The arbuscular mycorrhiza is the name given to a symbiotic association of plants and arbuscular mycorrhizal (AM) fungi. The AM symbiosis occurs widely throughout the plant kingdom and involves angiosperms, gymnosperms, pteridophytes and some bryophytes. The fungi that participate in this symbiosis are all members of the Glomeromycota, [1–3], a likely sister group to the Ascomycota and Basidiomycota. The symbiosis develops in the plant roots, where AM fungi form extensively branched hyphae, called arbuscules, in the cortical cells. In addition to growth within the root, the fungus develops a network of extra-radical hyphae that extends into the rhizosphere. Inorganic phosphate (Pi) and nitrogen (N), acquired by the extra-radical hyphae, are translocated to the arbuscules and released to the plant. In return the plant provides the fungus with carbon. P and N are the two mineral nutrients that plants require in the greatest quantities and therefore the symbiosis has an important influence on plant health and consequently on ecosystem function [4–6].

There has been significant progress in understanding the signaling associated with the early stages of development of AM symbiosis [7]. Some of the signal molecules have been identified, including strigolactones, which act as pre-contact signals to stimulate AM fungal growth and metabolism [8, 9]. In addition, several components of a plant 'symbiosis signaling' pathway required for fungal entry into the root have been cloned [10–17]. These genes were identified initially in legumes, where the symbiosis signaling pathway is required also for the interaction with rhizobia. However, orthologs of these genes have been identified in tomato and rice, where they are required for AM symbiosis [18, 19]. This further supports the hypothesis that the symbiosis signaling pathway arose initially for interactions with AM fungi, and in legumes it was co-opted to enable interactions with rhizobia [19–21]. Currently, for AM symbiosis, the downstream genes controlled by this pathway are not known. In general, events associated with fungal development in the cortex are less well understood, and genes controlling arbuscule formation, have yet to be identified. Arbuscule development is an active process for both symbionts. The fungal hypha penetrates the cortical cell wall and undergoes repeated dichotomous branching, while the plant cell undergoes significant cellular changes, including reorganization of the cytoskeleton, endoplasmic reticulum and development of the peri-arbuscular membrane, which envelops the arbuscule [22–26]. Reorganization of the cytoskeleton and endoplasmic reticulum is initiated prior to arbuscule formation suggesting that a mobile signal may trigger these events [26, 27]. Currently, information about the lipid and protein content of the peri-arbuscular membrane is limited, but it is clear that the membrane contains some unique proteins, including symbiosis-specific Pi transporters that function to import Pi into the cortical cell [28]. The transporters involved in N acquisition are as yet unknown, but a series of nuclear magnetic resonance (NMR) spectroscopy and gene expression studies indicate that the fungus assimilates N in the external hyphae, translocates this to the intra-radical hyphae as arginine, and then disassembles the arginine to retain the carbon molecules, while providing N to the plant, probably as ammonium (NH₄⁺) [29–31]. Although not shown directly, it is likely that NH₄⁺ transfer to the plant occurs at the arbuscule-cortical cell interface also.

Arbuscules are terminally differentiated hyphae with a finite life and following development, they undergo a degeneration phase in which the arbuscule branches collapse, and the arbuscule gradually disappears from the cell [22, 32–34]. This process is not well understood, but it is clear that arbuscule development and degeneration occur cell autonomously and arbuscules in neighboring cells may be in different phases of the life cycle. Although the arbuscule-cortical cell interface is central to the AM symbiosis, currently, we understand little of the processes that enable its development or that control its degeneration.

To gain insight into the molecular changes that accompany development of AM symbiosis, a variety of transcript profiling approaches have been undertaken. In rice, surveys using a whole genome chip identified mycorrhiza-responsive genes and revealed a partial overlap in the transcriptional responses to mutualistic and pathogenic biotrophic fungi [35]. In Medicago truncatula, suppressive subtractive hybridization, cDNA arrays and long oligonucleotide arrays have been used to document transcript changes associated with AM symbiosis [36–40]. The inclusion of M. truncatula symbiosis mutants enabled gene expression to be associated with different phases of the symbiosis, for example appressoria formation [41]. A mycorrhizal root includes a heterogeneous mix of colonized and non-colonized cells, and therefore spatial gene expression information is particularly important. In situ hybridization and/or analysis of promoter-reporter fusions demonstrated that genes whose expression was induced in colonized roots could be further separated into two groups; those expressed exclusively in cells containing arbuscules and those expressed in cells containing arbuscules and in neighboring non-colonized cells [28, 36, 38, 42–46]. These data indicate that the different cortical cells of the mycorrhizal root system have unique transcript profiles, and point to at least two signaling pathways active in regulating mycorrhiza-induced gene expression. While these approaches have advanced information about gene expression in colonized cortical cells, in situ hybridization or promoter-reporter gene studies are labor intensive and this precludes surveys of a large number of genes. Our current knowledge of genes expressed in cells containing arbuscules is limited to less than 35 genes. In addition, the genes that have been analyzed are those that have the highest transcript levels and relatively few receptors or transcriptional regulators have been investigated. Beyond plant gene expression in the colonized cortical cell, almost nothing is known about gene expression in the arbuscule.

Laser microdissection (LM) offers a new way to obtain RNA from a subset of cells, and coupled with transcript profiling approaches, it provides an opportunity to monitor gene expression in an individual cell type more easily than promoter-reporter gene fusions. LM was originally developed to isolate specific animal cells from complex tissues; a breakthrough for cancer and neuronal research [47, 48]. In plants, LM was used initially to isolate phloem cells from rice leaves [49]. Subsequently, epidermal and vascular tissues were isolated from maize coleoptiles [50] and then various cell or tissue types, including bundle sheath, shoot tip protoderm, shoot apical meristem, leaf primordium, seedling procambium, hypocotyls and root meristems have been harvested from maize, rice, radish and tomato [51, 52]. So far, most studies have focused on shoot tissues. In a few cases, LM has been used to isolate plant cells during interactions with other organisms including cyst or root-knot nematode feeding sites in soybean roots [53–57]. In addition, fungus-plant interactions were assessed in maize by dissecting parenchyma cells infected with Colletotrichum graminicola and the RNA samples were used for analysis of C. graminicola gene expression [58]. Recently, this technology was used to isolate cells from arbuscular mycorrhizal or non-mycorrhizal tomato roots [59] where it enabled the analysis of phosphate transporter gene expression in the plant and the fungus.

Toward the ultimate goal of understanding how arbuscules develop, we initiated experiments to identify genes expressed in cortical cells containing arbuscules. We first undertook a profiling experiment with a new M. truncatula GeneChip^®, and then developed LM of M. truncatula root cortical cells to enable cell-type specific analyses of gene expression by RT-PCR. This approach led to the identification of novel plant and fungal genes expressed in colonized cortical cells and in arbuscules, including members of the LysM-receptor like kinase family, GRAS transcription factor family, a novel ammonium transporter that is a strong candidate for mediating symbiotic ammonium transport, and AM fungal genes of the urea cycle.

Here, we used the Medicago GeneChip^® to extend the transcript dataset for mycorrhizal roots and coupled this with LM and RT-PCR to assess gene expression specifically at the cortical cell-arbuscule interface. At the time that this work was initiated, there were only a few reports of LM of root cells or roots associated with other organisms, consequently, it was necessary to test and modify existing protocols to develop a protocol suitable for M. truncatula mycorrhizal roots. The protocol developed was based largely on Kerk et al. [51] and the main modification was the nature of the embedding material. Following initial difficulties with the quality of the RNA extracted from paraffin-embedded tissues, we tested Steedman's wax with the idea that RNA degradation would be minimized by exposing roots to lower temperatures during embedding, and the use of ethanol for de-waxing would be less detrimental than xylene. This proved to be the case and this low-melting-temperature, ethanol-soluble wax enabled us to extract RNA from M. truncatula roots of superior quality. LM of approximately 400 M. truncatula root cortical cells, followed by two rounds of linear RNA amplification provided microgram quantities of RNA for RT-PCR analyses. These cell-type specific RNA samples were validated by monitoring the expression of three genes, MtPT4, MtSCP1 and MtBCP1, whose expression patterns have been described previously [28, 36, 38]. Recently, Balestrini et al. [59] reported the use of methacarn fixation and paraffin embedding for LM of cells from tomato roots colonized by the AM fungus G. mosseae. However, they did not couple this with linear RNA amplification and consequently, it was necessary to harvest a large quantity of cells, between 3,000 and 11,000 cells to obtain nanogram quantities of RNA. Although the linear RNA amplification used in our approach may lead to a 3' end bias in the transcript molecules, comparative analyses are possible with careful primer design.

The GeneChip analysis provided a greatly expanded view of genes induced in the AM symbiosis, including several transcription factors and receptors. As reported previously, development of the AM symbiosis is accompanied by significant alterations in gene expression including the expression of novel 'mycorrhiza-specific' genes. Surprisingly, transcription factors responsible for mycorrhiza-regulated gene expression are entirely unknown. Here we identified 5 members of the GRAS transcription factor family induced in mycorrhizal roots (AL386879, AL386880, TC105118, IMGAG|1755.m00026, and AJ499899). The GRAS family of proteins is unique to plants and its name derives from three members that were initially identified GAI, RGA, and SCR [74–76]. The Arabidopsis GRAS family contains at least 33 members including Scarecrow and Short Root, two proteins that play a central role in root development (reviewed by [77]. Two M. truncatula GRAS proteins, MtNSP1 and MtNSP2, are essential for the activation of Nod-factor-induced genes in the rhizobial symbiosis [62, 63]. Given their roles in root development and symbiosis, members of this family are likely candidates for regulating gene expression in mycorrhizal roots. Here, we confirmed by RT-PCR that two putative GRAS genes; AJ499899 and AL386880, were expressed in G. versiforme- or G. intraradices-colonized roots, with non-detectable or low levels of transcripts in mock-inoculated roots. Furthermore, AJ499899 and AL386880 transcripts were not detected in cortical cells from mock-inoculated roots but were detected in arbuscule-containing cortical cells. Consequently, it is possible that these GRAS factors regulate gene expression in colonized cortical cells. The current data do not preclude a role in other cell types and one GRAS factor, represented by EST AL386880 is also highly expressed in root nodules. It would be interesting to know whether expression is associated with cells containing symbiosomes [60].

In addition to transcription factors, receptor proteins operating in signaling pathways, involved in arbuscule development are also entirely unknown. In contrast, receptors involved in signaling during nodule development have been described. Genes predicted to encode LysM-RLK including NFR1 and NFR5 in Lotus japonicus [78, 79] and orthologs LYK3 and NFP in M. truncatula [80], and SYM10 in pea [78] are involved in Nod factor signaling and are likely to be Nod factor receptors. The M. truncatula LysM-RLK family contains at least 17 genes, which, based on their kinase domain phylogeny, are divided into three subfamilies. Subfamily II contains NFP and four other members of the family of which MtLYR1 is most similar in sequence to NFP. There are no ESTs for MtLYR1 and in previous studies, transcripts were detected in nodulated roots but not in nodules [65]. Here, we found that MtLYR1 (IMGAG|1197.m00015) transcripts increased in G. versiforme or G. intraradices-colonized roots. Moreover, the MtLYR1 gene was detected in cortical cells with arbuscules. Like NFP, MtLYR1 has three LysM domains and a Ser-Thr kinase domain that lacks the activation and P-loops and consequently, it probably lacks kinase activity [65]. It is possible that these 'inactive' kinases dimerize with an active receptor kinase to form a receptor with kinase activity and consistent with this, heterodimerization of NFR1 and NFR5 has been demonstrated. LysM domains recognize N-acetyl glucosamine or structurally similar molecules such as Nod factors [81–83]. It is possible that the LysM domain of MtLYR1 recognizes the N-acetyl glucosamine residues in chitin in the AM fungal cell walls and given its expression pattern, it could play a role in signaling during arbuscule development.

Cell biological analyses of the arbuscule-cortical cell interface suggest that the peri-arbuscular membrane increases the membrane area within the cortical cell at least 10-fold. Consistent with this, metabolite profiles of mycorrhizal roots indicate a significant increase in fatty acids [84]. The GeneChip analyses identified two new mycorrhiza-induced genes predicted to be involved in fatty acid metabolism; a putative palmitoyl-ACP thioesterase gene (TC99271) and a triacylglycerol/steryl ester lipase-like protein gene (TC110731) that show a 49- and 102-fold increase in transcript levels respectively. RT-PCR analysis indicated that TC99271 and TC110731 transcript levels increase in cortical cells with arbuscules. Fatty acid synthesis in plants is a cyclic process and the extension cycles involves the binding of acyl chains to a soluble acyl carrier protein (ACP), and specific thioesterases hydrolyze the newly formed acyl-ACP into fatty acids and ACP [85]. An increase in palmitoyl-ACP thioesterase gene expression suggests an increase in the biosynthesis of palmitic acid. Consistent with this, a recent metabolite profile of G. intraradices/M. truncatula roots showed that mycorrhizal roots contained high levels of palmitic and oleic acids during active phases of AM symbiosis [84]. Oleic acid (18:1Δ⁹) is created by a desaturation reaction catalyzed by a delta-9 desaturase. Surprisingly, a G. intraradices delta-9 desaturase gene was expressed highly in mycorrhizal roots and specifically in cells with arbuscules. Previous studies had speculated that the fungus might obtain palmitic acid from the plant, and even suggested that AM fungi might lack fatty acid synthase activity and thus be dependent on host lipid [86]. In the absence of a full genome sequence, it is not possible to determine whether this is correct, however, our findings of complementary gene expression patterns in the two symbionts is intriguing.

Other metabolic changes documented in mycorrhizal roots include an increase in amino acids [84]. In plants, there are at least three protein families involved in transport of small peptides, the ABC transporters, the di/tripeptide transporters (PTR family) and the oligopeptide (OPT) family. Two genes (AL387494 and TC97569) of the PTR family were expressed in mycorrhizal roots and TC97569 was detected in cells containing arbuscules. The GeneChip experiment also led to the identification of other new genes predicted to encode transport proteins including a copper transport protein (TC97522) and an ammonium transporter (IMGAG|1723.m00046) and the latter was expressed in cells with arbuscules. The identification of a mycorrhiza-induced ammonium transporter is particularly interesting. It is clear from many studies that in addition to Pi, N is also transferred from the fungus to the plant and recent analyses indicate that N is likely transferred to the plant as ammonium. This ammonium transporter is a strong candidate for a role in the acquisition of ammonium delivered via the fungus. Metabolic labeling studies and gene expression analyses led to a model in which inorganic N taken up by the fungal extra-radical mycelium is converted into arginine which is transported into the intra-radical mycelium, where it is broken down into amino acids. Ammonium, resulting from arginine breakdown is proposed to be exported to the plant [29]. In further support of the labeling studies, Govindarajulu et al. [29] showed elevated transcripts of G. intraradices ornithine aminotransferase, a urease accessory protein and an ammonium transporter in the intra-radical mycelium relative to the extra-radical mycelium.

Interestingly, the G. intraradices genes present on the M. truncatula array include genes predicted to encode enzymes of the urea cycle as well as amino acid metabolism. RT-PCR analysis showed that these genes were expressed in G. intraradices-mycorrhizal roots, and analysis of the cell-type specific samples, indicates that they are expressed in the arbuscules. These data extend the previous findings, showing that the genes of urea and amino acid metabolism are expressed in the arbuscules. In addition, our data set includes a gene predicted to encode arginase that breaks down arginine to release ornithine and urea. This is a key step for the release of ammonia and further supports the model proposed by Govindarajulu et al. [29]. In addition to the urea cycle, we detected G. intraradices asparagine synthetase and glutamine synthetase gene expression in arbuscules. This suggests that ornithine, released via arginase activity, may be cycled into asparagine and glutamine through the action of asparagine synthetase and glutamine synthetase. Interestingly, the metabolite profiling studies found higher levels of specific amino acids, including Glu, Asp, and Asn in mycorrhizal roots [84]. From these studies, it is not possible to determine whether the amino acids are in the plant or AM fungus, however, coupled with the gene expression studies, there is evidence to suggest that at least some of this amino acid accumulation may result from activities in the fungus.

Using a combination of arrays, LM and RT-PCR, we have extended the M. truncatula/G. intraradices mycorrhizal root transcriptome and identified novel M. truncatula and G. intraradices genes that are expressed in cortical cells harboring arbuscules. By virtue of their expression patterns, the genes expressed in this cell type are candidates for mediating aspects of development or function of the symbiotic arbuscule-cortical cell interface. Novel genes potentially involved in signaling and in N transport and metabolism are of particular interest. This dataset will be valuable for future studies aimed at determining the molecular events that control arbuscule development, and N transport and metabolism in the AM symbiosis.