Transcript Profiling of Poplar Leaves upon Infection with Compatible and Incompatible Strains of the Foliar Rust Melampsora larici-populina^,

Abstract

To understand key processes governing defense mechanisms in poplar (Populus spp.) upon infection with the rust fungus Melampsora larici-populina, we used combined histological and molecular techniques to describe the infection of Populus trichocarpa × Populus deltoides ‘Beaupré’ leaves by compatible and incompatible fungal strains. Striking differences in host-tissue infection were observed after 48-h postinoculation (hpi) between compatible and incompatible interactions. No reactive oxygen species production could be detected at infection sites, while a strong accumulation of monolignols occurred in the incompatible interaction after 48 hpi, indicating a late plant response once the fungus already penetrated host cells to form haustorial infection structures. P. trichocarpa whole-genome expression oligoarrays and sequencing of cDNAs were used to determine changes in gene expression in both interactions at 48 hpi. Temporal expression profiling of infection-regulated transcripts was further compared by cDNA arrays and reverse transcription-quantitative polymerase chain reaction. Among 1,730 significantly differentially expressed transcripts in the incompatible interaction, 150 showed an increase in concentration ≥3-fold, whereas 62 were decreased by ≥3-fold. Regulated transcripts corresponded to known genes targeted by R genes in plant pathosystems, such as inositol-3-P synthase, glutathione S-transferases, and pathogenesis-related proteins. However, the transcript showing the highest rust-induced up-regulation encodes a putative secreted protein with no known function. In contrast, only a few transcripts showed an altered expression in the compatible interaction, suggesting a delay in defense response between incompatible and compatible interactions in poplar. This comprehensive analysis of early molecular responses of poplar to M. larici-populina infection identified key genes that likely contain the fungus proliferation in planta.

Plants respond to microbial invasion by activating an array of inducible defense mechanisms (Nimchuk et al., 2003). After specific recognition of a pathogen, the hypersensitive response (HR) is a rapid and efficient plant resistance mechanism leading to cell death at the site of infection (Heath, 2000). Among the rapid defense mechanisms triggered in plant tissues are generation of reactive oxygen species (ROS) at the site of infection, cell wall thickening, and production of anti-microbial compounds and enzyme inhibitors (Glazebrook, 2005). Genes encoding pathogenesis-related (PR) proteins such as glucanases and chitinases are primary target genes triggered during the early response to pathogen attack (Van Loon and Van Strien, 1999) and are considered as a signature of the HR. Their expression could be directly targeted by pathogen-sensing systems through highly complex and interconnected networks of transduction pathways driving plant resistance (Katagiri, 2004). Several signal molecules, such as ethylene, salicylic acid (SA), or jasmonic acid, play an important role in these defense reaction-signaling networks (Shah, 2003), and recently a central role for the NONEXPRESSOR OF PR GENES1 (NPR1) protein has been highlighted (Dong, 2004). Considerable advances have been made in understanding plant resistance processes in the model plant Arabidopsis (Arabidopsis thaliana), particularly through extensive mutant analyses. It is now clear that the plant response to pathogen infection is associated with massive changes in gene expression (Schenk et al., 2000). Large-scale mRNA expression profilings have revealed that plants express similar sets of defense mechanisms in response to different pathogens (Tao et al., 2003; Eulgem, 2004). Dissection of the defense response at the molecular level has greatly helped in drawing a general model for pathogen resistance in plants (Nimchuk et al., 2003; Tao et al., 2003). However, it is not known whether such a model applies to long-term adaptation of resistance mechanisms in perennial plant species like trees. Such long-lived species are more prone to attacks by pathogens before reproduction, and their long generation time makes it impossible for them to match the evolutionary rates of a pathogen that goes through several generations every year. Annotation of the Populus trichocarpa genome has revealed an expansion of the NBS-Leu-rich repeat (LRR) resistance gene families (Tuskan et al., 2006), suggesting a possible adaptation to long-term exposure to pathogens.

The basidiomycete Melampsora larici-populina is responsible for the leaf rust disease in Populus species (Frey et al., 2005; Pinon and Frey, 2005). Urediniospore germlings of this obligate biotrophic fungus usually penetrate the host plant through stomatal openings, differentiate a series of infection structures in the intercellular space, and exhibit highly localized penetration of the host cell wall to establish a haustorium (Laurans and Pilate, 1999). Hyphae then proliferate in the leaf parenchyma and produce golden pustules filled with masses of urediniospores on the lower leaf surface. M. larici-populina causes severe economic losses in European poplar plantations and has been recently detected in North America (Newcombe and Chastagner, 1993). Selection for resistance to this biotrophic pathogen is thus an important challenge for poplar breeders (Dowkiw and Bastien, 2004). Severe damage occurs through decreased photosynthesis efficiency, early defoliation, and increased susceptibility to other pests and diseases (Gérard et al., 2006). To date, no resistant poplar cultivars are available, as new virulent strains of M. larici-populina are developing regularly. Sustainability of newly selected resistance requires a better understanding of the molecular mechanisms underlying Populus-Melampsora interactions. Most of the knowledge on lifestyle of fungal pathogens derives from studies in nonobligate biotrophs, and there are limited data about obligate biotrophs, such as rust fungi and powdery mildews. These fungi show a high level of compatibility with their host, but the reasons for the obligate biotroph status remains unknown (Mendgen and Hahn, 2002). To date, only sparse molecular data were obtained on defense mechanisms in poplar, and published data concern interactions with herbivores or viruses (Christopher et al., 2004; Smith et al., 2004; Major and Constabel, 2006; Ralph et al., 2006a). Molecular responses to Melampsora spp. invasion are unknown.

Here, we investigated the interaction between an interamerican hybrid poplar, P. trichocarpa × Populus deltoides ‘Beaupré’ and M. larici-populina. This poplar hybrid has been largely used in commercial poplar cultivation in Europe and harbors a qualitative resistance to M. larici-populina (Pinon and Frey, 2005). It is derived from a cross between P. trichocarpa, a species for which the genome sequence is available (Tuskan et al., 2006), and P. deltoides, a species from which rust resistance loci are inherited (Jorge et al., 2005). Efforts had been made to localize rust resistance-related genes in poplar pedigrees through genetic approaches (Cervera et al., 2004). Different studies successfully mapped rust-resistance quantitative trait loci in P. trichocarpa and P. deltoides (Lescot et al., 2004; Yin et al., 2004). To characterize the specific host-response to either virulent or avirulent strains of M. larici-populina, we carried out a combined molecular and histological analysis of time-course infection through scanning electron microscopy and quantitative PCR (qPCR) to detect rust progression in planta. We also investigated lignin and ROS production in plant tissue to determine major differences in plant response between the compatible and incompatible interactions. Then, rust-responsive genes were identified using either a suppression subtractive hybridization (SSH)-cDNA library of poplar leaves infected by the avirulent M. larici-populina strain or NimbleGen whole-genome expression arrays. Finally, expression profiles of Populus defense-related genes during the time course of infection were confirmed with additional transcriptome-based approaches (reverse transcription [RT]-qPCR and cDNA arrays).

RESULTS

Time Course of Compatible and Incompatible Interactions

Rust development in P. trichocarpa × P. deltoides ‘Beaupré’ leaves was monitored at the macroscopic and microscopic levels over a period of 10 d postinoculation (dpi) with either compatible (98AG31; pathotype 3-4-7) or incompatible (93ID6; pathotype 3-4) strains of M. larici-populina. In the compatible interaction, uredinia formation was visible under the abaxial epidermis 5 dpi, and by 6 to 7 dpi, uredinia emerged through the epidermis and formed orange pustules of 1- to 2-mm diameter (Fig. 1). Uredinia distribution was uniform on the leaf surface, and there were about 73 ± 6 pustules/cm². In the incompatible interaction (93ID6), no lesion or pustule was observed on the leaf surface over a period of 10 d. The abaxial epidermis showed very localized necrotic zones, and dark dots inside mesophyll tissues were visible in transparency after 6 dpi (Fig. 1). Control leaves inoculated with water showed no pustules or necrotic lesions after 10 d.

Figure 1.

Experimental design of time-course infection of detached leaves of P. trichocarpa × P. deltoides ‘Beaupré’ mock inoculated with water or inoculated with incompatible (avirulent, 93ID6) and compatible (virulent, 98AG31) strains of M. larici-populina. RNA was sampled in independent replicate experiments at 12, 24, and 48 hpi and were used for transcript profiling. Leaf tissues were also sampled at 2, 6, 12, 18, 24, 48, and 96 hpi and were then used for observation in the VPSEM, detection of ROS and lignin synthesis, and qPCR detection of pathogen growth in planta. Expression of disease (uredinia formation on abaxial epidermis) or resistance (localized hypersensitive reaction) was observed at 7 dpi. Necrotic lesions or uredinia on abaxial epidermis of foliar discs are shown at 7 dpi for the incompatible or the compatible interaction, respectively.

Infection structures developed by M. larici-populina at the leaf surface were monitored by aniline blue staining and light microscopy. Compatible and incompatible spores of M. larici-populina germinated within 2 h postinoculation (hpi) and germ tubes of different length (1 μm–1 mm) were observed on the leaf surface (Fig. 2A). Most of the germ tubes ramified, formed appressoria, and had successfully penetrated plant tissues at 2 hpi. Low-temperature variable pressure scanning electron microscopy (VPSEM) was used to follow in planta colonization of M. larici-populina by direct observations of infected leaf sections. Fungal structures on the leaf surface were similar to those observed with aniline blue staining for both interactions. Penetration through stomata occurred for about one-half of the successful events of germination between 2 and 6 hpi (Fig. 2B). At 6 and 12 hpi, substomatal vesicles were observed in the substomatal chambers (Fig. 2C), and infection hyphae were developing in the mesophyll tissue from these vesicles. At 18 and 24 hpi, infection hyphae extended into the mesophyll and in some cases reached the palisadic mesophyll (Fig. 2, D and E). The infection hyphae terminated their growth on a mesophyll cell forming haustorial mother cells, while other infection hyphae continued their course into the mesophyll after branching (Fig. 2, D and E). Infection structures inside the cells (i.e. haustoria) cannot be observed using VPSEM. Most of the compatible strain hyphae observed by VPSEM invaded both the spongy and the palisadic mesophyll by 48 hpi. The number of infection hyphae dramatically increased for the compatible strain at 96 hpi (Fig. 2F). Observations were further made at 5 and 7 dpi. In the compatible interaction, hyphae totally invaded the plant tissue around primary infection sites at 5 dpi, domes were formed in the spongy mesophyll, and spore-forming cells were differentiating (data not shown). By 7 dpi, domes corresponding to uredinial pustules released spores on the leaf abaxial surface (Fig. 2G). In the case of the incompatible interaction, the number of hyphae observed in planta was similar to that of the compatible interaction until 24 hpi, and a limited number of hyphae was observed at later time points. At 96 hpi, a few hyphae were ramified or extended into the palissadic mesophyll (Fig. 2H).

Figure 2.

Development of infection structures of compatible (A–G) strain 98AG31 and incompatible strain 93ID6 (H) of M. larici-populina during time-course infection of leaves of P. trichocarpa × P. deltoides ‘Beaupré’. A, Aniline blue-stained urediniospores producing several ramified germ tubes at vicinity of stomata and hyphal penetration through stomata without appressorium formation at 2 hpi. B, Urediniospores producing germ tubes and appressoria at vicinity of stomata and primary infection hyphae penetrating through stomata at 6 hpi. C, Substomatal vesicle formed under the abaxial epidermis and infection hyphae developing in the spongy mesophyll at 12 hpi. D, Infection hyphae colonizing the spongy mesophyll at 48 hpi. E, Infection hyphae developing from the spongy mesophyll to the palisade mesophyll at 48 hpi. F, Spongy and palisade mesophyll colonized by infection hyphae at 96 hpi. G, Uredinium formed on the abaxial epidermis releasing newly formed urediniospores at 196 hpi. H, Limited development of hyphae of the incompatible M. larici-populina strain 93ID6 in the mesophyll at 96 hpi compared to compatible strain in F. The hyphae of M. larici-populina were painted in red in F and H to help with visualization of fungal hyphae in plant mesophyll. Fungal infection structures are pinpointed by arrowheads. ap, Appressorium; gt, germ tube; ih, infection hyphae; inf, infection site; nf sp, newly formed spore; pih, primary infection hyphae; sp, spore; ssv, substomatal vesicle; ur, uredinium.

Fungal Growth in Leaves during Compatible and Incompatible Interactions

Assuming that the proportion of fungal and plant biomass present at any given time during an infection is equivalent to the proportion of fungal and plant DNA, quantitation of fungal nuclear ribosomal DNA (rDNA) internal transcribed spacer (ITS) can be used to estimate the extent of fungal growth in the plant (Boyle et al., 2005). Invasion of foliar tissues by compatible and incompatible strains was monitored in planta by qPCR quantification of M. larici-populina rDNA ITS. In planta growth of compatible and incompatible strains of M. larici-populina was similar during early stages of infection (2, 6, 12, and 24 hpi; Fig. 3). A drastic increase in fungal DNA mass of about 12-fold was observed between 24 and 48 hpi for the compatible strain, while the incompatible strain showed a slower increase (4-fold). The amount of fungal rDNA ITS decreased for the latter strain between 48 and 96 hpi, indicating a possible hyphal decay (Fig. 3). In contrast, growth of the compatible strain dramatically increased at 96 hpi, and DNA mass was more than 500-fold higher than the amount measured for the incompatible strain.

Figure 3.

Time-course infection of leaves of P. trichocarpa × P. deltoides ‘Beaupré’ by compatible (98AG31) and incompatible (93ID6) strains of M. larici-populina. Development of the two rust strains was followed in planta by specific amplification of the rDNA intergenic ITS region from total DNA extracted from inoculated leaf tissues at 2, 6, 12, 24, 48, and 96 hpi. Pathogen growth curves correspond to ΔCt of fungal ITS amplicons measured by quantitative PCR. Note the log scale. Estimates of fungal mass DNA are indicated on the graph for the two strains at 12, 24, 48, and 96 hpi. Amplifications were carried out on three biological replicates, and significant differences between the two M. larici-populina strains are indicated by a star (t test; P < 0.05).

Detection of ROS and Lignin Monomers

Leaf tissues inoculated with compatible and incompatible strains of M. larici-populina showed no endogenous ROS accumulation based on diaminobenzidine (DAB) staining (data not shown). In contrast, control leaves with hydrogen peroxide (H₂O₂) injection and wounded leaves exhibited DAB precipitates (data not shown). Phloroglucinol staining is considered to be specific for cinnamaldehyde end groups present in lignins (Nakano and Meshitsuka, 1992). At 2, 6, 12 (data not shown), 24, and 48 hpi, a light red coloration of vessels of major orders was observable on leaf inoculated with the incompatible strain, and no coloration was visible for the rest of the leaf tissues (Fig. 4A). An intense red coloration of leaf tissues was detected at 96 hpi in leaves inoculated with the incompatible strain (Fig. 4, A and B), whereas staining was restricted to lower orders of leaf vessels in the case of the compatible interaction or in control leaves (Fig. 4A). Infection hyphae of the compatible strain could be observed as yellowish dots in the mesophyll tissue at 96 hpi (Fig. 4B). In control leaf tissues, a light red coloration was restricted to vessels.

Figure 4.

Wiesner coloration revealing monolignol accumulation by a red coloration in leaf discs of P. trichocarpa × P. deltoides ‘Beaupré’ sampled on detached leaves inoculated with compatible (98AG31) or incompatible (93ID6) strains of M. larici-populina or mock inoculated with water. A, Cleared leaf discs at 24, 48, and 96 hpi in compatible and incompatible interactions and in the control treatment. B, A 10× close-up of leaf discs at 96 hpi. M. larici-populina infecting hyphae are visible by transparency in the case of the compatible interaction.

Defense-Related Genes in a SSH cDNA Library of Rust-Infected Leaves

SSH technology is a powerful approach to identify genes differentially expressed by cells or organisms in specific developmental stages (Diatchenko et al., 1996). A subtractive cDNA library from incompatible rust-infected leaves (12, 24, and 48 hpi), subtracted by RNA from noninoculated leaf tissues, was constructed. Assembly of 1,999 ESTs obtained from the 5′ and 3′ ends of 1,152 SSH cDNAs produced 967 nonredundant tentative consensus (TC) sequences corresponding to 486 different genes. These sequences have been deposited in the National Center for Biotechnology Information (NCBI) database (accession nos. CT027996–CT029994; CT033829). Among them, 357 ESTs (37%) corresponded to singletons, 610 ESTs (63%) were clustered in 129 TC, and approximately 90% of the ESTs were supported on both 5′ and 3′ ends. A BLASTN search showed that 467 of these ESTs have a homolog in the P. trichocarpa genome sequence in the Joint Genome Institute (JGI) database. The remaining TCs showed no significant similarity to any genes in the NCBI databases, suggesting that they are either absent from the current P. trichocarpa genome sequence assembly (version 1.1) or they might be specific to P. trichocarpa × P. deltoides ‘Beaupré’. The sequences showed no significant hits against M. larici-populina and other basidiomycetes genomes on the JGI portal (data not shown).

A summary of homology searches against GenBank using the BLAST algorithm is shown in Supplemental Table S1. ESTs from Rubisco and chlorophyll a/b-binding protein (CAB) genes were not identified in the SSH cDNA library, whereas they represent 24% of a ‘Beaupré’ cDNA leaf library (Kohler et al., 2003), indicating that the SSH library was efficiently depleted from constitutively expressed transcripts. This analysis indicated that the most prevalent ESTs represented genes directly connected with plant defense and nitrogen metabolism. Among these abundant ESTs, an EST matching an inositol-3-P synthase (I3PS) gene was the most frequently detected sequence (199 occurrences; 20% of ESTs; Table I). Several transcripts corresponding to defense-related genes (PR-1, PR-2, PR-3, and PR-5) and pathogen perception (receptor-like kinase [RLK], LRR, and ankyrin protein) were also among the most abundant SSH ESTs. For example, PR-1 transcripts represented about 4% of the SSH transcripts, indicating a striking up-regulation of its expression. A least three distinct types of chitinase-like genes, including basic (PR-8) and acidic (PR-3) chitinases, were expressed in rust-infected leaves. Transcripts coding for Asp aminotransferase, Asn synthetase, and NADH-Glu synthase, with no obvious direct defensive roles, were also abundant in the SSH library.

Table I.

Most prevalent rust-responsive transcripts in leaves of P. trichocarpa × P. deltoides ‘Beaupré’ inoculated with the incompatible strain 93ID6 of M. larici-populina measured by ESTs redundancy in a SSH-enriched cDNA library

nd, Not detected.

Contig IDa	Accession No. (5′)b	Accession No. (3′)b	Best Database Match (Species)c	P. trichocarpa Protein ID No.d	Percent (EST No.)
2YE21	CT029167	CT029166	I3PS (Mesembryanthemum crystallinum)	832275	20.4 (199)
1YH15	CT029702	CT029701	PR-1 (Vitis vinifera)	550049	3.8 (37)
3YM24	CT028131	CT028130	Thaumatin-like protein, PR-5 (Nicotiana tabacum)	747341	2.2 (21)
1YL15	CT029535	CT029534	Protein disulfide isomerase, PDI-3 (Cucumis melo)	230101	1.3 (13)
1YC18	CT029896	CT029895	Asp aminotransferase (Arabidopsis)	710171	1.1 (11)
1YH09	CT029711	CT029710	Chitinase (class I), PR-3 (Medicago sativa)	831333	0.9 (9)
2YE11	CT029187	CT029186	Cyt P450 (Oryza sativa)	680604	0.9 (9)
3YD08	CT028543	CT028542	Asp synthetase (Triphysaria versicolor)	722643	0.9 (9)
3YG19	CT028393	CT028392	Hypothetical protein (Glycine max)	579371	0.8 (8)
2YK09	CT028933	CT028932	Subtilisin-like Ser protease, AIR3 (Arabidopsis)	810987	0.8 (8)
1YI17	CT029657	CT029656	Chitinase, PR-3 (Psophocarpus tetragonolobus)	586583	0.6 (6)
1YK13	nd	CT029580	Acidic class III chitinase SE2, PR-3 (Beta vulgaris)	746640	0.6 (6)
2YB13	CT029291	CT029290	RNA helicase (O. sativa)	755273	0.5 (5)
2YB09	nd	CT029298	Asp synthetase (Helianthus annuus)	722643	0.5 (5)
1YK07	CT029592	CT029591	Wall-associated kinase (Arabidopsis)	758034	0.5 (5)
1YJ14	CT029624	CT029623	Senescence-related protein (Pyrus pyrifolia)	265740	0.4 (4)
2YI14	CT029011	CT029010	β-1,3-Glucanase, PR-2 (Fragaria × ananassa)	290846	0.4 (4)
1YP01	CT029388	CT029387	Receptor-like protein kinase, RLK (e-value < E-5)	756597	0.4 (4)
3YF23	CT028431	CT028430	LRR protein kinase (Arabidopsis)	826060	0.4 (4)
2YC08	CT029266	CT029265	Disease resistance protein (Arabidopsis)	788310	0.4 (4)
1YG06	CT029755	CT029754	Protein disulfide isomerase, PDI-2 (C. melo)	565873	0.4 (4)
3YD21	CT028524	CT028523	Ankyrin repeat protein (e-value < E-5)	571736	0.4 (4)
2YM21	CT028822	CT028821	β-1,3-Glucanase, PR-2 (Hevea brasiliensis)	769807	0.4 (4)
1YN04	CT029471	CT029470	NADH Glu synthase (Phaseolus vulgaris)	824538	0.4 (4)
1YG23	CT029729	CT029728	Subtilisin-like proteinase (O. sativa)	790236	0.4 (4)

^aRepresentative EST ID from assembly contig.

^bGenBank accession number of 5′ and 3′ sequences of transcripts.

^cBest database match (and corresponding species) obtained with a BLASTX query at NCBI.

^dProtein ID number of the best database match in P. trichocarpa ‘Nisqually-1’ (version 1.1) obtained from a BLASTN search on the JGI Web site (http://genome.jgi-psf.org/Poptr1_1/Poptr1_1.home.html).

Identification of Rust-Responsive Genes at 48 hpi Using Whole-Genome Oligoarrays

Microscopy observations (Fig. 2) and qPCR measurement of fungal DNA (Fig. 3) showed a shift in fungal progression between compatible and incompatible interactions at 48 hpi. A strong difference in lignin deposition at infection sites was observed in the case of the incompatible interaction at 96 hpi (Fig. 4), suggesting that the host molecular response probably initiated after the fungus entered into the mesophyll (12 hpi; Fig. 2, D and E) and when haustorial infection structures were differentiating. We thus investigated changes in gene expression in ‘Beaupré’ leaves at 48 hpi in compatible and incompatible interactions (Fig. 1) using the NimbleGen Populus whole-genome expression oligoarray (Tuskan et al., 2006). We identified 280 (0.4%) and 1,730 (2.6%) transcripts differentially accumulated (≥2-fold) in the compatible and incompatible interactions, respectively, compared to control leaves mock inoculated with water (Table II; Supplemental Table S2). Among these transcripts, 150 showed an increased concentration ≥3-fold in the incompatible interaction and only seven in the compatible interaction.

Table II.

Expression ratios of rust-responsive transcripts from P. trichocarpa × P. deltoides ‘Beaupré’ measured with the NimbleGen P. trichocarpa whole-genome expression oligoarray at 48 hpi in ‘Beaupré’ leaves inoculated with incompatible (I48) or compatible (C48) strains of M. larici-populina

NimbleGen Probe IDa	P. trichocarpa Protein ID No.b	I48-Fold- Regulationc	P Valuec	Best BLAST Hitd	AGI No.e
TREE0002S00038855	678883	30.6	8.77E-12	No hit (RISP)	No hit
TREE0002S00039697	272826	14.8	2.65E-10	GST U2 (Nicotiana benthamiana)	At1g17170
TREE0002S00058098	792358	9.6	5.49E-07	Hypothetical protein (Arabidopsis)	At1g58170
TREE0002S00060385	277644	7.6	1.57E-06	GST parC, auxin-regulated protein (N. tabacum)	At1g78380
TREE0002S00000329	711753	7.5	4.92E-06	PR protein (dirigent-like protein; Pisum sativum)	At1g64160
TREE0002S00058872	657351	7.2	1.67E-07	GST 18 (P. alba × P. tremula)	At2g29420
TREE0002S00001434	731797	7.1	8.96E-06	Late embryogenesis abundant protein, SAG21 (Arabidopsis)	At4g02380
TREE0002S00000421	712863	6.5	1.96E-05	UDP-Glc:protein transglucosylase-like protein (Lycopersicon esculentum)	At3g02230
TREE0002S00040010	276538	6.4	1.47E-05	GST parC, auxin-regulated protein (N. tabacum)	At1g17180
TREE0002S00035838	795681	6.3	7.03E-06	Acidic chitinase (P. tetragonolobus)	At5g24090
TREE0002S00032433	581980	6.2	1.07E-05	PR-1 protein (Arabidopsis)	At2g14610
TREE0002S00028427	746640	6.2	1.13E-05	Acidic endochitinase, glycoside hydrolase family 18 (Medicago truncatula)	At5g24090
TREE0002S00052008	820835	6.2	4.67E-06	GST 18 (P. alba × P. tremula)	At2g29420
TREE0002S00028575	748543	6.1	1.33E-05	PR-1 protein (Arabidopsis)	At2g14610
TREE0002S00000256	710544	6.0	7.90E-05	Late embryogenesis abundant protein, SAG21 (Arabidopsis)	At4g02380
TREE0002S00015535	769807	5.5	1.10E-05	β-1,3-Glucanase (H. brasiliensis)	At4g16260
TREE0002S00028427	746640	5.4	1.13E-05	Acidic endochitinase, glycoside hydrolase family 18 (M. truncatula)	At5g24090
TREE0002S00024163	417599	5.3	1.20E-05	RLK5 (Arabidopsis)	At5g25930
TREE0002S00063111	248394	5.3	3.19E-05	PSII CP43 protein (Panax ginseng)	AtCg00280
TREE0002S00000638	714634	5.3	1.57E-05	Hypothetical protein (P. deltoides × Populus maximowiczii)	At4g19950
TREE0002S00040235	279076	−4.1	1.75E-03	rRNA intron-encoded homing endonuclease (O. sativa)	No hit
TREE0002S00063251	256788	−4.2	1.18E-02	Lys decarboxylase (O. sativa)	At5g06300
TREE0002S00047596	735328	−4.2	5.07E-04	Ribosomal protein 40S S9 (Solanum demissum)	At5g39850
TREE0002S00047882	640496	−4.2	1.16E-04	β-Tubulin (Gossypium hirsutum)	At5g12250
TREE0002S00063570	Cp_orf79	−4.2	1.17E-03	Hypothetical chloroplastic protein (Spinacia oleracea)	No hit
TREE0002S00062228	418172	−4.5	7.99E-04	Magali Spm transposable element 60I2G03 (P. deltoides)	No hit
TREE0002S00040045	277030	−4.6	5.33E-04	Hypothetical chloroplastic protein (Saccharum officinarum)	No hit
TREE0002S00063598	cp_orf61	−4.6	4.47E-03	Hypothetical protein SpolCp101 (S. oleracea)	AtCg00300
TREE0002S00035479	795166	−4.8	1.99E-03	No hit	No hit
TREE0002S00030128	596748	−4.8	5.91E-04	GDSL-like lipase/acylhydrolase (O. sativa)	At1g28640
TREE0002S00032920	585246	−5	1.32E-03	Hypothetical chloroplast ATPase (Ycf2; P. alba)	AtCg00860
TREE0002S00039944	275797	−5.6	2.95E-02	Hypothetical chloroplastic protein (Cuscuta reflexa)	No hit
TREE0002S00056518	771095	−5.8	7.19E-03	No hit	No hit
TREE0002S00040050	277108	−5.9	3.51E-03	Hypothetical chloroplastic protein (N. tabacum)	No hit
TREE0002S00031244	587419	−6	1.17E-02	No hit	No hit
TREE0002S00063597	cp_ycf15	−6.6	1.97E-03	Hypothetical protein (Orf77/Ycf15-A; Arabidopsis)	AtCg00870
TREE0002S00023106	195834	−7	3.12E-03	Hypothetical auxin-induced protein (Capsicum annuum)	At4g34800
TREE0002S00042443	820269	−7.2	3.04E-02	Hypothetical protein (Arabidopsis)	At4g25030
TREE0002S00060995	290970	−8	2.98E-03	Lys decarboxylase-like protein (O. sativa)	At5g06300
TREE0002S00040072	277323	−8.1	3.46E-03	Hypothetical chloroplastic protein (N. tabacum)	No hit
NimbleGen Probe IDa	P. trichocarpa Protein ID No.b	C48-Fold- Regulationc	P Valuec	Best BLAST Hitd	AGI No.e
TREE0002S00055699	594680	4.2	2.22E-03	Anthocyanin acyltransferase-like protein (Arabidopsis)	At5g39050
TREE0002S00029168	837131	3.6	9.66E-04	Ferredoxin-nitrite reductase (Betula pendula)	At2g15620
TREE0002S00037742	811643	3.3	3.16E-03	CuZn-superoxide dismutase (P. tremula × P. tremuloides)	At2g28190
TREE0002S00009339	568530	3.2	1.87E-02	LRR-containing hypothetical protein (Arabidopsis)	At5g55540
TREE0002S00067017	544845	3	4.55E-02	Integrase polyprotein (M. truncatula)	At2g15650
TREE0002S00001434	731797	3	9.18E-03	Late embryogenesis abundant protein, SAG21 (Arabidopsis)	At4g02380
TREE0002S00048762	679519	3	3.74E-03	Polyubiquitin UBQ10 (Arabidopsis)	At4g05320
TREE0002S00052689	566171	2.9	7.56E-03	β-Ketoacyl-CoA synthase (O. sativa)	At5g43760
TREE0002S00034768	794254	2.8	2.72E-03	Aldo/keto reductase (M. truncatula)	At3g53880
TREE0002S00041247	291991	2.7	1.37E-02	Rubisco, large chain (Lophocolea martiana)	AtCg00490
TREE0002S00063977	588636	2.7	4.92E-03	gag/pol polyprotein (S. demissum)	At2g14380
TREE0002S00060457	279164	2.7	7.55E-03	RLK (Arabidopsis)	At4g27300
TREE0002S00063843	660789	2.7	1.07E-02	No hit	No hit
TREE0002S00028989	826800	2.6	8.29E-03	Ser/Thr protein kinase (M. truncatula)	At5g09890
TREE0002S00030792	583368	2.6	1.43E-02	No hit	No hit
TREE0002S00048783	748355	2.6	1.02E-02	Pollen coat protein-like (Arabidopsis)	At5g38760
TREE0002S00007567	561703	2.6	5.65E-03	No hit	No hit
TREE0002S00061579	172038	2.6	2.93E-03	Hypothetical protein (Arabidopsis)	At1g08440
TREE0002S00002619	818390	2.5	1.66E-02	Calcium-binding protein (Atriplex nummularia)	At3g50360
TREE0002S00024949	421059	2.5	3.45E-03	Δ-Pyrroline-5-carboxylate synthetase (G. max)	At2g39800
TREE0002S00059712	263964	−4.3	1.58E-02	Lys decarboxylase-like protein (O. sativa)	At5g06300
TREE0002S00020096	648814	−4.6	3.00E-02	Hypothetical splicing factor (Arabidopsis)	At2g16940
TREE0002S00023106	195834	−4.6	2.82E-02	Hypothetical auxin-induced protein (C. annuum)	At4g34800
TREE0002S00037394	793544	−4.8	1.58E-03	Hypothetical protein (O. sativa)	No hit
TREE0002S00060995	290970	−5.1	1.67E-02	Lys decarboxylase-like protein (O. sativa)	At5g06300
TREE0002S00047970	817423	−5.3	2.73E-03	Hypothetical protein (Arabidopsis)	At5g42050
TREE0002S00055491	591262	−5.4	1.21E-02	Cys proteinase (Alnus glutinosa)	At2g27420
TREE0002S00056518	771095	−5.4	1.14E-02	No hit	No hit
TREE0002S00031048	585596	−5.5	3.10E-03	Calcium-binding protein (Arabidopsis)	At4g13440
TREE0002S00047850	827481	−5.8	2.75E-02	Zinc finger protein PHD family (G. max)	At2g02470
TREE0002S00054306	588374	−6.3	4.23E-04	No hit	No hit
TREE0002S00058200	789011	−6.4	4.83E-03	Pollen coat oleosin-Gly rich protein (Arabidopsis)	At5g07565
TREE0002S00066249	660216	−6.6	6.71E-03	No hit	No hit
TREE0002S00031244	587419	−6.7	8.78E-03	No hit	No hit
TREE0002S00064258	681711	−6.9	2.92E-03	Magali Spm transposable element 60I2G03 (P. deltoides)	No hit
TREE0002S00029354	589352	−7.1	3.54E-02	Hypothetical protein, sec34 homolog (Arabidopsis)	At1g73430
TREE0002S00050427	274646	−7.4	3.04E-03	Phosphoprotein phosphatase (Arabidopsis)	At1g05000
TREE0002S00061299	298144	−8.7	3.82E-03	Lys decarboxylase-like protein (O. sativa)	At5g06300
TREE0002S00044995	583302	−10.5	6.67E-03	Calcium-binding EF-hand family protein-like (Arabidopsis)	At2g44310
TREE0002S00047844	732714	−11	1.50E-02	Hypothetical protein (O. sativa)	At5g24610

^aProbe ID number on NimbleGen Populus expression array version 2.0 (NCBI GEO platform GPL2699).

^bProtein ID number of corresponding best gene model in the P. trichocarpa ‘Nisqually-1’ genome sequence (version 1.1) at http://genome.jgi-psf.org/Poptr1_1/Poptr1_1.home.html.

^cExpression ratios (and associated P value) calculated between normalized transcript concentration of inoculated versus mock-inoculated (water) Populus leaves obtained with duplicated probes on NimbleGen whole-genome oligoarray and three biological replicates, ratios <1.0 were inverted and multiplied by −1 to aid their interpretation.

^dBest database match (and corresponding species) obtained with a BLASTX query at NCBI.

^eAGI number of best Arabidopsis homolog.

Incompatible Interaction

The transcript showing the highest rust-induced accumulation (32-fold) in the incompatible interaction corresponded to a P. trichocarpa gene model (protein identification [ID] no. 678883) with no sequence similarity in the nonredundant NCBI database or the Arabidopsis genome. The corresponding genomic sequence is located at the beginning of scaffold 5,059 of the P. trichocarpa genome assembly and is truncated in 5′. This transcript showed a strong homology with several Populus ESTs in NCBI dbEST. These ESTs are longer in their 5′ and 3′ ends of nucleotide sequences and encode an 82-amino acid polypeptide with no ortholog in databases. SignalP 3.0 (http://www.cbs.dtu.dk/services/SignalP/) analysis identified a signal peptide of 24-amino acid length, and Phobius (http://phobius.cgb.ki.se/) predicted a noncytoplasmic localization for the protein, suggesting a putative localization outside of the plant cell.

Genes encoding enzymes known to be associated with the host defense response were highly induced at 48 hpi in the incompatible interaction and presented no significant regulation in the compatible interaction. These genes encode several types of PR proteins, such as PR-1 homologs, basic glucan-endo-1,3-β-glucosidase (PR-2), thaumatin-like, and osmotin-like proteins (PR-5), which were also identified in the SSH cDNA library and PR-10-like proteins. All those PR transcripts showed an increase ≥3-fold compared to the mock-inoculated treatment.

Among other transcripts showing an important induction in the incompatible interaction (≥5-fold accumulation), we detected transcripts coding for components of the signaling pathways (calmodulins), an RLK, and the I3PS identified in the SSH library. Transcripts related to secondary metabolism and cell wall synthesis, such as dirigent-like proteins, chalcone-, flavonol-, and tropinone reductases, were also detected among transcripts significantly accumulated in the incompatible interaction supporting the observed accumulation of phloroglucinol-stained lignin monomers (Fig. 4A).

Several rust-induced genes corresponded to different members of the glutathione S-transferase (GST) gene family. Twelve different GST transcripts showed at least a 2-fold accumulation in the incompatible interaction compared to the mock-inoculated treatment. The most strongly accumulated GST transcripts consisted of two phylogenetic groups of sequences (group 1, protein ID nos. 276538, 277644, and 272826; and group 2, protein ID nos. 657351 and 820835; Table II) that shared a high sequence similarity (≥90%). Alignment of the oligonucleotide probe sequences matching these GST sequences revealed a significant overlap that may have resulted in possible cross hybridizations between transcript species. Nevertheless, all transcripts that were strongly accumulated belonged to the tau GST class and not to other classes (i.e. phi, theta, and zeta GSTs) described so far in plants (Dixon et al., 2002; Wagner et al., 2002).

There were a few transcripts (292) showing a decrease in concentration at 48 hpi in the incompatible interaction. These include several chloroplastic transcripts, as well as different transcripts coding transposable elements, including the magali Spm-like transposable elements (protein ID no. 418172) located within a rust-resistance locus in P. deltoides (Lescot et al., 2004). A transcript encoding a Lys decarboxylase also showed a dramatic decrease in concentration at 48 hpi (8-fold). Most of these transcripts also presented a decrease in concentration in the compatible interaction (Table II).

Compatible Interaction

Only a few transcripts were significantly up- or down-regulated in the compatible interaction (158 and 122, respectively). The highest increase (4.2-fold) corresponded to a transcript coding for an anthocyanin acyltransferase-like protein. Several transcripts corresponding to signaling components (Ser/Thr kinase, calcium-binding protein) and LRR-containing proteins, including an RLK, were significantly induced (≥2.5-fold). Interestingly, a transcript coding a Δ1-pyrroline-5-carboxylase (protein ID no. 421059) showed a 2.5-fold induction. This protein is involved in Pro biosynthesis, and a regulation of transcripts involved in the catabolism of Pro was previously described specifically in the flax (Linum usitatissimum)-Melampsora lini compatible interaction (Ayliffe et al., 2002).

Identification of Rust-Responsive Genes at 48 hpi Using cDNA Microarrays

Before the availability of the NimbleGen whole-genome expression oligoarrays, we carried out a series of transcript profilings of rust-infected (incompatible interaction) and mock-inoculated (control) P. trichocarpa × P. deltoides ‘Beaupré’ leaves at 48 hpi using 28 K Platform for Integrated Clone Management (PICME) cDNA microarrays. Thus, these datasets obtained on different plants (year 2004) than those for oligoarray-based expression profiling (year 2005) were compared to the oligoarray expression profiles.

We identified 1,614 transcripts corresponding to 1,055 P. trichocarpa gene models that were significantly accumulated or decreased (≥2-fold) at 48 hpi in the incompatible interaction compared to control leaves. Among these transcripts, 218 showed an increased concentration ≥3-fold and 136 a decreased concentration ≤3-fold. Transcripts accumulated in response to the rust infection are mostly identical to those described in the oligoarray expression profiling (Table III; Supplemental Table S3). The highest levels of rust induction (≥8-fold) were observed for transcripts coding PR proteins, such as PR-1, PR-2, PR-5, PR-8, and PR-10, and the rust-induced secreted protein (RISP) number 678883. GST transcript corresponding to protein ID number 820835 showed a 5.9-fold induction. In addition, two transcripts coding GSTs not significantly regulated in the whole-genome oligoarray analysis and belonging to the same rust-induced tau GST class (see above) were detected as significantly induced on PICME microarrays (protein ID no. 670248, 9.4-fold; protein ID no. 658124, 6-fold).

Table III.

Expression ratios of rust-responsive transcripts in leaves of P. trichocarpa × P. deltoides ‘Beaupré’ measured with the PICME 28 K cDNA Populus microarray at 48 hpi in ‘Beaupré’ leaves inoculated with the incompatible (I48) strain of M. larici-populina versus mock-inoculated leaves

The highest expression ratio is presented for gene models (P. trichocarpa protein ID no.) represented by different cDNA probes on the array.

PICME ID No.a	P. trichocarpa Protein ID No.b	Best BLAST Hit (Species)c	I48-Fold- Regulationd	Bayes-lnP Valued	AGI No.e
AJ773363	669475	Thaumatin-like protein, PR-5 (Actinidia deliciosa)	28.8	5.40E-07	At4g11650
AJ773115	751998	β-1,3-Glucanase, PR-2 (Fragaria × ananassa)	28.7	3.00E-06	At4g16260
AJ777100	595857	Basic PR protein, PR1 (ATPRB1; Arabidopsis)	25.7	1.02E-05	At2g14580
AJ769638	711753	PR protein, dirigent-like protein (P. sativum)	23.8	9.26E-05	At1g64160
CA822510	678883	No hit (RISP)	19.4	1.41E-03	No hit
AJ777741	769807	β-1,3-Glucanase, PR-2 (H. brasiliensis)	15.1	5.45E-06	At4g16260
AJ774585	746640	Acidic endochitinase, PR-3 (M. truncatula)	14.4	2.33E-06	At5g24090
AJ770319	645750	LRR, putative disease resistance protein (Arabidopsis)	13.1	1.14E-04	At1g74170
CF235659	732511	Heavy metal transport/detoxification protein (M. truncatula)	13.1	1.72E-04	At3g07600
AJ774712	792358	Dirigent protein (Arabidopsis)	12.7	1.71E-04	At1g58170
CB239336	654817	Ascorbate oxidase (L. esculentum)	11.6	2.05E-05	At4g39830
CB239610	754908	Calmodulin-binding protein, NPGR1 (Arabidopsis)	11.1	1.68E-06	At1g27460
AJ779296	758549	Hypothetical protein (Arabidopsis)	10.9	4.82E-03	At1g61667
AJ780675	171587	Putative receptor protein kinase, PERK1 (G. max)	10.6	1.22E-04	At1g52290
CA825881	286375	Hypothetical protein (Arabidopsis)	10.5	7.55E-03	At3g17380
CA824683	654672	No hit	10.5	3.02E-04	No hit
CF229868	567801	Hypothetical protein (Arabidopsis)	10.4	1.4E-02	At5g10695
AJ780826	748543	PR protein, PR-1 (Arabidopsis)	10.3	4.59E-06	At2g14610
AJ777924	722998	Heavy metal transport/detoxification protein (M. truncatula)	9.7	2.40E-04	At3g07600
AJ771768	641396	E-class Cyt P450, group I (M. truncatula)	9.6	2.86E-05	At1g12740
AJ769643	837541	ras-related GTP-binding protein (N. tabacum)	−4.2	3.01E-03	At5g45130
AJ780783	197959	Ribosomal protein 30S S5 (Arabidopsis)	−4.2	1.11E-02	At2g33800
AJ775019	282386	Ser/Thr protein kinase (Arabidopsis)	−4.2	9.31E-05	At3g53030
CA821048	565812	Hypothetical protein PGR5 (Arabidopsis)	−4.3	1.40E-04	At2g05620
AJ780576	824324	Triacylglycerol/steryl ester lipase-like protein (M. truncatula)	−4.3	9.86E-05	At5g14180
CA821054	431886	Rubisco activase (G. hirsutum)	−4.4	1.20E-04	At2g39730
AJ772360	568201	Thi4 thiamine biosynthetic enzyme (Citrus sinensis)	−4.5	5.36E-04	At5g54770
AJ773302	814251	CAB (Daucus carota)	−4.5	7.30E-05	At5g54270
AJ774495	800945	Nine-cis-epoxycarotenoid dioxygenase4 (P. sativum)	−4.6	8.37E-04	At4g19170
AJ768470	249554	Dienelactone hydrolase (M. truncatula)	−4.6	1.39E-04	At1g35420
CA820871	657524	Galactinol synthase, isoform GolS-1 (Ajuga reptans)	−4.7	2.80E-04	At1g60470
AJ773153	589333	Peroxiredoxin Q (P. trichocarpa × P. deltoides)	−4.7	1.61E-05	At3g26060
AJ774710	671325	PSI light-harvesting antenna CAB (B. vulgaris)	−4.7	8.23E-05	At3g47470
AJ773223	817727	Thi1 thiamine biosynthetic enzyme (Picrorhiza kurrooa)	−4.8	2.89E-05	At5g54770
CA821119	744667	Light-harvesting complex II type III CAB (Vigna radiata)	−5.1	4.44E-05	At5g54270
AJ770887	724963	PSI light-harvesting CAB (N. tabacum)	−5.3	2.33E-05	At3g54890
CF232422	597692	SMAD/FHA (M. truncatula)	−5.6	1.34E-03	At3g13780
AJ774616	551646	Pepsin A (Arabidopsis)	−7.8	5.22E-05	At1g09750
AJ772852	261110	Pro-rich protein (G. max)	−9.5	1.79E-05	At2g45180
CF229292	825296	Lipid-binding protein (Arabidopsis)	−12.5	2.8E-02	At3g53980

^aProbe ID number on the PICME 28K cDNA Populus microarray (http://www.picme.at/).

^bProtein ID number of corresponding best gene model in the P. trichocarpa ‘Nisqually-1’ genome sequence (version 1.1) at http://genome.jgi-psf.org/Poptr1_1/Poptr1_1.home.html.

^cBest database match (and corresponding species) obtained with a BLASTX query at NCBI.

^dExpression ratios calculated between inoculated versus mock-inoculated (water) normalized transcript concentration values and associated P value obtained from two independent biological replicates, ratios <1.0 were inverted and multiplied by −1 to aid their interpretation.

^eAGI number of best Arabidopsis homolog.

Several Cyt-P450 transcripts showed contrasting expression profiles, some being strongly induced and others repressed as observed on the whole-genome oligoarray. Several other transcripts related to redox regulation also showed a slight decrease in concentration at 48 hpi in the incompatible interaction (e.g. peroxidases, thioredoxin), whereas other transcripts in the same cellular category were strongly induced (e.g. GSTs, protein disulfide isomerase, and peroxidases). Several transcripts involved in the photosynthetic machinery and carbon metabolism (e.g. light-harvesting complex CAB, PSI and PSII polypeptides, Rubisco, RuBisCO activase) and two different transcripts involved in thiamin biosynthesis showed a decreased concentration at 48 hpi in the incompatible interaction.

Validation of Rust-Regulated Genes Using cDNA Macroarrays and RT-qPCR

Expression data were further carried out at 12, 24, and 48 hpi in both compatible and incompatible interactions by either a RT-qPCR approach or reverse-northern on a Populus 4 K cDNA Nylon macroarray (Supplemental Table S4) with different sets of biological replicates than those used in the expression profiling experiments described above.

We measured transcripts coding for PR-1, PR-5, and PR-10 proteins as typical genes triggered by host defense reactions at 48 hpi in leaf tissues. The transcripts coding for I3PS (protein ID no. 832275), the dirigent-like protein (protein ID no.711753), NPR1 (protein ID no. 253241), and the RISP (protein ID no. 678883) that showed the highest transcripts accumulation based on the whole-genome oligoarrays, cDNA microarrays, or SSH library sequencing were also measured. Genes coding for a PSI center reaction subunit (protein ID no. 711610) and the small subunit of the Rubisco (protein ID no. 813777) that were slightly down-regulated during both types of interactions were included in the set of genes tested by RT-qPCR.

Strong accumulation of the selected rust-induced transcripts was confirmed by RT-qPCR amplification in leaf tissues challenged by the incompatible strain of M. larici-populina compared to mock-inoculated tissues. Maximum induction of PR genes was reached at 48 hpi (Fig. 5), and, interestingly, NPR1 transcript showed a peak of expression at 24 hpi. The transcript coding the RISP (protein ID no. 678883) with the highest accumulation (32-fold) detected at 48 hpi with whole-genome oligoarray profiling showed a different profile with RT-qPCR. A strong induction (7-fold) was measured at earlier time points in the incompatible interaction, and a lower induction level was detected at 48 hpi. We thus measured the level of this latter transcript by RT-qPCR with the RNA samples used to perform the whole-genome oligoarray hybridizations, and we observed a 56 (±8)-fold accumulation at 48 hpi (Fig. 5). This observation confirmed that the rate of induction is influenced by the physiological status of the infected plants rather than strong technical biases in array measurement. In some cases, RT-qPCR revealed a late induction of transcripts coding PR proteins (e.g. PR-1 and PR-10; Fig. 5) at 48 hpi in the compatible interaction with lower levels than those reached in the incompatible interaction, whereas array analysis did not reveal such induction.

Figure 5.

RT-qPCR expression patterns for transcripts of proteins PR-1, PR-5, PR-10, I3PS, NPR1, dirigent-like protein, RISP number 678883, ribulose 1,5-bisphosphate carboxylase oxygenase, arabinogalactan protein, and PSI center reaction IV. Total RNA of mock-inoculated or inoculated leaves of P. trichocarpa × P. deltoides ‘Beaupré’ with either compatible (black bars) or incompatible (white bars) strains of M. larici-populina 12, 24, and 48 hpi was isolated, and aliquots of 1 μg were used for first-strand cDNA synthesis. PCR was performed with 2 μL of first-strand cDNA. A control with no RT in the first-strand cDNA synthesis reaction mix was included to control for the lack of genomic DNA. The Populus ubiquitin transcript (GenBank ID CA825222) was used as a control for transcript not regulated by rust. Inset, Data for the RISP number 678883; transcript concentration was measured by RT-qPCR on the RNA samples used for the whole-genome oligoarray analysis at 48 hpi (n = 3).

Comparison of the Different Transcript Profiling Approaches

Genes showing striking differences in transcript concentration during the incompatible interaction were detected by the different transcript profiling approaches, i.e. SSH cDNA sequencing, whole-genome oligoarrays, cDNA microarrays, cDNA macroarrays, and RT-qPCR, indicating consistency of the various approaches (Fig. 5; Table IV), although the regulation ratio may vary. For example, the RISP transcript that showed the highest accumulation (32-fold) based on the whole-genome oligoarrays was represented by several ESTs on the cDNA microarrays and showed a level of accumulation over 10-fold (protein ID no. 678883; Table III). In contrast, the strongly induced transcript coding PR-5 protein (protein ID no. 669475) showed a 28.8-fold induction based on the cDNA microarray and was quite abundant in the SSH library (2% of the cDNA clones), whereas a lower level was detected on the whole-genome oligoarray. The I3PS transcript that was highly abundant in the SSH library (approximately 20.4% of the cDNA clones) only showed approximately 5-fold accumulation in both whole-genome oligoarray and cDNA microarray analyses. In addition to bias resulting from the different technologies (full-length cDNA versus 60-mer oligonucleotide probes), differences in mRNA accumulation detected between the various profiling approaches likely reflect the fact that RNA were extracted from different sets of biological replicates with delayed plant defense response due to the variable physiological status of cuttings grown in greenhouse. A high proportion of rust-induced genes were identified in the SSH library. This technique presents the potential of identifying rare transcripts or genes expressed locally that may be missed in microarray expression profiling (e.g. statistics stringency in array analysis). Thus, this approach is not redundant but complementary to array-based transcriptome profiling.

Table IV.

List of significantly rust-induced transcripts detected using different transcriptome-based expression analyses in leaves of P. trichocarpa × P. deltoides ‘Beaupré’ inoculated by the incompatible M. larici-populina strain 93ID6 compared to mock-inoculated (water) control leaves at 48 hpi

Expression ratios (fold-induction) measured with NimbleGen whole-genome oligoarrays, PICME cDNA microarrays, or Nylon-based cDNA macroarrays are given. Transcripts were selected when identified through at least two different approaches and based on their expression ratios (at least 3-fold induction in one of the transcriptomic assay). Matches of a transcript sequence with an EST from a rust-responsive Populus leaves cDNA library (SSH) are indicated (×). –, Data are not available (not detected, not significant, or not present on array or in cDNA library).

Populus Protein ID No.a	NimbleGen Populus Oligoarray	PICME cDNA Microarray	cDNA Macroarray	SSH EST	Best BLAST Hit (Species)b	AGI Homologc
678883	30.6	19.4	4.3	–	RISP (P. trichocarpa × P. deltoides)	No hit
669475	3.9	28.8	23.3	×	Thaumatin, PR-5 protein (A. deliciosa)	At4g11650
832275	5	4.2	26.9	×	I3PS (Nicotiana paniculata)	At5g10170
769807	5.5	15.1	19.2	×	Glucan endo-1,3-β-d-glucosidase, PR-2 (H. brasiliensis)	At4g16260
180318	3.5	9.4	10.9	–	Osmotin-like protein (G. hirsutum)	At4g11650
761866	2.6	3.6	7.4	–	Hypothetical protein (G. max)	At4g16380
731797	7.1	3.5	2.8	–	Senescence-associated gene 21, SAG21 (Arabidopsis)	At4g02380
820835	6.2	6.0	3.9	–	GST18 (P. alba × P. tremula)	At1g10370
826324	2.3	4.0	6.0	–	Kunitz trypsin inhibitor 3, Pop-TI3 (P. trichocarpa × P. deltoides)	At1g73325
727757	3.5	4.4	5.3	–	1,4-Benzoquinone reductase (P. armeniaca)	At4g27270
726588	3.5	2.0	5.6	–	Hexose transporter (V. vinifera)	At5g26340
706088	3.6	4.0	3.6	–	Tropinone reductase (Arabidopsis)	At1g07440
800516	2.5	3.7	3.8	–	Hypothetical protein (Arabidopsis)	At4g27450
175325	2.6	3.4	2.9	–	Phosphoglycerate dehydrogenase-like protein (Arabidopsis)	At4g34200
266069	2.2	2.1	3.1	–	S-Adenosyl-methionine decarboxylase uORF (O. sativa)	At5g15950
595857	4.5	25.7	–	–	PR-1 protein (V. vinifera)	At2g14580
711753	7.5	23.8	–	×	Disease resistance protein dirigent-like protein (P. sativum)	At1g64160
746640	6.2	14.4	–	×	Acidic chitinase (class III), PR-3 (M. truncatula)	At5g24090
732511	2.1	13.1	–	–	Heavy metal transport/detoxification protein (M. truncatula)	At3g07600
792358	9.6	12.7	–	–	Hypothetical protein (dirigent-like protein; Arabidopsis)	At1g58170
567801	3.5	10.4	–	×	No hit	No hit
748543	6.1	10.3	–	–	PR-1 protein (Arabidopsis)	At2g14610
641396	3.2	9.6	–	–	Cyt P450, E-class P450, group I (M. truncatula)	At1g12740
714634	5.3	9.2	–	–	Hypothetical protein (P. deltoides × P. maximowiczii)	At4g19950
722643	3.1	9.0	–	×	Gln-dependent Asn synthetase type II (P. vulgaris)	At3g47340
290737	3.1	8.3	–	–	Protein disulfide isomerase-like protein, PDI-3 (C. melo)	At1g60420
712863	6.5	4.0	–	×	UDP-Glc transglucosylase-like protein, SlUPTG1 (L. esculentum)	At3g02230
554725	2.2	6.3	–	–	RLK LRR protein (S. tuberosum)	At1g34210
755695	2.9	6.2	–	–	NIMIN1 (Arabidopsis)	At1g02450
710544	6.0	4.8	–	–	Senescence-associated gene 21, SAG21 (Arabidopsis)	At4g02380
823332	3.6	5.7	–	–	ATP-binding cassette transporter (PtrATH2; Arabidopsis)	At3g47730
259624	3.0	5.5	–	–	Dicyanin (blue copper-binding protein; L. esculentum)	At5g20230
829298	2.7	5.1	–	–	Peroxidase (P. trichocarpa)	At3g49120
593975	2.3	5.2	–	–	Dolichyl-diphosphooligosaccharide-protein glycotransferase, DGL1 (Arabidopsis)	At5g66680
673066	4.1	3.1	–	–	Reversibly glycosylated polypeptide (T. aestivum)	At5g15650
422974	4.1	2.5	–	–	ATP-binding cassette transporter, abc2 homolog (Arabidopsis)	At3g47780
575987	3.9	3.6	–	–	Hypothetical protein (Arabidopsis)	At3g22160
828962	3.2	3.9	–	×	GST (Caragana korshinskii)	At1g17180
659435	3.9	2.1	–	–	Wound-induced protein WI12 (Arabidopsis)	At3g10985
202273	3.8	3.4	–	–	LRR receptor-like Ser/Thr protein kinase, RLK5 (Arabidopsis)	At1g09970
646543	3.6	3.7	–	–	Hydroxymethylglutaryl-CoA synthase 2 (H. brasiliensis)	At4g11820
711719	3.7	2.9	–	–	Heat shock protein 70 (Cucumis sativus)	At5g42020
826060	3.1	3.6	–	×	Putative disease resistance protein (LRR; Arabidopsis)	At2g25470
822404	2.5	3.5	–	–	60S Acidic ribosomal protein PO (Euphorbia esula)	At2g40010
827148	2.4	3.5	–	–	Hypothetical protein, putative prenylated rab receptor (O. sativa)	At3g13710
758330	2.8	3.5	–	–	Dirigent-like protein pDIR9 (Picea engelmannii × Picea glauca)	At1g55210
418525	3.5	2.5	–	–	Pro-rich lipid-binding protein (G. max)	At2g10940
586871	3.4	2.2	–	–	No hit	No hit
818818	2.7	3.3	–	×	Heat shock protein 90, SHD (Arabidopsis)	At4g24190
830497	3.2	3.1	–	×	Putative membrane protein (S. tuberosum)	At1g36580
580121	2.2	3.1	–	×	Hypothetical protein (Arabidopsis)	At4g21700
726993	3.1	3.1	–	–	Quinone reductase (Arabidopsis)	At4g27270
793953	2.7	3.1	–	–	Cyt P450 (Persea americana)	At4g13290
823674	2.5	3.1	–	–	Hevein-like (P. tremula × P. alba)	At3g04720
837012	3.0	2.2	–	–	Phosphogluconate dehydrogenase (Arabidopsis)	At3g02360
822983	2.1	–	9.5	–	Ubiquitin-conjugating enzyme (Arachis hypogaea)	At1g78870
712259	2.3	–	6.9	–	Developmental protein DG1118 (Arabidopsis)	At1g73030
556400	2.5	–	6.1	–	40S Ribosomal protein S19 (Arabidopsis)	At5g61170
829835	2.1	–	5.1	–	trans-Caffeoyl-CoA 3-O-methyltransferase, CCoAMT-2 (P. trichocarpa)	At4g34050
729723	2.2	–	4.8	–	CuZn-superoxide dismutase (P. alba × P. tremula)	At1g08830
580782	2.2	–	4.8	×	14-3-3 protein (Populus × canescens)	At3g02520
558203	2.4	–	4.1	–	Translation elongation factor (Arabidopsis)	At5g19510
721294	2.5	–	4.0	–	Cytochrome c1 (Arabidopsis)	At5g40810
831066	2.7	–	4.0	–	Monosaccharide transporter (P. tremula × P. tremuloides)	At1g11260
256724	2.6	–	3.9	–	Cytosolic l-ascorbate peroxidase (Codonopsis lanceolata)	At3g09640
830146	2.0	–	3.9	×	Ribosomal protein 60S L3 (S. tuberosum)	At1g43170
552351	2.6	–	3.8	–	Acidic ribosomal protein 60S P1 (Zea mays)	At1g01100
675956	2.4	–	3.6	–	Cytosolic l-ascorbate peroxidase (G. max)	At3g09640
575698	2.6	–	3.5	–	Enolase (O. sativa)	At2g36530
770546	2.6	–	3.5	–	Ribophorin (Arabidopsis)	At2g01720
834126	3.3	–	2.1	–	Pentameric polyubiquitin (Nicotiana sylvestris)	At4g02890
553829	3.3	–	2.9	–	Pectin methylesterase (P. tremula × P. tremuloides)	At3g14310
724015	3.1	–	3.1	–	S-Adenosyl-methionine synthetase (P. trichocarpa × P. deltoides)	At1g02500
654817	–	11.6	3.2	–	Ascorbate oxidase (L. esculentum)	At4g39830
727577	–	9.4	3.1	–	PR protein (P. deltoides × P. maximowiczii)	At1g24020
667694	–	7.9	3.2	–	Sinapyl alcohol dehydrogenase-like protein (P. tremula × P. tremuloides)	At4g37970
827390	–	7.3	7.5	–	PR protein (P. deltoides × P. maximowiczii)	At1g24020
728031	–	5.5	4.9	–	Nitrate-induced NOI protein (Arabidopsis)	At5g55850
421939	–	3.6	3.2	–	Zinc finger DNA-binding protein (C2H2 domain; Catharanthus roseus)	At1g27730
588763	–	2.5	4.1	–	Ethylene response factor, AP2 family, EREBP (Arabidopsis)	At5g47230
669494	4.1	–	–	×	Lipoprotein A, expansin, blight-associated protein p12 (O. sativa)	At4g30380
565873	3.91	–	–	×	Protein disulfide isomerase, PDI-3 (Quercus suber)	At1g60420
569255	3.84	–	–	×	Fe(II)/ascorbate oxidase (Arabidopsis)	At4g10500
795101	3.55	–	–	×	Protein disulfide isomerase-like protein, PDI-3 (C. melo)	At1g60420
586583	3.42	–	–	×	Acidic chitinase, PR-3 (M. truncatula)	At5g24090
670783	3.16	–	–	×	No hit	No hit
751998	–	28.7	–	×	β-1,3-Glucanase, PR-2 (Fragaria × ananassa)	At4g16260
831333	–	8.5	–	×	Chitinase, basic, PR-8 (Elaeagnus umbellata)	At3g12500
680604	–	7.6	–	×	Cyt P450 E-class, group I (M. truncatula)	At1g12740
760140	–	7.3	–	×	Syringolide-induced protein B13-1-1 (G. max)	At4g39830
824336	–	6.6	–	×	Formate dehydrogenase (Quercus robur)	At5g14780
802053	–	6.1	–	×	β-Galactosidase (Fragaria × ananassa)	At3g13750
832905	–	5.1	–	×	Cyt P450 (Cicer arietinum)	At4g37370
579371	–	3.7	–	×	Hypothetical protein (zinc finger; G. max)	At4g16380
829431	–	3.7	–	×	Heat shock protein 70 (C. sativus)	At5g42020
816882	–	3.7	–	×	α-Mannosidase (Arabidopsis)	At3g26720
826206	–	3.6	–	×	Taxadiene 5-α hydroxylase, Cyt P450 (O. sativa)	At5g36110
729244	–	3.1	–	×	Peroxidase precursor (Q. suber)	At5g05340
836795	–	–	4	×	F0F1-ATPase γ-subunit (Ipomoea batatas)	At2g33040
822499	–	–	3.1	×	ras-related protein RAB8-1 (N. tabacum)	At3g46060
676990	–	–	3.1	×	Hypothetical protein (Arabidopsis)	At3g29780

^aProtein ID number of corresponding best gene model in the P. trichocarpa ‘Nisqually-1’ genome sequence (version 1.1) at http://genome.jgi-psf.org/Poptr1_1/Poptr1_1.home.html.

^bBest database match and corresponding species obtained with a BLASTX query at NCBI.

^cAGI number of best Arabidopsis homolog.

DISCUSSION

Reactions that lead to programmed cell death in incompatible interactions between plant and hemi-biotrophic or necrotrophic pathogens preventing the pathogen to spread in plant tissues have been largely described in several plant pathosystems (Heath, 2000). In contrast, interactions involving biotrophic pathogens, with their sophisticated type of pathogenesis that keeps plant cells alive and minimizes tissue damage in susceptible hosts, are poorly known. The uredinial stage of rust fungi is generally taking place through stomatal penetration, and most studies suggest that host compatibility requires the ability for the fungus to avoid or negate prehaustorial defenses within the substomatal cavity of the host leaf, breach the mesophyll plant cell wall to form the first haustorium, and develop a biotrophic interaction with the living invaded cell to support further fungal growth (Schulze-Lefert and Panstruga, 2003; Glazebrook, 2005; Spanu, 2006). Recent work described haustorially expressed secreted proteins of the rust biotrophic pathogen M. lini (Catanzariti et al., 2006) that led to HR when expressed in planta, indicating a probable direct R-Avr recognition system in flax challenged by an avirulent strain of M. lini (Dodds et al., 2006).

In this study, we describe at the microscopic, histological, and transcriptomic levels a novel pathosystem involving P. trichocarpa × P. deltoides ‘Beaupré’ challenged by urediniospores of the leaf rust basidiomycete M. larici-populina. ‘Beaupré’ is resistant to M. larici-populina isolate 93ID6 (pathotype 3-4) and susceptible to isolate 98AG31 (pathotype 3-4-7; Barrès et al., 2006). Unexpectedly, there was no significant difference in fungal growth from spore germination to contact with mesophyll cell between these two isolates. Spore germlings of both isolates were able to penetrate the leaf through stomatas in the first hours after inoculation, forming primary infection structures (i.e. substomatal vesicles) in the spongy mesophyll and reaching mesophyll cells for infection. Appressoria were formed on the leaf surface but are not a prerequisite to penetrate inside plant tissues, as frequent direct hyphal penetration through stomatas were observed. Based on the amount of fungal rDNA, differences of growth in planta were only noticeable between compatible and incompatible isolates after 24 hpi (Fig. 3). The compatible isolate then spread in leaf tissues and colonized the whole mesophyll, while the incompatible strain remained in the spongy mesophyll (Fig. 2, F and H). Fungal growth increased until the compatible strain developed spore-forming cells and produced typical golden pustules filled with masses of urediniospores on the lower leaf surfaces between 6 and 7 dpi. At this stage, the resistant phenotype was generally characterized by the presence of scattered necrotic lesions and the absence of macroscopic symptoms. Similar responses have been reported in P. deltoides × Populus nigra ‘Ogy’ inoculated with virulent and avirulent isolates of M. larici-populina (Laurans and Pilate, 1999).

We were not able to detect H₂O₂ accumulation through DAB staining in leaf tissues challenged by the incompatible strain of M. larici-populina. H₂O₂ production possibly occurred transiently and only in plant cells challenged by M. larici-populina during infection. Supporting this assumption, several genes encoding enzymes of the redox regulation pathways such as GSTs, ascorbate peroxidases, and superoxide dismutase were highly up-regulated at 48 hpi. Studies conducted at the protein level confirmed the up-regulation of thioredoxin and peroxiredoxin during Populus-Melampsora interaction (Rouhier et al., 2004; Vieira Dos Santos et al., 2005). The lack of H₂O₂ accumulation has been previously reported in plants interacting with biotroph pathogens (Glazebrook, 2005). This may reflect the specific and complex biotrophic relationships between rust and living host cells. As observed by Laurans and Pilate (1999) and in this study, the necrotic tissues are highly localized to a limited area (Fig. 1), supporting the fact that H₂O₂ production related to HR was not spreading far from infection sites in leaf tissues.

Phloroglucinol staining confirmed that a massive production of monolignols was induced upon inoculation of plant tissues by the incompatible rust strain (Fig. 4). Such compounds are believed to play a role in plant defense (Dixon, 2001). Several genes of the phenylpropanoid pathways and dirigent-like proteins encoding genes were induced at 48 hpi in inoculated leaves prior to observation of the maximum level of phloroglucinol staining. Strong production of secondary metabolites in colonized leaves likely led to the synthesis of phytoalexins and lignin deposition in secondary cell walls restricting the fungal proliferation, as shown in the cowpea (Vigna unguiculata)-Uromyces vignae interaction and in other pathosystems involving biotrophic fungal pathogens (Heath, 1997, 2000). Callose synthase and Phe ammonia-lyase genes are often reported as marker genes of lignin and callose deposition, although the levels of induction may strongly vary from one interaction to another. In this study, the homologs of Populus Phe ammonia-lyase and callose synthase were not significantly induced during the incompatible interaction at the time points tested. Induction of dirigent proteins encoding transcripts has been reported in conifers submitted to wounding or insect attacks (Ralph et al., 2006b), and their products may participate in a general stress response in trees submitted to biotic or abiotic stress.

Studying the transcriptome of rust-infected leaves with whole-genome oligoarray harboring more than 45,000 putative gene models from the P. trichocarpa genome sequence (Tuskan et al., 2006) identified 2,397 rust-responsive genes (approximately 3% of the total set of arrayed genes) that may play a role in defense reactions in the incompatible interaction or, conversely, in supporting fungal growth in susceptible plants. As expected from previous studies in model species (Schenk et al., 2000; Mahalingam et al., 2003), many functional groups of genes were found to be involved in the defense response, including signal transduction pathways components, genes stimulated during biotic or abiotic stress responses, and genes of primary and secondary metabolisms. A striking alteration in steady-state RNA populations took place at 48 hpi, whereas few genes were regulated at earlier stages of interaction when tested by RT-qPCR or cDNA macroarrays. This suggests that gene expression in response to rust infection is activated later after the fungal cells entered into contact with the leaf surface and colonized the substomatal cavity. Recognition of the pathogen is likely taking place upon triggering of specific recognition mechanisms when fungal hyphae are elongating within the mesophyll and attempt to penetrate the plant cell wall barrier to form haustorial structures (Fig. 2). Presence of strongly inducible genes in the incompatible interaction indicates that the rust infection has had a significant impact on the leaf transcriptome despite the limited amount of fungal mycelium (Fig. 3) during the early stages of invasion. The observed differences between the transcriptomes of leaves infected by either the compatible or incompatible isolates were striking. The incompatible strain elicited up-regulation of a wide range of defense-related genes (e.g. PR proteins, GSTs), whereas these genes were not induced by the compatible strain.

Several genes that may participate in the perception of the rust pathogen by sensing avirulence products released by invading hyphae were differentially accumulated in the incompatible interaction between P. trichocarpa × P. deltoides and M. larici-populina. We identified transcripts coding for putative LRR disease resistance proteins (protein ID nos. 645750 and 826060) that showed an induction of their expression (Table IV). Transcripts coding for LRR receptor protein kinases (protein ID nos. 417599 and 171587), showing homology with RLK5 and PERK1, respectively, were also induced during the incompatible interaction. PERK1 encodes a putative receptor kinase with an extracellular domain with sequence identity to cell wall-associated extensin-like proteins. PERK1 transcripts are rapidly expressed upon mechanical wounding in response to Sclerotinia sclerotiorum and SA and methyl jasmonate application (Silva and Goring, 2002). RLK5, also named HAESA, is involved in the control of floral abscission in Arabidopsis (Haffani et al., 2004). These receptor kinases belong to a large family that had been shown to play a role in microbial sensing, and they may likely participate in pathogen perception in poplar. In flax, the M. lini products of the avirulence gene Avr567 are expressed in haustoria and recognized inside plant cells (Dodds et al., 2004). This recognition occurs through a direct interaction with the L resistance genes products of flax (Dodds et al., 2006). L genes are members of the NBS-LRR resistance gene family (Ellis et al., 1999). We did not identify any members of this family among rust-induced genes. These plant proteins are responsible for the detection of pathogens in natural conditions and should be constitutively expressed; thus, their expression may not necessarily be under transcriptional control during the infection process. None of the rust-induced putative receptors and LRR proteins is located in the superclusters of NBS-LRR R genes containing the MER locus (Tuskan et al., 2006) or close to the rust-resistance loci identified in poplar by genetic mapping (Cervera et al., 2004; Lescot et al., 2004; Jorge et al., 2005). Strikingly, a significant induction of RLK transcripts was also detected in the compatible interaction at 48 hpi (2.7-fold).

Upon recognition of the pathogen aggression, a complex network of signaling enzymes and molecules relays the information in the plant cell to the nucleus, where specific defense-related gene expression is triggered. As shown in several other plant-fungus interactions, components of the signaling pathways were induced in the incompatible interaction between P. trichocarpa × P. deltoides ‘Beaupré’ and M. larici-populina. These include transcripts from the calcium- and ethylene-related pathways, calmodulin, calreticulin, 1-aminocyclopropane-1-carboxylate oxidase, 14-3-3 proteins, and ethylene-responsive element-binding protein (EREBP) and MYB family transcription factors. NPR1 is an important regulator of PR gene expression through binding to transcriptional regulators TGA elements and a low accumulation (approximately 2-fold) of its transcript has been reported in different pathosystems (Glazebrook, 2005). P. trichocarpa × P. deltoides ‘Beaupré’ NPR1 transcript (protein ID no. 253241) was accumulated at 24 and 48 hpi in poplar leaves during the incompatible interaction (Fig. 5) prior the strong induction observed for targeted PR genes. Interestingly, we also detected the induction of a transcript coding for NPR1/NIM1-interacting protein (NIMIN-1) at 48 hpi (Table IV). NIMIN-1 directly interacts with NPR1 and can modulate its activity and expression of PR genes in Arabidopsis (Weigel et al., 2005). A gene that showed among the highest levels of induction, using the different transcript profiling approaches, encoded an I3PS. The protein is involved in the production of inositol-P, a metabolite that could lead to the production of various products such as phosphatidylinositides, cell wall components, or oligosaccharides of the raffinose series (Loewus and Murthy, 2000). Phosphatidylinositols, like inositol 1,4,5-trisphosphate, are important secondary messengers of the cell transduction pathways that play a crucial role in calcium homeostasis in plant cells (Munnik et al., 1998). Involvement of calcium-related signaling in plant-rust interaction has been previously reported (Xu and Heath, 1998). Several calcium-binding proteins encoding genes are induced in poplar leaves during the incompatible interaction with M. larici-populina, and I3PS may contribute to the production of phosphatidylinositols involved in calcium regulation. Considering the strong induction of I3PS transcript in poplar leaves during the incompatible interaction with M. larici-populina, addressing the exact role of inositol-P in response to either biotic or abiotic stress requires further investigation.

Within transcripts with the highest rust-induced accumulation (>10-fold) in the incompatible interaction, several encoded PR proteins, such as PR-1, PR-2 (1,3-β-glucanase), PR-3 (acidic chitinase), PR-5 (thaumatin-like protein), PR-8 (basic chitinase), and PR-10 (ribonuclease). All these enzymes are known for their antifungal properties (Van Loon and Van Strien, 1999) and are typical SA-induced marker genes of plant response to bacterial and fungal attacks (Van Loon and Van Strien, 1999; Schenk et al., 2000). There is evidence that Uromyces rust species are susceptible to apoplastic PR proteins, including PR-1 (Rauscher et al., 1999). Interestingly, a gene encoding a PR-10 protein was also activated in epidermal cells of resistant cowpea challenged by the cowpea rust fungus and prior of the fungus entering the cell lumen (Mould et al., 2003). We identified many expressed hypothetical proteins among rust-responsive genes. Of interest is the RISP whose transcript showed the strongest (32-fold) induction in the incompatible interaction in the whole-genome oligoarray analysis. The RISP may play a role in early defense against M. larici-populina, and it remains to address its exact role to determine whether it is a novel PR protein in Populus.

In a compatible biotrophic interaction, the invading hyphae are able to alter the host-plant metabolism in such a way that increasing amounts of nitrogen and carbon metabolites are mobilized and translocated to fungal cells (Mendgen and Hahn, 2002; Panstruga, 2003; Both et al., 2005). In the compatible interaction between P. trichocarpa × P. deltoides ‘Beaupré’ and M. larici-populina, we did not detect a striking induction of metabolism-related elements, and only a few genes encoding transporters and enzymes of the carbon metabolism were slightly induced with the cDNA-macroarray experiment. We observed a 2.5-fold induction of a Populus transcript encoding a Δ¹-pyrroline-5-carboxylate synthetase (protein ID no. 421059; Table II) in the compatible interaction, whereas no induction of this gene was observed in the incompatible interaction. This enzyme catalyzes the two first steps of Pro synthesis. The fis1 transcript encodes a Δ¹-pyrroline-5-carboxylate dehydrogenase, which is specifically expressed in the flax-M. lini compatible interaction (Ayliffe et al., 2002). FIS1 is involved in the catabolism of Pro to Glu, and a possible link between such catabolic activity and the fungal metabolism remains unclear.

In conclusion, the rust-responsive genes from P. trichocarpa × P. deltoides ‘Beaupré’ presented here are a valuable resource for further functional genomics studies addressing mechanisms of durable resistance in a perennial species, Populus, and other Salicaceae. Comparative analysis of compatible and incompatible interactions showed the stimulation of several known genes involved in plant defense reactions to biotrophic pathogens like PR proteins targeted by plant recognition systems (i.e. R genes). New candidate genes, such as I3PS and RISP, which may participate to a specific Populus response to rust, were also detected. The accumulation of most rust-responsive transcripts occurred at a late stage (48 hpi) when fungal hyphae penetrate the mesophyll cells, although a few transcripts were induced at earlier time points. It appears that a perennial species, such as Populus, does not use specific arrays of defense proteins. Expansion of NBS-LRR genes as well as PR protein gene families in the Populus genome (Tuskan et al., 2006) may underlie specific recognition systems for an efficient targeting of certain PR proteins within expanded families. Further analyses, including gene inactivation, will address whether the observed quantitative differences in gene expression between the compatible and incompatible infections are sufficient to explain the drastic difference in the interaction of both M. larici-populina strains with ‘Beaupré’. Rust-responsive genes will be used in genetic studies, including ecotilling, comparing qualitative and quantitative resistances to rust in various poplar cultivars to help in marker-assisted breeding of new genotypes for durable resistance against M. larici-populina.

MATERIALS AND METHODS

Plant Materials and Growth Conditions

All experiments were performed on rooted cuttings of the hybrid poplar (Populus trichocarpa × Populus deltoides) ‘Beaupré’. For the analysis of compatible and incompatible poplar-rust interactions, P. trichocarpa × P. deltoides ‘Beaupré’ plants were grown for 12 weeks in a greenhouse from dormant cuttings in 5-L pots containing a sand-peat (50:50, v/v) mixture, with an initial fertilization of 1.45 g L⁻¹ CaO and 6 g L⁻¹ of slow release 13:13:13 N:P:K fertilizer (Nutricote T 100; Fertil). ‘Beaupré’ plants were watered daily with deionized water under 16-h/8-h photoperiod in greenhouse conditions with supplemental artificial light to complement to a minimum illumination of 200 μmol s⁻¹ m⁻² during the winter season. After 12 weeks, young trees were >1 m high and presented 10 to 14 fully expanded leaves.

Inoculation Procedures

Two isolates of Melampsora larici-populina were used in this study: the virulent 98AG31 (pathotype 3-4-7) and avirulent 93ID6 (pathotype 3-4) isolates (Barrès et al., 2006). The urediniospores of the two M. larici-populina isolates were multiplicated on detached leaves of P. deltoides × Populus nigra ‘Robusta’, which is susceptible to all M. larici-populina strains and collected at 10 dpi. The spores were dried and were kept in a dry atmosphere at 1°C. Fully expanded leaves from leaf plastochrony index (LPI) 5 to 9 were detached from several ‘Beaupré’ plants and spray inoculated on their abaxial surface with an urediniospore suspension in water-agar (0.1 g L⁻¹) adjusted to 100,000 urediniospores mL⁻¹ or with water-agar as a control (mock-inoculated leaves). Inoculations were done by pooling three leaves of different LPI from different plants (i.e. LPI 5 from plant 1, LPI 8 from plant 2, and LPI 9 from plant 3) for each treatment and each time point. The inoculated leaves (i.e. 27 leaves in total) were then incubated with the abaxial surface uppermost, floating on deionized water in 22.5- × 22.5-cm petri dishes, at 19 ± 1°C under continuous artificial illumination (fluorescent light, 25 μmol s⁻¹ m⁻²), for various durations. The material harvested at different time points in the different treatments consisted of 7 cm² (30-mm diameter) leaf discs randomly sampled on the overall leaf surface. The leaf discs were immediately snap-frozen in liquid nitrogen and transferred to −80°C until further analysis.

Microscopy Analysis and Scanning Electron Microscopy Analysis

Fungal infection structures at leaf surface were observed by clearing inoculated leaf discs (control, compatible, and incompatible) in boiling ethanol (70% v/v) for 10 min in a water bath, followed by 10 min incubation in an aniline blue solution (10 mg/mL; Sigma-Aldrich) and washes in distilled water. Observations by light microscopy were carried out at a magnification of 400× on a OPTIPHOT system (Nikon).

To obtain a high density of M. larici-populina urediniospores on the abaxial surface of poplar leaves for scanning electron microscopy observation, dry inoculations were performed with an air pistol (DIANA model 3; Mayer and Grammelspacher) with about 1 mg of spores (approximately 4 × 10⁵ spores) inoculated for each shot. For each combination of time point (2, 6, 12, 24, 48, 96, 120, and 192 hpi) × treatment (compatible and incompatible interactions), two leaves were inoculated, and observations were carried out on three leaf fragments of 1 cm², snap-frozen immediately after harvesting. Samples were fractured in liquid nitrogen and were attached to aluminum stubs on a Peltier stage (−50°C). They were then examined under a variable pressure scanning electron microscope (model 1450VP; Leo). Backscattered secondary electron images were observed at an accelerating voltage of 15 kV, a working distance of 10 mm, and at a pressure chamber of about 30 Pa. Digital images of samples (abaxial side and transversal section) were captured using the microscope software and edited with Adobe Photoshop CS2 (Adobe Systems France SAS) to adjust brightness and contrast or to artificially paint fungal hyphae colonizing the leaf tissues.

ROS and Lignin Monomers Detection

Production of ROS was investigated in inoculated leaf discs (ø 30 mm) using the DAB technique according to Thordal-Christensen et al. (1997). Leaf discs were incubated in a solution containing 0.2% (w/v) of 3,3-diamino-benzidine tetrahydrochloride (Sigma-Aldrich) in a shaking bath for 24 h at 25°C in the dark. Other infiltration procedures (e.g. vacuum-forced infiltration) were also tested. An evapotranspiration system was set with complete leaves where petioles were standing in DAB solution, as described by Orozco-Cardenas and Ryan (1999), for up to 24 h. Then, leaf discs were cleared in boiling ethanol (70%, v/v) in a water bath for 10 min and washed in water. Positive controls for DAB coloration were performed by injection of several H₂O₂ solutions of various concentrations (0.5%–20%; Sigma-Aldrich) through syringes on uninfected leaves and by wounding tissues of both noninoculated and inoculated leaves. Magnification for observation was of 100×. All observations were carried out on three replicates of each treatment (control, compatible, and incompatible) at 2, 6, 12, 24, 48, and 96 hpi. Light microscopy was carried on an OPTIPHOT system (Nikon).

Lignin monomer production was examined using the Wiesner coloration procedure with standard protocol according to Nakano and Meshitsuka (1992). Control and inoculated leaf discs were incubated in a phloroglucinol solution (2% w/v) for 5 min in the dark. Disks were then incubated in a 6 n HCl solution for 5 min, rinsed, and kept in water for further observations. Observations were carried out on replicates at 2, 6, 12, 24, 48, and 96 hpi and captured with a digital camera Sony F707 (Sony).

DNA and RNA Extraction

Total DNA was extracted from leaf tissues with the DNeasy Plant Mini kit (Qiagen) from 100 mg of frozen (−80°C) material. RNA was removed by the addition of ribonuclease A during extraction. DNA quality was verified by electrophoresis on agarose gel, and DNA quantity was measured by spectrophotometry (Sambrook et al., 2001).

Total RNA extraction was performed with the RNeasy Plant Mini kit (Qiagen) from 100 mg of pooled (−80°C) foliar discs harvested from leaves of various LPI and various individual poplar plants for each treatment considered. Pooling of samples from different trees and LPI helped in minimizing the variations between individual RNA samples. Extraction from leaf tissue was modified as described in Kohler et al. (2004), and DNase I (Qiagen) treatment was included in the RNA extraction procedure according to the manufacturer's instructions to eliminate traces of genomic DNA. Quality and quantity of RNA samples were checked by electrophoresis on agarose denaturing gel and by spectrophotometry for cDNA library construction and cDNA array experiments (Sambrook et al., 2001), while quality and quantity of RNA samples used for hybridization with NimbleGen Populus arrays were assessed on the RNA analyzer Experion (Bio-Rad) following the manufacturer's recommendation.

Amplification of rDNA ITS by qPCR

Development of the compatible and incompatible rust strains was followed in planta by specific amplification of the nuclear rDNA ITS on total DNA extracted from inoculated leaf tissues (Boyle et al., 2005). The same amount of 100 ng DNA was used for each qPCR amplification. Amplifications were performed in 1× iQ SYBR Green Supermix (Bio-Rad) with 0.3 μm of specific 5′- and 3′-primers for M. larici-populina (GenBank accession no. AY375268; M. larici-populina primers ITS-Mlp-F, GAGCGCACTTTAATGTGACTC; ITS-Mlp-R, ACTTAATTAAGTTGATAGGG; P. Frey, C. Husson, and J. Pinon, unpublished data) with a MJ-opticon2 DNA engine (Bio-Rad). Assuming a signal intensity proportional to amplified ITS sequences, we considered the pathogen growth as the relative difference (2^−ΔCt) of fungal ITS amplicons calculated for compatible or incompatible interactions compared to mock-inoculated tissues at 2, 6, 12, 24, 48, and 96 hpi. Amplifications were carried out in triplicates, and a statistical analysis (t test) based on ΔCt curves was performed with the software Statview 5.0 (SAS Institute) for Mac OS 9. A standard curve was drawn for conversion of Ct values to pathogen DNA mass (Boyle et al., 2005).

Sequencing of cDNAs from SSH Library

Double-stranded cDNAs corresponding to mRNAs expressed in P. trichocarpa × P. deltoides ‘Beaupré’ leaves upon infection with the incompatible strain of M. larici-populina at 12, 24, and 48 hpi (tester probe; 1/3, 1/3, 1/3) and cDNAs from ‘Beaupré’ mock-inoculated leaves at 12, 24, and 48 hpi (driver probe; 1/3, 1/3, 1/3) were separately obtained by using the SMART-PCR cDNA Synthesis kit (BD Biosciences). The mixed-tester cDNA pool was subtracted by the mixed-driver probe (SSH) following the manufacturer's instructions (PCR-Select cDNA Subtraction kit; BD Biosciences; Diatchenko et al., 1996). These subtracted cDNAs were subcloned in pGEM-T plasmids (Promega) and were used to transform Escherichia coli DH10b bacteria (Invitrogen).

Purification by rolling circle amplification and Dye Terminator sequencing of plasmid DNA from SSH clones were performed at the GENOSCOPE (Centre National de Séquençage) on ABI3730xl DNA analyzers (Applied Biosystems). Raw sequence data was edited using SEQUENCHER 4.2 (Gene Codes) for Mac OS X. Leading and trailing vector and polylinker sequences were removed by SEQUENCHER filters. Groups of sequences were assembled into clusters using the contig routine of SEQUENCHER and parsed using dedicated Perl scripts.

NimbleGen Populus Expression Oligoarrays

The P. trichocarpa whole-genome expression oligoarray version 2.0 (NimbleGen Systems) consisted of 65,965 probe sets corresponding to 55,970 gene models predicted on the P. trichocarpa genome sequence version 1.0 and 9,995 aspen cDNA sequences (Populus tremula, Populus tremuloides, and P. tremula × P. tremuloides). The Populus version 2.0 oligoarray (S. DiFazio, A. Brunner, P. Dharmawardhana, and K. Munn, unpublished data) is fully described in the platform GPL2699 stored in the Gene Expression Omnibus (GEO) at NCBI (http://www.ncbi.nlm.nih.gov/geo).

For hybridization with whole-genome oligoarray, a series of three replicates was obtained from mock-inoculated P. trichocarpa × P. deltoides ‘Beaupré’ leaves (control) and leaves infected with either compatible or incompatible strains of M. larici-populina at 48 hpi. Preparation of samples, hybridization procedures, and data acquisition and normalization were performed at the NimbleGen facility (NimbleGen Systems) following the manufacturer's procedures. Average expression levels were calculated for each gene from the independent probes on array and were used for further analysis. Log2-transformed data were calculated and were subjected to the CyberT statistical framework (http://www.igb.uci.edu/servers/cybert/; Long et al., 2001; Baldi and Hatfield, 2002) by using the ‘bayesreg’ script in the R statistical package (http://www.r-project.org/) with the following parameters: bayesT(aData,3,3,TRUE,1,TRUE,101,9). Statistical analyses were conducted separately for compatible and incompatible interactions versus mock-inoculated tissues in a simple paired data comparison model with all gene probe duplicates considered independently. Gene probes that showed a mean intensity close to the background level (i.e. lower than 200) in all experimental conditions were removed. Transcripts for which probe duplicates both showed a P value lower than 0.05 were considered as significantly regulated. A subset of transcripts that showed a 3-fold increase or 3-fold decrease in abundance in one treatment compared to the control was then selected. All materials and procedures descriptions comply MIAME standards set for array data (Brazma et al., 2001). The complete expression dataset is available as series accession number GSE7098 in the GEO at NCBI.

cDNA Microarrays

The PICME (http://www.picme.at/) Populus microarray, composed of 28,000 elements, including 23,500 cDNAs, is described in the platform GPL4874 stored in the GEO at NCBI. This set of cDNAs corresponds to approximately 10,000 different predicted gene models in the P. trichocarpa genome sequence (Tuskan et al., 2006). For hybridization with PICME microarrays, a series of three replicates was obtained from mock-inoculated P. trichocarpa × P. deltoides ‘Beaupré’ leaves (control) and leaves infected with the incompatible strain of M. larici-populina at 48 hpi. Sample details and hybridization procedures are fully described in sample series GSM162978 and GSM162979 deposited in the GEO at NCBI. Microarray images were acquired on an Axon GenePix 4000B scanning device (DIPSI) at a resolution of 10 μm. Quantitation of signals was done with the GenePix Pro 5 software (Axon, DIPSI) with automatic detection of background levels by block and lowess selected for normalization of signals. Mean intensities in each channel were used for quality control, and log-transformed ratios were calculated on the ratios of mean signals. Log-ratios were then submitted to a t test using the CyberT Web site. Only two replicates were retained for the statistical analysis, as one replicate suffered from postprocessing hybridization problems. Based on the statistical analysis, a gene was considered significantly up- or down-regulated if the t test associated Bayes-lnP value was lower than 0.01 and the ratio ≥3-fold. All materials and procedures descriptions comply with MIAME standards set for array data (Brazma et al., 2001). The complete expression dataset is available as series accession number GSE7049 in GEO at NCBI.

cDNA Macroarrays

The P. trichocarpa × P. deltoides ‘Beaupré’ macroarray, composed of 4,600 cDNA, is fully described in the platform GPL4887 stored in the GEO at NCBI. For hybridization with cDNA macroarrays, a series of three independent biological replicates was obtained from mock-inoculated P. trichocarpa × P. deltoides ‘Beaupré’ leaves (control) and leaves infected with either compatible and incompatible strains of M. larici-populina at 12, 24, and 48 hpi. Hybridization, image acquisition, and analysis were performed as previously described (Duplessis et al., 2005; Gupta et al., 2005). Data quality assessment was performed with the Cyber-T Web interface. Based on the statistical analysis, a gene was considered significantly up- or down-regulated if the t test associated ln-P value was lower than 0.05 in at least one treatment at any time point. For the final analysis of expression patterns, fold-change of gene expression in compatible or incompatible interaction compared to mock-inoculated leaves were averaged, and genes having their expression falling between 0 and 1 were multiplied by −1 and inverted to facilitate their interpretation. All materials and procedure descriptions comply with MIAME standards set for array data (Brazma et al., 2001). The complete expression dataset and samples details are available in series GSE7121 described in the GEO at NCBI.

RT-qPCR Analysis

To allow the amplification of specific transcripts by RT-qPCR, we designed primers from the P. trichocarpa gene models coding for the PR proteins PR-1, PR-5, and PR-10 (protein ID nos. 550049, 669475, and 827390, respectively), I3PS (protein ID no. 832275), NPR1 (protein ID no. 253241), dirigent-like protein (protein ID no. 711753), RISP (protein ID no. 678883), ribulose bisphosphate carboxylase oxygenase (protein ID no. 813777), PSI reaction center subunit IV (protein ID no. 711610), and arabinogalactan protein (protein ID no. 573930). The primers were designed in the coding sequence, and amplified fragments showed a length ranging between 160 and 271 nucleotides. The primers list is detailed in Supplemental Table S5. A BLASTN against the P. trichocarpa genome sequence was performed for each primer sequence to verify the absence of cross annealing in other regions of the P. trichocarpa genome sequence.

RNA samples used for RT-qPCR corresponded to an additional biological replicate (year 2005) to the ones used in the transcriptomic analyses mentioned above. First-strand cDNAs were synthesized from 1 μg DNase-treated total RNA using the iScript cDNA synthesis kit (Bio-Rad) in a total volume of 20 μL according to the manufacturer's instructions. Two microliters of RT products were amplified by PCR in 1× iQ SYBR Green Supermix (Bio-Rad) with 0.3 μm of specific 5′- and 3′-primers with a MJ-opticon2 DNA engine (Bio-Rad). The specific 5′ and 3′ primers (see Supplemental Table S5) were used, and the ubiquitin-specific primers used as a relative control were the same as described in Kohler et al. (2004; P. trichocarpa protein ID no. 732892; GenBank ID CA825222). Fold-changes in gene expression between inoculated and mock-inoculated poplar leaves were based on ΔCt calculation. ΔCt corresponded to Ct of one selected gene subtracted by Ct of ubiquitine, and fold-change expression was based on calculation of the ΔΔC(t) (Livak and Schmittgen, 2001) that corresponded to ΔC(t) in inoculated tissues subtracted by the ΔC(t) in mock-inoculated tissues.

Accession Numbers

Sequences from the SSH library described in this article can be retrieved in GenBank under accession numbers CT027996 to CT029994 and CT033829.

Supplemental Data

The following materials are available in the online version of this article.

• Supplemental Table S1. Complete list of transcripts sequenced from a cDNA SSH library of P. trichocarpa × P. deltoides ‘Beaupré’ leaves inoculated with the incompatible isolate 93ID6 of M. larici-populina.

• Supplemental Table S2. List of expression ratios of P. trichocarpa × P. deltoides ‘Beaupré’ transcripts measured with the NimbleGen P. trichocarpa whole-genome expression oligoarray at 48 hpi between ‘Beaupré’ leaves inoculated with compatible (C48) or incompatible (I48) strains of M. larici-populina and mock-inoculated (water) ‘Beaupré’ leaves at 48 hpi.

• Supplemental Table S3. List of expression ratios of P. trichocarpa × P. deltoides ‘Beaupré’ transcripts measured with the PICME Populus 28 K cDNA microarrays at 48 hpi between ‘Beaupré’ leaves inoculated with incompatible strain of M. larici-populina (I48) and mock-inoculated (water) ‘Beaupré’ leaves.

• Supplemental Table S4. Complete list of expression ratios of P. trichocarpa × P. deltoides ‘Beaupré’ transcripts measured with Nylon-based ‘Beaupré’ cDNA macroarrays at 12, 24, and 48 hpi during time-course infection of ‘Beaupré’ leaves with compatible (C) and incompatible (I) strains of M. larici-populina.

• Supplemental Table S5. Specific 5′ and 3′ primers designed for RT-qPCR amplification of Populus transcripts.