Huanglongbing (HLB) is considered a devastating citrus disease and affects most of the main citrus growing areas worldwide. It has become a serious threat to Brazil and USA, the largest producers of sweet orange and frozen concentrated orange juice, since the first reports of the disease in 2004 [1] and 2005 [2], respectively. In Brazil, the causal agents of HLB are the Gram-negative and phloem-restricted bacteria Candidatus Liberibacter asiaticus (CaLas) and Ca. L. americanus (CaLam). Another species, Ca. L. africanus (CaLaf), occurs in Africa and is not found in South, Central or North America. Both CaLas and CaLam are transmitted by the Asian citrus psyllid (Diaphorina citri); however, they can also be transmitted with different levels of efficiency by grafting of infected tissues [3, 4].

Once a citrus tree is infected, the time taken for HLB symptoms to appear varies according to the genotype; the physiological stage and age of the plant; the tissue used as the source of bacteria or the intensity of inoculation by the psyllid; and the season of the year. In leaves, typical symptoms include yellowing, asymmetric blotchy mottling and nutrient deficiency. Small, poor quality and lopsided fruits are frequently observed [5]. There is a clear disturbance of the transport system between source (leaves) and sinks (meristems, fruits, roots), with starch accumulation in plastids, chloroplast disruption, necrotic phloem (probably associated with plugging of sieve cells via callose deposition) and phloem-proteins accumulation [6–8]. Ultrastructural changes in the phloem and adjacent tissues can be observed in plants infected with CaLas, from the early (asymptomatic) stages of infection (three months), progressing to phloem degeneration nine months later [8]. However, before the complete degeneration of the phloem vessels, the bacteria move to newly growing tissues to restart the infection process [8]. Compared with CaLas, CaLam is less effectively transmitted both by vector and by grafting, is more sensitive to higher temperatures and reaches lower titer levels under the same experimental conditions. Both bacteria induce indistinguishable HLB symptoms in susceptible genotypes [9].

Using two different metagenomic approaches, two research groups reported the full genome of CaLas [10, 11]. The authors showed that the bacterium has a small (1.23 Mb) genome, limited capacity for aerobic respiration and depends on several primary metabolites. It lacks type III and type IV secretion systems, avirulence factors, and degradative enzymes, which probably avoid the elicitation of defense response based on the products of plant cell wall degradation. However, the authors did not discount the possibility that CaLas may elicit a plant response mediated by a pathogen-associated molecular patterns (PAMP) and/or secrete an unknown effector that suppresses the host defense [11]. Thus, together with the absence of important housekeeping genes associated with the difficulty to culture the bacterium in artificial medium [12, 13], its capacity to multiply both in plants and insects, and its intracellular habitat in the phloem sieve tubes suggest a parasitic, rather than a pathogenic, lifestyle. Additionally, two highly related, circular prophages have been found integrated into the CaLas genome [14]. One of them, SC1, is apparently fully functional, with a lytic cycle that appears to be activated in plants, but not in psyllids. The other prophage, SC2, is apparently in the lysogenic phase both in plants and psyllid infected by CaLas. Whether or not the CaLas prophages play a direct role in its competitiveness with CaLam remains to be determined; however, Zhang et al. suggested that this possibility could not be ruled out [14]. Taken together, these features clearly suggest an improved fitness of CaLas when compared with CaLam, and may explain why, in recent years, this species has become prevalent in Brazil. Nevertheless, CaLam is still found in the field and there is at least one report of this species in China [15].

Several studies on the transcriptional response of citrus plants showing HLB symptoms have been reported; however, all of them focused on CaLas infection. Overall, they show complex gene expression modulation in response to CaLas, but no specific mechanism associated with the infection has been identified yet [6, 16–20]. It should be noted that the studies have focused on CaLas because of its worldwide importance; however, finding common features - and differences - between the two species, and differences in the response of citrus to them, may lead to a better understanding of the pathogenesis of HLB and how citrus plants respond to the disease. The present study evaluated the transcriptional reprogramming of leaf tissue of sweet orange during the symptomatic stage of infection with CaLam, using a robust, customized, 385K-microarray chip containing approximately 32,000 unigenes from Citrus sinensis L. Osb. A set of candidate genes from this and from previous studies was used to validate the expression profile in symptomatic and asymptomatic PCR-positive leaf tissue of sweet orange infected with CaLas or CaLam. Based on these results, we described the main alterations in gene expression during the late stages of infection of CaLam compared with CaLas and discussed how this information could be useful for genetic engineering to control HLB.

HLB is considered the most destructive citrus disease worldwide, and in Brazil it is caused by CaLam and CaLas. CaLas and CaLam are transmitted by the Asian citrus psyllid and are restricted to the phloem of infected citrus, where they can multiply and spread, causing a severe imbalance in the translocation of nutrients and other important metabolites. Phloem is the main trafficking pathway of nutrients, defensive compounds and signaling molecules throughout the plant; thus, several relevant biological processes are affected in citrus infected with Liberibacters.

Our study aimed to investigate the transcriptome reprogramming of citrus upon infection with CaLam. Microarray analysis identified 514 DEGs, which were grouped into gene ontology (GO) categories and ranked according to the most representative GO terms, as calculated by the GSEA approach. Among the molecular functions overrepresented, we found that gene transcripts encoding three zinc transporters (ZIP1, ZIP4 and ZIP5) were overexpressed. High levels of ZIP1 transcripts (FC = 30.48) were also observed in the transcriptome of ‘Navel’ and ‘Madam Vinous’ sweet oranges infected with CaLas, although ZIP1 was not found among the proteins evaluated in the citrus proteome [19]. Similarly, transcripts for a ZIP5 transporter were highly induced in the infected susceptible genotype (Cleopatra mandarin) but not in the tolerant genotype (US-897) [17]. The overexpression of transcripts for Zn transporters in citrus in the late stages of infection with CaLam or CaLas are expected, because the symptoms in the leaves of susceptible citrus plants often resemble those of zinc deficiency. In fact, in some citrus genotypes, the concentrations of Zn and Fe in infected plants was found to be approximately half those in healthy plants, which indicates that zinc homoeostasis is significantly affected during infection [24]. As an important micronutrient in plants, zinc has structural (e.g., carbonic anhydrase) and catalytic functions (e.g., superoxide dismutase CuZn-SOD) in many of enzymes. In addition, zinc is involved in the maintenance of membrane integrity and protection of cell structural components against the oxidative damage caused by ROS. As a result, several studies demonstrated that an imbalance in the intracellular zinc concentration affects not only optimal plant growth, but also, in some cases, the susceptibility/tolerance of plants to certain pathogens [25, 26]. In susceptible citrus infected with Liberibacters, zinc deficiency has, to some extent, been associated with inefficient translocation of this mineral by the roots. An important issue to be addressed is whether Liberibacters could directly or indirectly reduce the availability of this metal as a strategy to favor the infection process [24], as observed for Xanthomonas oryzae in rice [27].

In addition to nutritional deficiencies observed during symptom progression of HLB, several studies reported that CaLas infection dramatically affects carbohydrate metabolism [8, 17]. The imbalance of carbohydrate partitioning causes an accumulation of starch in infected leaves during the progression of HLB symptoms. Evidence for this observation include increases in the starch content in symptomatic leaves, microscopic observations of starch accumulation in phloem parenchyma cells of infected leaves, induction of transcripts encoding enzymes related to starch biosynthesis (such as ADP-glucose pyrophosphorylase (AGPase)) and repression of transcripts related to starch breakdown [6, 16, 17, 19]. Similarly to CaLas infected leaves, we found an induction of transcripts encoding key enzymes involved in starch biosynthesis and repression of those related to starch breakdown in CaLam infected leaves. Real time PCR of the starch synthase gene confirmed its upregulation in symptomatic leaves infected with CaLam and showed increased expression from asymptomatic to symptomatic leaves (Figure 3). High levels of sucrose and glucose in symptomatic leaves infected with CaLas have also been documented [16, 19]. The increase in glucose levels explains the significant induction of transcripts for glucose-6-phosphate/phosphate transporter 2 (GPT2) in the transcriptome of citrus leaves infected with CaLas [17]. In our microarray analysis, GPT2 was also significantly induced in symptomatic leaves infected with CaLam compared with the control. RT-qPCR analysis showed a significantly increased expression of transcripts for GPT2 in symptomatic and asymptomatic leaves infected with both CaLas and CaLam (Figure 3).

Transcripts encoding enzymes related to raffinose metabolism, another class of sugars found in phloem sap, were also modulated during CaLam infection. Among them, a galactinol synthase (GLS8) increased by seven-fold in infected plants. Galactinol synthase is the first enzyme in the synthesis of RFOs and regulates the partitioning between sucrose and RFOs, whereas raffinose synthases catalyze the synthesis of RFOs from sucrose and galactinol. High intracellular levels of RFOs have been correlated with osmoprotection in plant cells; however, recent studies reported RFOs as potential scavengers of ROS, suggesting a novel role for RFOs in protection against oxidative stress [28]. We hypothesized that the induction of enzymes involved in the biosynthesis of RFOs may be an attempt to reduce the level of sucrose in the phloem of infected leaves, although this strategy does not appear to be effective in preventing the accumulation of starch in leaves and the subsequent consequences on the translocation of nutrients from source to sink organs. Furthermore, RFOs could also function as potent antioxidants to minimize the oxidative stress that occurs near to the necrotic sieve elements formed during CaLam infection.

Transcripts encoding three UDP-glucosyltransferases (UGT84A1; UGT76B1 and a putative UGT) were overexpressed in symptomatic leaves upon infection with CaLam. UGTs catalyze the transfer of glucosyl residues from UDP-glucose to a wide range of secondary metabolites and hormones, such as salicylic acid (SA). A transcript for UGT76B1, which was slightly induced in our microarray analysis, has been reported as a key player in the crosstalk between SA-jasmonic acid (JA) signaling. The knockout of UGT76B1 in Arabidopsis led to enhanced resistance to Pseudomonas syringae, but an increased susceptibility to Alternaria brassicicola[29]. In HLB-infected citrus trees, transcripts encoding different UGTs were modulated in leaves [17] and in fruit tissues infected with CaLas [18]. Among them, transcripts encoding a UGT73B3 were more abundant in infected leaves of the tolerant hybrid compared with Cleopatra, the susceptible citrus genotype. Unlike UGT76B1, knockout of UGT73B3 and its homolog UGT73B5 in Arabidopsis led to increased susceptibility to Pseudomonas syringae[30]. The modulation of several UGTs either in CaLam or CaLas infected plants reveals the potential application of UGTs as target genes for genetic engineering.

During the symptomatic stage of HLB, the deposition of callose and P-proteins has been observed in the sieve pores of the sieve elements of the leaf phloem [8, 16]. P-proteins are structural proteins involved in sealing off damaged sieve elements by plugging up the sieve plate pores. This dynamic and reversible mechanism is frequently accompanied by a long-term solution to sieve tube damage: callose deposition in the sieve pores. Several transcriptome studies of citrus infected with CaLas have reported a strong induction of transcripts encoding the phloem protein PP2-B15 in symptomatic leaves of sweet oranges compared with the control [6, 16, 17]. In our microarray analysis, we found transcripts encoding two different phloem proteins, PP2-B10, which was highly induced and PP2-B14, which was moderately induced. We tested the expression of PP2-B10 by RT-qPCR and confirmed an increase in the level of this transcript in asymptomatic and symptomatic CaLam infected leaves compared with control leaves (Figure 3). PP2-B15 was also assayed by RT-qPCR and showed a similar pattern of overexpression found to PP2-B10 in symptomatic and asymptomatic leaves infected with CaLam. In relation to CaLas, our RT-PCR assays confirmed an induction of PP2-B15 during infection, but PP2-B10 was not significantly modulated. With regard to callose deposition, we found that transcripts encoding a cytochrome P450 (CYP83B1), an enzyme involved in glucosinolate biosynthesis, were four-fold induced in the microarray analysis. Transcripts encoding a CYP83B1 were upregulated in symptomatic flavedo of citrus fruits upon CaLas infection [18]. Indole glucosinolates (IGs) are secondary metabolites derived from tryptophan, which have a well-characterized role in insect resistance [31]. However, Clay and co-workers reported that IGs and their breakdown products were required for callose deposition, which was effective in restricting Pseudomonas syringae growth in Arabidopsis[22]. According to the microarray analysis, transcripts encoding a homogentisate phytyltransferase (VTE2), the first enzyme of the tocopherol biosynthetic pathway, were also induced. Tocopherols are antioxidants that have roles in protecting chloroplast membranes and the photosynthetic apparatus from oxidative damage. However, studies have shown that tocopherol has an important role in regulating the phloem loading in low-temperature adaptation [32, 33]. Intriguingly, loss of VTE2 function-mutants, which were deficient in tocopherol, exhibited an inhibition of photoassimilate transport, followed by an increase of solute sugar and consequently starch, and callose deposition in phloem parenchyma transfer cell walls adjacent to the companion cell/sieve element complex [33], leading to a phenotype that resembles HLB in citrus. Impairment of phloem loading is a major consequence of Liberibacter infection; therefore, modulation of tocopherol biosynthesis in citrus by overexpression of VTE2 during early infection could be an interesting approach for increasing the phloem translocation of nutrients and for minimizing the symptoms.

Considering that CaLam infection affected different biological processes in citrus, it is not surprising that the expression of many TFs were differentially modulated. Microarray analysis identified transcripts for 38 TFs that were differentially expressed in symptomatic leaves infected with CaLam. The most highly induced TF was a myb-like gene, which regulates the expression of several genes in response to phosphate during sucrose starvation in Arabidopsis[34, 35]. In addition to having regulatory roles in the defense response upon infection with different pathogens [36–38], several MYB genes have been reported as key regulators of sugar-responsive genes, such as α-amylase during sugar starvation in rice [39]. Interestingly, the same myb-like gene was nearly 200-fold induced in symptomatic leaves of susceptible plants infected with CaLas, but not in the tolerant genotype, indicating that the upregulation of this gene could be associated with the susceptibility of citrus to Ca. Liberibacter spp. or, to some extent, to the manifestation of symptoms [17]. Whether this myb-like gene is involved in regulating the expression of defense response genes or sugar metabolism genes in response to CaLam and CaLas infection remains to be proven.

Among the differentially expressed defense-related gene transcripts in CaLam-infected citrus leaves were several for receptor-like proteins and a LysM receptor-like kinase (CERK1). Although the differential expression of transcripts encoding a CERK1 could not be confirmed by RT-qPCR in CaLam infected leaves, this gene was induced in asymptomatic leaves infected with CaLas. CERK1 is a receptor implicated in the perception of chitin, an essential component of the cell walls of all fungi, which acts as elicitor of the defense response in plants [40]. Despite the recognition of the fungal PAMP chitin by CERK1, a recent study showed that this receptor was able to recognize the bacterial type III effector protein, AvrPtoB [41]. Although bacteria do not contain chitin, other carbohydrates with similar structures to chitin, or even an unknown bacterial PAMP, could be potential ligands of the LysM domain of CERK1 [42, 43]. CaLas does not have the type III secretion system or the degradative enzymes of type II [11]. However, it has been suggested that lipopolysaccharides (LPS) and LPS fragments, or even an unknown non-type III effector, could be the potential PAMP of CaLas [11]. In addition, we found that transcripts encoding three PR proteins (PR3, PR4 and PR6) were highly induced in our microarray data. PR3 and PR4 belong to the chitinase class, whereas PR6 proteins are defined as a subclass of serine proteinase inhibitors. Plant proteinase inhibitors play distinct physiological roles, including seed dormancy and protection against proteolytic enzymes of herbivores and pathogens. One of the first studies of the transcriptome of citrus leaves infected with CaLas reported an induction of PR6 only in asymptomatic leaves collected 5–9 weeks after inoculation [6]. In that paper, PR6 was annotated as a putative protease inhibitor (At2g38870). Our RT-qPCR data showed that both CaLas and CaLam induced PR6 in asymptomatic and symptomatic stages during infection. PR6 of Arabidopsis was induced in leaves upon infection with Botrytis cinerea and the resistance against this fungus was enhanced in transgenic lines overexpressing this gene [43]. To the best of our knowledge, few reports have demonstrated the induction of PR6 or other serine-proteinase inhibitors in plants upon infection with bacterial pathogens [44]. Further studies are needed to determine the role of PR6 in the Liberibacters vs. Citrus spp. interaction.

Among the gene transcripts moderately induced in the microarray analysis, we found an NDR1/HIN1-like3 (NHL3) gene. RT-qPCR showed a slight induction in the expression of this gene only in symptomatic leaves infected with CaLam. The expression of NHL3 was suppressed in Arabidopsis upon inoculation with a virulent Pseudomonas syringae (Pst DC3000); however, inoculation with a mutant Pst DC3000, which is deficient in the delivery of effector proteins of the Type III-secretion system, induced the expression of NHL3 [45].

Among the repressed gene transcripts related to defense response, we identified transcripts for two carbonic anhydrases: salicylic acid-binding protein 3 (SABP3) and beta carbonic anhydrase 2 (CA2). SABP3 has a high affinity for SA and its downregulation has been detected in CaLas infected leaves of citrus [6]. In addition, Slaymaker (2002) showed that in tobacco, SABP3 has antioxidant properties and has a role in the hypersensitive response [46]. Similarly, transcripts for CA2 were also reported to decrease significantly in leaves of Arabidopsis 12 hours post-inoculation with Pseudomonas syringae[47], which leads us to speculate that the downregulation of these genes could be associated with susceptibility of citrus to Liberibacters.

In addition to defense response genes, several transcripts encoding enzymes directly involved in oxidative stress were differentially modulated. Among them were transcripts for CSD1 and CSD2, which were induced and repressed, respectively, in symptomatic leaves infected by CaLam. This was also observed in transcriptome of citrus leaves upon infection with CaLas [6, 17, 19]. RT-qPCR analysis showed that transcripts encoding an NADPH oxidase (RbohD), which were slightly induced in the microarray analysis, were significantly repressed during both CaLam and CaLas infection (Figure 3). Increased expression of RbohD was responsible for accumulation of ROS during the defense response of Arabidopsis against Pseudomonas syringae and Peronospora parasitica[23]. In plants, ROS are continuously produced in mitochondria, chloroplasts, and peroxisomes as by-products of aerobic metabolic processes, such as respiration and photosynthesis. However, under stress conditions, the balance between production and scavenging of ROS is disturbed. During biotic stress, the production of ROS is termed the oxidative burst and is one of the first steps in the response of plants to pathogen attack. High levels of ROS can also be produced as a result of uncoupling, or inhibition, of the photosystem machinery in the chloroplast and during photorespiration [48]. It is not clear whether oxidative stress in late stages of HLB is part of a defense mechanism of citrus or a secondary effect caused by depletion of photosynthetic apparatus and necrosis of sieve elements. In any case, the role of the oxidative stress during CaLam or CaLam infection needs to be further investigated.

This study aimed to unravel the changes in the transcriptome of symptomatic leaves of sweet orange upon infection with CaLam, using a customized 385K microarray chip containing about 32,000 Citrus sinensis cv Pera unigene transcripts. A large number of citrus transcripts and biological processes were significantly altered upon CaLam infection. Among the changes we highlight induction of transcripts for zinc transporters, differential modulation of transcripts encoding enzymes related to sugar metabolism, depletion of photosynthesis, induction of several defense-related genes, and modulation of transcripts encoding enzymes regulating ROS production. We found several biological processes differentially modulated in leaves during the symptomatic phase of CaLam infection, which are similarly affected by CaLas infection. Most of these biological processes probably reflect a secondary, rather than a primary, effect of the infection process. For example, the depletion of photosynthesis probably results from a mechanism of feedback inhibition caused by the accumulation of sucrose and glucose in leaves during infection [17]. Moreover, the impairment of the phloem vessels caused by deposition of P-proteins and callose appears to be a late and unsuccessful strategy of defense, because it does not prevent the bacteria from spreading through the plant. The blockage of the phloem vessels also affects the translocation of important nutrients through the plant. In this sense, PP2 gene silencing or silencing of genes related to the callose deposition would be a promising strategy to reduce the severity of symptoms of HLB, allowing the transport of nutrients through the phloem. However, the silencing of callose genes has been shown to favor the spread of Xanthomonas citri subsp. citri, (syn. Xanthomonas axonopodis pv. citri) leading to the development of the canker disease in citrus [49].

Microarray analysis showed several citrus transcripts that were differentially expressed in symptomatic compared to control plants are annotated as genes responsive to infection by bacterial pathogens, according to sequence homology to Arabidopsis genes. The identification of large number of transcripts coding for PR proteins, receptor-like proteins, NBS-LRR and transcription factors (such as WRKY and MYB) shows that even a susceptible citrus genotype is able to actively respond to infection by CaLam, as reported for CaLas. The development of HLB disease symptoms leads us to believe that the perception of the pathogen by the host and the subsequent activation or repression of genes involved in resistance is delayed or is insufficient to protect the plant from the pathogen. For this reason, certain defense-related genes that are able to increase the perception of the pathogen by the host and/or trigger a systemic defense response to CaLam and CaLas infection have been selected as candidates for citrus genetic engineering in our laboratory.