Advanced Breeding for Biotic Stress Resistance in Poplar

Poplar is one of the most important forest trees because of its high economic value. Thanks to the fast-growing rate, easy vegetative propagation and transformation, and availability of genomic resources, poplar has been considered the model species for forest genetics, genomics, and breeding. Being a field-growing tree, poplar is exposed to environmental threats, including biotic stresses that are becoming more intense and diffused because of global warming. Current poplar farming is mainly based on monocultures of a few elite clones and the expensive and long-term conventional breeding programmes of perennial tree species cannot face current climate-change challenges. Consequently, new tools and methods are necessary to reduce the limits of traditional breeding related to the long generation time and to discover new sources of resistance. Recent advances in genomics, marker-assisted selection, genomic prediction, and genome editing offer powerful tools to efficiently exploit the Populus genetic diversity and allow enabling molecular breeding to support accurate early selection, increasing the efficiency, and reducing the time and costs of poplar breeding, that, in turn, will improve our capacity to face or prevent the emergence of new diseases or pests.


Introduction
The term "poplar" identifies a group of about 30 tree species, belonging to the genus Populus, which are traditionally clustered into six sections known as Populus, Tacamahaca, Leucoides, Turanga, Abaso, and Aigeiros [1], although phylogenies based on nuclear rDNA and mitochondrial sequences corroborate the polyphyletic origin of some sections [2]. Comparative genomics is contributing to unravelling the evolution and classification of poplar species, showing that white poplar (P. alba) is genetically close to black cottonwood (P. trichocarpa), although they are currently included in Populus and Tacamahaca sections, respectively [3]. Unlike these clusters, a high level of sequence conservation has been detected between the European aspen (P. tremula) and quacking aspen (P. tremuloides), in agreement with their classification in the same Populus section [4].
In its native habitats, poplar shapes natural forest and woodlands but it is also the most intensively domesticated forest tree species [5]. The genome sequence of black cottonwood was the first one obtained for forest trees [6] and nowadays large genomic resources are available for poplar species. The easiness in transformation, clonal propagation, and the fact that the reproductive stages can be reached in a reasonable short time have fostered the use of poplar as a model species for genetic and genomic studies for forestry and tree breeding [7,8]. The most economically relevant poplar species exploited for breeding and for the generation of commercial hybrids include black poplar (P. nigra) and eastern cottonwood (P. deltoides), both belonging to Aigeiros section, quaking aspen (P. tremuloides), European aspen (P. tremula), and white poplar (P. alba), included in Populus section, black Beyond fungi and insects, bacteria classified in the Erwinia genus, Xanthomonas populi, and Lonsdalea populi cause poplar canker resulting in consistent losses of wood biomass yield. Lastly, leaf viral pathogens (i.e., poplar mosaic virus, poplar decline virus, and arabis mosaic virus) also represent serious threats [15][16][17][21][22][23] (Table 2).
For woody plants, including poplar, biotic stresses are the results of complex biological interactions among hosts, pests and pathogens, and environmental conditions [24]. In both natural populations and specialized poplar plantations, these concurrent combinations of factors may cause a high rate of tree mortality [25]. Climate change is altering the bioclimatic factors that characterize poplar farming areas, and, in general, the growing areas of perennials and agricultural lands, and is expected to enhance the vulnerability to environmental stresses with a reduction of productivity [26]. In crops, several observations indicate that climate-warming drives the movement of pests and diseases poleward [27]. Similarly, current evidence points out that climate change underlies the expansion of the geographic ranges of diseases, affecting forest species towards higher latitudes and the emergence of pathogens in new territories where susceptible plantations are located [28,29]. In poplar, simulation models predict an increase of the average degree of susceptibility to Melampsora larici-populina for the next decades  in New Zealand [30]. Under the new climate scenarios, the understanding of the relationship between changing environmental conditions and the spread of poplar pests and pathogens is pivotal to drive breeding activities and to adopt mitigating policies to tackle or anticipate the emergences of diseases in new territories.
Traditional methods for the prevention of poplar pests and diseases are based on the selection of clones well adapted to soil (pH, salinity, and calcium content) and climatic conditions with some resistance to drought. Well adapted clones can avoid abiotic stresses that can foster the incidence of biotic attacks. Good clonal adaptability coupled to the choice of proper cultural practices (i.e., low plant density, polyclonality, fertilization, and irrigation) and the use of resistant clones represent prevention systems that are commonly adopted in several plantation areas.
The recent advances in genomics, quantitative trait loci (QTLs) mapping, genomic prediction (GP), and genome editing (GE) have the potential to accelerate and improve the traditional poplar breeding, and more in general tree breeding, to cope with old and emerging pests and diseases [31]. In this review, we firstly revised the existing diversity for biotic stress resistance in untapped poplar germplasms and its relevance to support improvement efforts to create new clones resistant to biotic stresses. Secondly, we examined the most recent advancements in genetics, genomics, and biotechnology that are currently used, or might be used, for improving the selection of new poplar clones, highlighting how genomic tools could be deployed for ameliorating and fastening poplar breeding for biotic stress resistance.

Harnessing Poplar Diversity to Improve Pest and Disease Resistance
The narrow genetic diversity of cultivated commercial clones of poplar and the intensive poplar farming adopted worldwide are exacerbating the need of new resistant genotypes. Unlike spontaneous forests, poplar farming is largely based on a few F 1 interspecific clones that have been selected to maximize wood production in specific environments [32]. Moreover, several commercial clones do not often offer appropriate resistance levels to cope with the emergence of pests and diseases moving in new territories as consequence of climate change. This is especially true in the Mediterranean area where a single clone, the Italian selection 'I-214', is the most used particularly in monoclonal stands [16]. For instance, leaf rust can cause up to 60% of yield losses in poplar stands [33], while, as mentioned above, up to 10% reduction of potential production has been reported for WPA [7].
To manage the attack of some of these pathogens and pests (i.e., leaf rust, leaf spot, shoot blight, bacterial canker, and WPA) several breeding programmes targeting biotic stress resistance have been developed worldwide [34][35][36][37][38][39][40][41][42]. All these conventional breeding programmes rely on the availability of germplasm diversity as a source of donor genotypes carrying resistant alleles, which are pivotal to succeed in developing high-performance clones. Consequently, the identification of new sources of resistance to biotic stresses in untapped poplar germplasms and the breeding of a new generation of clones represent the main long-term strategy for coping with the effects of climate change and contrasting the spread of diseases.
Several studies have assessed the existence of genetic diversity for biotic resistance and there is evidence that individuals sampled in natural forests can provide new alleles at resistance loci [43]. The analysis of trait diversity carried out in 47 genotypes of P. nigra collected in several location in Europe has shown various levels of resistance against S. musiva [43]. The same pattern has been substantiated in eastern cottonwood (P. deltoides), for which a correlation between genomic diversity and resistance to different lineages of S. musiva has been highlighted [44]. Following the same research line, it has been shown that trees sampled from wild populations of black poplar (P. nigra) exhibit genetic variation for resistance against M. larici-populina [45], and the identification of two highly resistant trees corroborates the finding that untapped genetic materials sampled in natural forests might be relevant sources of resistance to sustain poplar breeding programmes [46].
While large studies have pointed out diversity for resistance to biotic stresses, the bioclimatic parameters that characterize the geographic origin of poplar germplasm might be useful to carry out initial screening as there is a high chance to identify resistant trees in areas subjected to the natural occurrence of pathogens. For instance, eastern cottonwood and black cottonwood (P. trichocarpa) trees originating from more humid environments exhibit, in general, higher levels of leaf rust resistance, compared to trees sampled in drier environments, which have evolved under weaker pathogen selection intensity [47]. This approach has been largely substantiated for cereal crops [48] and might be used to mine large poplar collections and assemble diversity panels to improve biotic stress resistance.
The effective exploitation of untapped poplar germplasms requires field trials and rapid assays to discover and evaluate resistant alleles effective against different pathogen races [34]. Typically, these assays include tests carried out in controlled conditions (i.e., greenhouses or growth chambers) or in laboratories. Moreover, the reliability of these assays should depend as less as possible on testing conditions or tree age and dimensions. In poplar, the expression of resistance against fungal pathogens often occurs during juvenile stages and has fostered the development of phenotyping methodologies in controlled conditions, facilitating breeders' work [49]. Assays for evaluating resistance against M. brunnea and X. populi have been developed and, currently, allow early and accurate selection of resistant poplars at juvenile stages [49]. A better control of the environmental factors in the juvenile evaluation improves the efficiency of early selection for Melampsora species [49], for which the resistance is often evaluated on one-year-old plants in controlled conditions [50]. Other assays for testing the resistance to M. larici-populina in laboratory conditions are based on the scoring for the appearance or absence of fungal spores in poplar leaf disks upon artificial inoculation [51]. Interestingly, the estimation of V. populina resistance in controlled conditions has been also validated using randomized clonal trials organized in sites where this pathogen severely attacks poplar plantations [52].
Breeding for insect resistance exacerbates the need of solid phenotyping methodologies and testing assays as insects are known to exhibit preferences towards certain individuals in monoclonal plantations [34], hampering the identification of resistant genotypes. Nevertheless, at least for three different species (WPA, G. aceriana, and S. salicis), methods for assessing resistances in controlled conditions have been developed [34], and, particularly, the test for evaluating the susceptibility to WPA is currently used for P. ×canadiensis breeding programmes and for genetic studies [7,34,49].
To fully harness advantageous alleles, poplar germplasms exhibiting the relevant biotic resistance must enter in actual breeding programmes, which are mainly based on hybrid breeding to systematically exploit heterosis [8,36,49,53]. In Europe, the most successful poplar clones belong to P. ×canadiensis, which derives from enforced crossing between eastern cottonwood and black poplar, P. ×wettsteinii originated by inter-specific crosses between European aspen and quacking aspen, and P. ×tomentosa, obtained by crossing white poplar and European aspen (Table 3). Table 3. List of the hybrid clones obtained in poplar breeding programmes.

Hybrid Cross
P. ×canadiensis P. deltoides × P. nigra P. ×generosa P. trichocarpa × P. deltoides P. ×tomentosa P. alba × P. tremula P. ×wettsteinii P. tremula × P. tremuloides P.×interamericana P. deltoides × P. trichocarpa The exploitation of heterosis in hybrid breeding usually requires the development of heterotic groups, including lines that may or may not be genetically related and that can exhibit heterosis in F 1 generation. In the context of poplar breeding, heterotic groups coincide with the different sets of parental lines belonging to different species, making F 1 inter-specific hybrids the most effective strategy to maximize heterosis [6,34,52,54]. Using both field trials and assays in controlled conditions, the public breeding programme of poplar carried out in Italy has applied semi-reciprocal recurrent selection to develop several groups of parental lines of eastern cottonwood and black poplar, which show improved resistance to WPA for eastern cottonwood and resistance to WPA, Marssonina, Venturia, and Melampsora species for black poplar [8,36,53].

Genome Sequences Are Pivotal to Improve Poplar Breeding for Biotic Stress Resistance
The genome sequence of black cottonwood is a landmark for plant science as it was the first sequenced genome of a perennial tree species [6]. The assembly statistics shows that this reference sequence spans about 550 Mb, is organized in 19 chromosomes, and, as for other outcrossing species, is highly heterozygous [6,55] (Table 4).

Species
Overall, more than 45,000 protein-coding genes were annotated on poplar genomes and their functional characterization represents a powerful tool to help dissecting the genetic bases of biotic stress resistance and accelerating poplar breeding [4,59]. For instance, the gene ontology (GO) classification carried out on European aspen and quacking aspen genome sequences allowed the identification of genes putatively involved in disease resistance [4]. In a similar way, the genome sequence of white poplar revealed a loss of disease-resistance genes [3].
A deeper assessment of the genetic diversity in poplar has been achieved through the analysis of pan-genome, which has allowed to detect structural variants, including insertion/deletions (INDELs) and copy number variations (CNVs), in closely related poplar species (P. nigra, P. deltoides, and P. trichocarpa) [62]. Interestingly, the GO classification carried out on genes exhibiting copy number variants shows that many of them are implicated in resistance to stresses and diseases [62].
Poplar genomic assemblies have enabled comparative in silico analyses between unrelated species, demonstrating great potential for unravelling the functional role of genes. By comparing the genome sequences of poplar and the model plant Arabidopsis thaliana, several candidate genes associated to disease resistance were identified [14]. For instance, the genome assembly of black cottonwood was mined using the sequences of 15 Arabidopsis MLO proteins, associated to susceptibility to powdery mildew, allowing the mapping of 26 MLO genes on 14 black cottonwood chromosomes. On the base of Arabidopsis-poplar phylogeny, four of these genes were considered as potential candidates involved in poplar-powdery mildew resistance, paving the way for future studies aimed at managing the spread of this disease in poplar plantations [63]. Following the same approach, a genome-wide analysis of WRKY transcription factors encoding genes on poplar genomes led to the identification of 100 unique genes, 61 of which displaying modulated expression in response to biotic stresses [64]. In addition, Hidden Markov Models (HMMs) were applied for a comprehensive genome-wide analysis of pentatricoptide repeat (PPR) genes on black cottonwood genome and 626 of such genes were discovered, 154 of which are modulated in response to abiotic stresses and M. brunnea attack [65].
Beyond comparative analyses, genome sequences are enabling the identification and mapping of molecular markers, which, in turn, contribute to accelerating QTL mapping and Genome Wide Association Scan (GWAS) applications for the identification of loci underlying biotic stress resistance. The analysis of the genetic diversity carried out in a panel of 1038 black cottonwood trees, mostly sampled across their natural range, identified more than 7.4 M genetic variants including rare single nucleotide polymorphisms (SNPs) and INDELs [66]. Similarly, targeted resequencing of 579 eastern cottonwood individuals, sampled across populations spanning 15 states in the USA, allowed to detect more than 500k SNPs [67].
Exploiting available genome sequences, three SNP arrays were developed for highthroughput genotyping of poplar: the Illumina iSelect Infinium 34K array interrogates 34,131 SNPs derived from the resequencing of 34 wild black cottonwood accessions and located within or close to 3543 genes [68]; the 12K Infinium array was designed using SNPs identified by resequencing 51 black poplar trees and includes markers within or close to candidate genes for rust resistance, wood properties, water-use efficiency, and bud phenology [69]; and, lastly, a new Affimetrix multispecies SNP array (4TREE array) has been developed within the EU-funded H2020 B4EST project (https://b4est.eu/ (accessed on 23 March 2022)) and includes 13,400 poplar SNPs obtained from the sequencing data of 90 black poplar genotypes and 30 eastern cottonwood trees, respectively (data not published). Although the availability of poplar SNP arrays offers unprecedented possibilities for enabling genetic studies and improving poplar breeding, arrays are optimized for the species used for their development and cover only a fraction of the existing genetic variability within a species [70]. Recently, exome sequencing and protocols based on reduced-representation sequencing, such as Genotyping-by-Sequencing (GbS) or singleprimer enrichment technology (SPET), have been used to fingerprint poplar populations and overcome the limitations of SNP ascertainment bias in genotyping arrays [7,67,71,72].

An Overview of Known Loci Co-Segregating with Biotic Stress Resistance in Poplar
The most studied poplar disease is leaf rust and MXC3 is one of the first QTLs cosegregating with the resistance against the leaf rust agent Melampsora × columbiana and was mapped in a high-resolution local genetic map using Bulk Segregant Analysis (BSA) and 19 Amplified Fragment Length Polymorphism (AFLP) markers [13]. The sequences of these markers were used to screen a bacterial artificial chromosome (BAC) library of black cottonwood and positive BAC clones were subsequently assembled in a large contig, which was further extended using chromosome walking. The resulting physical map was then integrated with a high-resolution genetic map but, owing to the low recombination frequency, MXC3 was not positionally cloned [13] (Table 5). Table 5. Summary of poplar QTLs associated to pest and disease resistance. For each QTL, the pathogen, the genetic resources used for the identification, the name, the LG/chromosome, the physical or genetic position, the peak marker/markers, and the candidate genes are reported when indicated in the corresponding reference/references. For MXC3 and MER [73], the QTLs associated to the response to insects [74], ORPM_26, and WPMS_15-PMGC_2839 [75], and qMLSD-VI-1 and qMLSD-XVI-2 [76] the physical positions have been determined by blasting the sequences of the corresponding primers on P. trichocarpa genome (https://phytozome-next.jgi.doe.gov/info/Ptrichocarpa_v4_1 (accessed on 21 March 2022)).      A second QTL for resistance against the leaf rust caused by M. larici-populina is the MER locus, which was shown to co-segregate with 11 AFLPs. These markers allowed the identification of 17 recombinants from 512 progenies of three interspecific crosses between eastern cottonwood and black poplar, using the same eastern cottonwood resistant parent, and a high-resolution local genetic map, covering 3.4 cM including the MER locus, was obtained. The sequencing of the AFLPs revealed similarities with nucleotide binding sites leucine-rich repeat (NBS-LRR) Resistance ®genes for three of them. Like for MXC3, MER was not positionally cloned (Table 5) [80].
Using a two-way pseudo-testcross mapping strategy, a genetic map highly saturated with AFLP and Simple Sequence Repeat (SSR) markers was constructed from 171 individuals of a F 1 population, derived from the cross between one hybrid clone (P. trichocarpa, 93-968 × P. deltoides, ILL-101) and one eastern cottonwood genotype (P. deltoides clone D109). This resource was used to map both MXC3 and MER loci and markers co-segregating with these loci were subsequently used to screen and assembly shot-gun sequence data form draft Populus genome scaffolds, generated in the pre-genomic era, to identify candidate genes for MXC3 [73] (Table 5 and Figure 1). A second QTL for resistance against the leaf rust caused by M. larici-populina is the MER locus, which was shown to co-segregate with 11 AFLPs. These markers allowed the identification of 17 recombinants from 512 progenies of three interspecific crosses between eastern cottonwood and black poplar, using the same eastern cottonwood resistant parent, and a high-resolution local genetic map, covering 3.4 cM including the MER locus, was obtained. The sequencing of the AFLPs revealed similarities with nucleotide binding sites leucine-rich repeat (NBS-LRR) Resistance ® genes for three of them. Like for MXC3, MER was not positionally cloned (Table 5) [80].
Using a two-way pseudo-testcross mapping strategy, a genetic map highly saturated with AFLP and Simple Sequence Repeat (SSR) markers was constructed from 171 individuals of a F1 population, derived from the cross between one hybrid clone (P. trichocarpa, 93-968 × P. deltoides, ILL-101) and one eastern cottonwood genotype (P. deltoides clone D109). This resource was used to map both MXC3 and MER loci and markers co-segregating with these loci were subsequently used to screen and assembly shot-gun sequence data form draft Populus genome scaffolds, generated in the pre-genomic era, to identify candidate genes for MXC3 [73] (Table 5 and Figure 1).

Figure 1.
Projection of known loci co-segregating with resistance to pests and diseases on the reference sequence of P. trichocarpa. For MXC3, MER [73], ORPM_26, WPMS_15-PMGC_2839 [75], qMLSD-VI-1, and qMLSD-XVI-2 [76], the physical position has been determined blasting the sequences of the corresponding primers on the reference genome. Loci co-segregating with the resistance to insect, leaf rust, Sphaerulina musiva, WPA, Schizoempodium mesophyllincola, Marsonnina brunnea, and other biotic stresses are reported in brown, orange, light green, blue, purple, red, and light blue, respectively. The 19 chromosomes of P. trichocarpa are coloured according to the density of annotated genes. A qualitative resistance locus (R 1 ), effective towards four leaf rust strains, and nine leaf rust resistance QTLs, two of which (including R US ) with broad-spectrum effects, have been mapped deploying 389 markers and a F 1 segregating population of 343 individuals, derived from an interspecific cross of P. deltoides × P. trichocarpa [81] (Table 5). Fine-mapping of R 1 and R US , conducted enlarging the interspecific population of P. deltoides × P. trichocarpa populations [81] to 1415 individuals, genotyped by fragment-based molecular markers linked to resistance, localized the two loci on chromosome 19 (Table 5). Physical maps for the two dominant and recessive R US haplotypes were generated from the P. trichocarpa parent, heterozygote at the locus. The genetic and physical maps were then anchored to the genome sequence, discovering clusters of NBS-LRR and serine-threonine kinasis genes for R 1 , and NBS-LRR genes for R US [82] (Table 5 and Figure 1). This latter example shows how the availability of poplar genome assemblies has drastically improved QTL mapping, leading to the anchoring on chromosomes and fostering the identification of candidate genes inside the confidence interval of QTLs, without the need of tedious and expensive positional cloning or physical mapping.
Besides leaf rust, candidate genes for resistance to other pests were identified combining mapping and genomic data. Two major leaf spot resistance QTLs were mapped using a P. deltoides F 1 population of 84 individuals, genotyped with 1398 AFLPs and 72 SSRs, and the sequences of molecular markers in linkage with this trait were aligned to poplar genome [76] (Table 5 and Figure 1).
To identify common pathways of response to insect attack, a linkage map was built from 350 individuals of an interspecific cross, genotyped using fragment-based molecular markers, and assessed for the response to seven categories of insects causing leaf damage at two time points [74]. Fourteen genomic regions on nine linkage groups (LGs) correlated with plant/insect interaction and a three-step approach to combine QTL mapping and genomic information was then applied to find out candidate genes and metabolic mechanisms associated to insect response. This approach was based on: (1) searching for co-location with genes implicated in the shikimate-phenylpropanoid pathway; (2) searching for colocation with QTLs controlling leaf traits; and (3) functional classification through GO enrichment analyses of the genes detected at the level of the QTL confidence intervals [74]. As a result, similar response to different insects was discovered and 15 genes implicated in the production of phenolic glycoside were mapped within 9 QTLs, providing new understandings of the interactions between poplar and insects to be exploited in breeding programmes [74] (Table 5 and Figure 1).
The molecular bases of WPA resistance were investigated using 131 individuals of a F 1 segregating population derived from the cross between a WPA resistant eastern cottonwood and a susceptible black poplar. This interspecific population was genotyped through GbS and available SSRs, leading to 5667 polymorphic markers that were used to create highresolution maps for the parental lines. One major and two minor QTLs co-segregating with WPA resistance and explaining more than 65.8% of the phenotypic variance were mapped on three different LGs and candidate genes were identified [7] (Table 5 and Figure 1). GbS was also used for fingerprinting an interspecific F 1 population of 300 trees derived from enforced crossing of the leaf rust resistant eastern cottonwood clone I-69 and a susceptible P. simonii genotype. Genetic maps were created using 1222 polymorphic SNPs and 11 rust resistance QTLs were mapped on nine different chromosomes (Table 5 and Figure 1). R gene clusters were identified withing the confident intervals of two QTLs (Table 5) and a 611-bp deletion associated to variation in rust resistance was discovered in a R gene, providing a marker to develop molecular diagnostic tools for rust resistance [84].

Transcriptomics for Disease Resistance in Poplar
Transcriptomics is a valuable tool to compare the response to pathogen or pest attacks in closely related species or varieties of the same species, providing a snapshot of the molecular mechanisms activated or repressed during infections. This brings to the prediction of pathways and candidate genes implicated in the response to pathogens that might be useful in breeding programmes [96]. In poplar, transcriptional changes and differentially expressed genes (DEGs) upon attacks of M. medusae, M. larici-populina, and M. brunnea were initially analysed using expression microarrays in P. ×canadiensis, P. ×generosa, and P. deltoides, respectively [97][98][99][100].
The introduction of Next Generation Sequencing (NGS) for studying gene expression through RNA-Seq is generating comprehensive transcriptome datasets, providing the expression level, sequence variations, and transcriptional structure (i.e., alternative splicing) of each gene [101]. RNA-Seq was applied to analyse the response to S. musiva of two resistant and two susceptible poplar clones and pathways implicated in the resistance (oxidation-reduction, protein fate, secondary metabolism, and accumulation of defence-related gene products) or in susceptibility (hypersensitive response loci) were discovered [102]. Likewise, a comprehensive transcriptomic experiment was conducted to investigate the response to the transition from the biotrophic to the necrotrophic phase of the hemibiotrophic fungus M. brunnea of two susceptible poplar species, P. deltoides and P. alba × P. alba var. pyramidalis, at three critical time-points. Pathways activated by infections were discovered and differences were observed during the progress of infection and among the two genotypes: P. deltoides is more responsive to initial attack, while the main interaction between P. alba × P. alba var. pyramidalis and the fungus occurs at the necrotrophic phase. This information is potentially used to guide the development of poplar resistant varieties to leaf spot [103].
The transcriptome of eight different poplar clones belonging to three different species was analysed in response to the attack of herbivory insects and allowed the identification of common herbivory-induced genes and signalling pathways, which represent potential regulators of poplar response to insects [104].
RNA-Seq can also be directed for monitoring the expression of microRNAs (miRNAs). This application was used to study compatible and incompatible interactions between P. szechuanica and avirulent and virulent isolates of M. larici-populina: miRNAs differentially expressed in the two interactions, including miRNAs regulating disease-resistance genes, kinases, and transcription factors that, probably, contribute to mount the response to rust infection, were discovered [105].

Integration of Transcriptomic and Genomic Data
The integration of QTL mapping or GWAS with transcriptomic analyses has a great potential for sustaining breeding as it allows to better prioritize candidate genes underlying specific traits. Following this approach, leaf rust resistance loci R 1 , R US [81,82] , and MER [73], previously mapped on chromosome 19, were further dissected using DEGs located on the same target genomic regions [83] ( Table 5). The sequences of the molecular markers flanking these three loci were aligned to poplar reference sequence to obtain their corresponding physical positions (Table 5 and Figure 1). The orthologous sequences of eastern cottonwood, which was the donor of R 1 and MER resistance, were assembled and candidate R genes were functionally annotated by BLAST searches on available databases. This analysis was integrated with an RNA-Seq experiment to compare the transcriptional response to leaf rust, at two time-points from the inoculum, studying two eastern cottonwood genotypes displaying opposite behaviour to infection (highly resistant T-120 and highly susceptible D-896). Three candidate R genes (one located on R US and two inside MER) were identified among the DEGs mapped within the target loci (Table 5) and were analysed for sequence variations in the contrasting genotypes in order to characterize haplotypic variants that could represent tools for marker-assisted selection (MAS, the use of molecular markers to select plants with better performance) [83].
Like for leaf rust resistance, a GWAS aimed at identifying genes underlying resistance to S. musiva was carried out using a panel of 1000 black cottonwood trees and allowed the detection of nine associations, six of which located at the level of genes encoding for receptors. To better analyse the putative role of these genes in the fungal response, RNAs from resistant and susceptible genotypes, collected at three time-points after infection, were sequenced and DEGs located within the associated loci and showing expression patterns compatible with resistance or susceptibility were discovered [86] (Table 5). Validations by binding assays and in vivo over-expression were also performed [86].
Recently, an RNA-Seq experiment carried out on P. tomentosa LM50 upon inoculation with M. brunnea, at three time-points, was integrated with a multigene association analysis conducted on a panel of 435 unrelated P. tomentosa individuals. This germplasm collection was high-throughput genotyped using 29,399 SNPs obtained from direct sequencing and filtered for the presence on the DEGs discovered by the RNA-Seq experiment as being implicated in M. brunnea response. Following this approach, it was possible to find out many key genes implicated in poplar response to M. brunnea with potential roles in regulating photosynthesis and plant growth, highlighting the genetic interaction between different pathways during pathogen attack [106].
A different approach is based on the conjugation between GWAS and the identification of expression-QTL (eQTLs). eQTLs are DNA variants that contribute to variation in the expression levels of a gene in different genotypes and can represent cisor trans-regulatory elements responsible of differences in the expression of a phenotype [107]. They can be detected using transcriptomic techniques in association with mapping methods that calculate the linkage between variation in expression and genetic polymorphisms. Following this approach, the poplar PtHCT2 gene, encoding for a hydroxycinnamoyl-CoA:shikimate hydroxycinnamoyl transferase 2 (Potri.018G105500), was associated to the response to biotic stresses (Table 5 and Figure 1) [103]. In more detail, a GWAS was conducted on 917 P. trichocarpa accessions, genotyped by resequencing, and regions associated to the accumulation of secondary metabolites, including the PtHCT2 region, were discovered. To confirm the implication of this gene in the antioxidant response, an RNA-Seq experiment was conducted on leaves and xylem of 390 and 444 accessions, respectively. Sequencing results were used to perform correlations with three secondary metabolite abundances and the expression of nine PtHCT genes. Significant associations were discovered only for PtHCT2. An eQTL mapping was then performed using the transcript abundance of PtHCT genes, as the phenotypic variable, in the GWAS experiment. Interestingly, a cis-eQTL was discovered regulating PtHCT2 in both leaves and xylem and two SNPs in this region affected the core of a W-box element, the binding site for WRKY transcription factors, known to be involved in defence response. This association was also confirmed by transient expression and co-expression networks, developed from the RNA-Seq data described by Muchero et al. [86] that revealed that PtHCT2 responds to S. musiva [103].

Breeding for Biotic Stresses in Poplar: Future Perspective
In recent years, genetic studies have allowed to dissect the genetic bases of biotic resistances in poplar and this information has the potential to be exploited for tree breeding, particularly for implementing MAS and, consequently, marker-assisted breeding (MAB). Under certain scenarios, it was shown that MAS can be used to replace or improve truncation selection based on phenotypic values [108] and has the potential to accelerate the identification of the best individuals in a segregating population or germplasm collection, reducing the time and costs of breeding ( Figure 2) [109].
truncation selection based on phenotypic values [108] and has the potential to accelerate the identification of the best individuals in a segregating population or germplasm collection, reducing the time and costs of breeding ( Figure 2) [109]. The knowledge acquired in QTL mapping and GWAS shows that, in several cases, a few QTLs or QTNs with large effects underlie the genetic bases of biotic stress resistance [7]. Moreover, high-density DNA arrays and NGS, coupled with methods based on reduced-representation sequencing, have drastically cut down genotyping costs in poplar, as they have for other species [68,69,72]. Both factors are fostering the use of MAS in poplar breeding as, even though the bottleneck related to the long juvenile phases of perennial plants remains, it allows early selection of a limited number of individuals carrying the desired traits to bring to sexual maturity, thus reducing the spaces and costs for field trials [110]. While MAS offers undoubtedly advantages to accelerate selection, its application in plant breeding requires to appropriately choose and validate the QTLs or QTNs that underlie target traits, which implies testing QTLs/QTNs over different site-by-season combinations or in different backgrounds as their effects might disappear in other pedigrees [111]. Currently, a large fraction of QTLs for biotic stress resistance identified in poplar have not been yet validated, posing further challenges to their exploitation in poplar breeding.
The application of GP, which aims to estimate the breeding value of individuals using only genotypic information [112], for the improvement of poplar is still in its infancy. Nevertheless, the lesson learned from other crops and simulation studies points out that this methodology might be a game-changer for accelerating tree breeding as it allows increasing the genetic gain per unit time and improving selection accuracy [113]. While simulation studies applied to woody perennials corroborate the advantage of GP for the simultaneous improvement of thousands of minor-effect loci [114], its application for The knowledge acquired in QTL mapping and GWAS shows that, in several cases, a few QTLs or QTNs with large effects underlie the genetic bases of biotic stress resistance [7]. Moreover, high-density DNA arrays and NGS, coupled with methods based on reducedrepresentation sequencing, have drastically cut down genotyping costs in poplar, as they have for other species [68,69,72]. Both factors are fostering the use of MAS in poplar breeding as, even though the bottleneck related to the long juvenile phases of perennial plants remains, it allows early selection of a limited number of individuals carrying the desired traits to bring to sexual maturity, thus reducing the spaces and costs for field trials [110]. While MAS offers undoubtedly advantages to accelerate selection, its application in plant breeding requires to appropriately choose and validate the QTLs or QTNs that underlie target traits, which implies testing QTLs/QTNs over different site-by-season combinations or in different backgrounds as their effects might disappear in other pedigrees [111]. Currently, a large fraction of QTLs for biotic stress resistance identified in poplar have not been yet validated, posing further challenges to their exploitation in poplar breeding.
The application of GP, which aims to estimate the breeding value of individuals using only genotypic information [112], for the improvement of poplar is still in its infancy. Nevertheless, the lesson learned from other crops and simulation studies points out that this methodology might be a game-changer for accelerating tree breeding as it allows increasing the genetic gain per unit time and improving selection accuracy [113]. While simulation studies applied to woody perennials corroborate the advantage of GP for the simultaneous improvement of thousands of minor-effect loci [114], its application for predicting biotic stress resistance in poplar might not outperform MAS as GWAS and QTL mapping studies showed that, in many cases, few loci explain a large fraction of the phenotypic variation. Large-scale simulation studies calibrated on poplar might provide evidence of advantages and limitations of GP for breeding for biotic stress resistance under different scenarios.
Alternative ways to accelerate the genetic gain in poplar breeding are based on genetic modification. Recently, genetic-engineering applications in poplar have been reviewed and include studies related to the generation of transgenic poplars resistant to diseases by expressing antibacterial and antifungal genes (i.e., the ones encoding for osmotin, glucanases, chitinases, lysozyme, and thaumatin) or insect R genes (CRY and genes encoding for proteinase inhibitors) [115]. Even though the production of transgenic plants is a faster and feasible alternative to traditional breeding, it is strongly influenced by the random insertion of the transgene that can result in inter-transformant variations in gene expression, due to the integration in genomic regions with low transcriptional activity, epigenetic control, or sequence-specific gene silencing [116]. Moreover, the use of genetically modified organisms (GMOs) is limited by environmental concerns and regulatory issues that are particularly strict for tree species [117].
The development of the new GE technologies based on CRISPR (Cluster Regularly Interspersed Short Palindromic Repeats)/Cas is revolutionizing the concept of genome modification because allows the introduction of insertions, deletions, or the substitutions of single base or short sequences at specific sites in a target genome in highly efficient, simple, and versatile way [118][119][120][121][122] (Figure 2).
Even if GE in poplar is mainly focused for the improvement of growth in plantations and the use in the pulp and bio-refinery industries [123,124], applications to understand the mechanisms involved in poplar disease resistance have been reported: the CRISPR/Cas9mediated knockout of two WRKY genes, WRKY18 and WRKY35, contributed to unravelling their involvement in P. tomentosa resistance to Melampsora [125]; through molecular, genetic, and biochemical approaches, including inactivation by CRISPR/Cas9, the MYB115 transcription factor has been demonstrated to regulate fungal resistance in P. tomentosa by activating the biosynthesis of proanthocyanidins [126]. These examples highlight how CRISPR/Cas9 can offer a great contribution to the discovery and validation of genes underlying poplar resistance to pathogens and represent the most promising frontier for the fast and efficient generation of genome-edited resistant poplar plants, by strengthening endogenous defences through the replacement of native unfunctional alleles or promoter regions, leading to high level of expression of functional genes [124,127].

Conclusions: Are Current Knowledge on Poplar Sufficient to Support Advanced Breeding for Biotic Stress Resistance?
Biotic stresses represent major threats in both natural populations and plantations of poplar and forest trees in general. The extent of biotic stresses is intensifying as global warming is rapidly enlarging the distribution and incidence of pests and pathogens towards higher latitudes and their occurrence in new territories [28,29]. For this reason, the development of new tools to speed up poplar breeding for stress resistance/resilience is pivotal to contrast the spread of diseases.
One of the main issues of poplar breeding is tied to the biology of poplar that, as for other perennials, displays late expression of relevant traits and late development of reproductive structures, characteristics that prolong selection cycles and, consequently, the duration of breeding programmes. In addition, long breeding cycles should be supported by long-term financial investments. Nonetheless, market needs can change over time and the new clones developed by traditional breeding might not guarantee the expected return investments.
The usual intensive cultivation of monoclonal plantations of a small number of elite cultivars, strongly adapted to the pedo-climatic conditions of specific territories, is often criticized by environmental associations, more favourable to greater biodiversity as in natural forests, and restrictions have been posed also on the cultivation of hybrid clones in protected areas and bio-conservation sites along rivers. In this condition, selected clones of native species, alone or in mixed clones, could represent a useful alternative to interspecific hybrids in more sustainable poplar plantations. For this reason, it is necessary targeting poplar genetics to guarantee the sustainability of wood productions in the areas where the spread and emergence of pests and pathogens are expected, also because of the higher frequency and incidence of abiotic stress conditions. It is also important to improve the annual rate of genetic gain in poplar breeding programmes using fast and precise applications and exploiting untapped sources of genetic variability, taking advantage of the large poplar genetic variation worldwide and creating large genetic panels that should be shared between research groups from different countries and tested in multiple environments. The molecular characterization of these collections, in combination to common garden phenotypic data, can contribute to the unlocking of this genetic diversity available to breeders and allows the acquisition of new knowledge (i.e., multi-trait associations, specific adaptive profiles, gene pools, and mixture of genotypes) to be integrated into breeding and deployment strategies. The development and use of more diverse genotypes and the maintenance of large well-adapted populations will result in a significant increase in frequency of potentially beneficial allelic variants.
The annotated poplar genomes and the new genotyping platforms have largely contributed to the mining of the interaction between poplar and pests or pathogens, but the current knowledge is still insufficient to face climate-change challenges. For example, because of the rapid LD decay measured for poplar, many molecular markers are needed to capture the effects of all QTLs. Moreover, strong LD between markers and causative variants that control the traits of interest is desirable to achieve high values of predictive ability in GP. The available SNP arrays, developed using candidate-genomic-region approaches [68,69], represent valuable tools for genetic studies but do not provide the sufficient marker density to uncover all the genetic determinants that underly a specific trait, especially the genetic variants with small effects on phenotype. The high-processive and cost-effective NGS platforms (i.e., Nova-Seq and Illumina HiSeq 4000) could contribute increasing the panel of molecular markers (including SNPs, INDELs, or CNVs) to be used in next-generation studies.
The understanding of the relationship between changing environment and the spread of poplar diseases, using accurate prediction models of the effects of climate change on poplar plantations, is pivotal to define new breeding objectives based on the enlargement of the genetic backgrounds in parentals for the generation and selection of new clones with large biodiversity and adaptability to changing environments.
Finally, the activation of new cooperation projects, involving geneticists, eco-physiologists, wood technologists, and stakeholders, as poplar growers and wood users, would lead to the definition of short-and medium-term objectives for the development of innovative genetic materials with wide variability to cope with future highly uncertain environments and for the definition of cost-and time-effective strategies that meet rapid changes even considering the ecological and economic contexts.

Conflicts of Interest:
The authors declare no conflict of interest.