA, response to auxin). Other up-regulated terms seemed to suggest the way in which the plant is controlling the fungus invasion, such as reinforcing the cell wall (lignin biosynthesis) or storing toxic antifungal compounds, most of them derived from flavonoids, isoprenoids, phenylpropanoids or fatty acids ( Amil-Ruiz et al. , 2011 ). One of the highest up-regula­ted enriched terms was "oxidation-reduction process", probably related with the accumulation of reactive oxygen species (ROS) in the early stages of the fungus infection. The role the oxidative burst and the associated cell death is not clear in the necrotrophic pathogenicity, although the general idea is that the early activation of ROS production induces the defense mechanisms, but a sustained production and the associated cell death promotes susceptibility to necrotrophs and resistance to biotrophs ( Mengiste, 2012 ). In lentil, the resistance has been associated to both responses: the inhibition, such as in the cultivar 'CDC Robin' ( Sari et al. , 2017 ) and the promotion, such as in 'ILL7537' ( Khorramdelazad et al. , 2018 ), of the programmed cell death.

In 'ILL5588', the down-regulated enriched terms are mainly related with photosynthesis functions, a usual response after a pathogen invasion ( Huot et al. , 2014 ).

L. c. orientalis shared several up-regulated enriched GO terms with 'ILL5588', and although they were in a lower number and belonged to a more general GO level, some coincidences suggest a common mechanism of defense with 'ILL5588' (lignin biosynthesis, response to JA, flavonoid biosynthesis, or phenylpropanoid bio­synthesis). Many down-regulated terms were related with photosynthesis as found in 'ILL5588'. One of the down-regulated terms (cutin biosynthesis) seems to be directly related with the defense, since cutin is supposed to protect all the aerial parts of the plants. However, although the relationships of cutin with the response mechanisms is proved, its role its controversial: while in tomato the level of cutin in the cuticle is correlated with the resistance to Botrytis cinerea , in Arabidopsis mutants with lower levels of cutin in the leaves showed an increase in the resistance to B. cinerea ( Fich et al. , 2016 ).

A remarkable difference between the susceptible 'Lupa' and the resistant genotypes was the down-regulation of the response to JA and the up-regulation of the genes for gibberellin biosynthesis. Gibberellin promotes the degradation of DELLA proteins, which strengths the SA signaling pathway and weakens the JA pathway, thus increasing the susceptibility to necrotrophs ( Navarro et al. , 2008 ). The role of DELLA proteins in the resistance to A. lentis is supported by Khorramdelazad et al. (2018) , who found an over-expression of the coding genes in the resistant genotype ILL7537, suggesting a main role in the systemic acquired resistance signaling.

Summarizing the results for these three Lens geno­types, it can be concluded that in response to Ascochyta , the reaction by the host plant was multiple and although it presents genotype-associated peculiarities, some com­mon points can be observed. Thus, the JA pathway was activated in resistant genotypes, while it was down-regulated in the susceptible one, and the processes related with the lignin biosynthesis are also induced exclusively in the two resistant genotypes. The response to chitin, the salicylic pathway and the auxin response were activated only in the resistant L. c. culinaris genotype 'ILL5588', while the giberellin synthesis was only induced in the susceptible L. c. culinaris cv. 'Lupa'. Finally, the general defense responses and the metabolism of ROS products were activated in all the three genotypes. Because the hormone pathways are interconnected, either synergism (JA and auxin pathways) or antagonism (JA and SA pathways, SA and auxin pathways) among them, the balance of all signals would be the determinant of the level of resistance.

Since similar results have been previously published in other lentil genotypes ( Khorramdelazad et al. , 2018 ; Sari et al. 2018 ), we can conclude that our results agree with the up-regulation of JA signaling as a general response of the resistance to Ascochyta blight infection in lentil.

The authors would like to thank Dra. Ana Llorente for helping in the greenhouse and laboratory work.