PHOSPHATE RESPONSE 1 family members act distinctly to regulate transcriptional responses to phosphate starvation

Abstract

To sustain growth when facing phosphate (Pi) starvation, plants trigger an array of adaptive responses that are largely controlled at transcriptional levels. In Arabidopsis (Arabidopsis thaliana), the four transcription factors of the PHOSPHATE RESPONSE 1 (PHR1) family, PHR1 and its homologs PHR1-like 1 (PHL1), PHL2, and PHL3 form the central regulatory system that controls the expression of Pi starvation-responsive (PSR) genes. However, how each of these four proteins function in regulating the transcription of PSR genes remains largely unknown. In this work, we performed comparative phenotypic and transcriptomic analyses using Arabidopsis mutants with various combinations of mutations in these four genes. The results showed that PHR1/PHL1 and PHL2/PHL3 do not physically interact with each other and function as two distinct modules in regulating plant development and transcriptional responses to Pi starvation. In the PHR1/PHL1 module, PHR1 plays a dominant role, whereas, in the PHL2/PHL3 module, PHL2 and PHL3 contribute similarly to the regulation of PSR gene transcription. By analyzing their common and specific targets, we showed that these PHR proteins could function as both positive and negative regulators of PSR gene expression depending on their targets. Some interactions between PHR1 and PHL2/PHL3 in regulating PSR gene expression were also observed. In addition, we identified a large set of defense-related genes whose expression is not affected in wild-type plants but is altered in the mutant plants under Pi starvation. These results increase our understanding of the molecular mechanism underlying plant transcriptional responses to Pi starvation.

Introduction

Phosphorus (P) is an essential macronutrient for all organisms. Although P is abundant in the environment, the amount of its major form for plant uptake, phosphate (Pi), is quite limited in both natural ecosystems and farmlands because of its low diffusion rate, absorption by soil, and conversion to organophosphates by microorganisms. Pi deficiency in soils substantially reduces plant growth and has become a major constraint for agricultural production (Nussaume et al., 2011; Cong et al., 2020).

To cope with the limited availability of Pi, plants have evolved sophisticated responses, including the alteration of root system architecture; the increase of high-affinity Pi transporter activities on the root surface; the induction and secretion of acid phosphatases (APases), ribonucleases, and organic acids; the remodeling of membrane lipids; and the accumulation of anthocyanins and starches in leaves (López-Arredondo et al., 2014). These adaptive responses are accompanied by a dramatic genome-wide reprograming of gene transcription (Wu et al., 2003; Misson et al., 2005; Morcuende et al., 2007; Bustos et al., 2010; Osorio et al., 2019; Hani et al., 2021). The genes encoding Pi transporters, APases, ribonucleases, and genes involved in organic acid export and anthocyanin biosynthesis are highly induced under Pi starvation. Pi starvation also induces the expression of two microRNAs, miRNA399 and miRNA827, which cleave the transcripts of PHOSPHATE2 (PHO2) and NITROGENE LIMITATION ADAPTATION (NLA). PHO2 and NLA are involved in degrading PHOSPHATE TRANSPORTER1 (PHT1) proteins (a family of transporters for Pi uptake in roots) and PHOSPHATE1 (PHO1, a Pi exporter responsible for translocating Pi from roots to shoots) via the 26S proteasome pathway (Paz-Ares et al., 2022). With the cleavage of PHO2 and NLA mRNAs, PHT1 and PHO1 proteins are stabilized, which enhances Pi uptake in roots and Pi translocation from roots to shoots. The high induction of miRNA399 is followed by a rapid rise in the levels of two long non-coding RNAs (lncRNAs), INDUCED BY PHOSPHATE STARVATION 1 (IPS1) and At4 which antagonize the activity of miRNA399 through a mechanism called target mimicry (Franco-Zorrilla et al., 2007). Because the majority of these adaptive responses are transcriptionally regulated, it is important to understand how such reprograming of gene transcription is controlled at the genomic level.

PHOSPHATE RESPONSE 1 (PHR1) was identified in Arabidopsis (Arabidopsis thaliana) as a key regulator of transcriptional responses to Pi starvation (Rubio et al., 2001). It is a transcription factor with an MYB domain and a coiled-coil domain. PHR1 binds to a cis-element called P1BS (PHR1-binding sequence) with an imperfect palindromic sequence of 5′-GNATATNC-3′ (Rubio et al., 2001). This cis-element is prevalent in the promoters of many Pi starvation-induced (PSI) genes (Bustos et al., 2010). Knockout of PHR1 reduced the induction of many PSI genes, including those encoding the high-affinity Pi transporters PHOSPHATE TRANSPORTER1;1 (Pht1;1) and Pht1;4, an acid phosphatase ACID PHOSPHATASE 5 (ACP5), and an RNase RNASE 1 (RNS1), the miRNA399, and the lncRNAs IPS1 and At4. The transcriptional activity of PHR1 is regulated by certain SPX domain-containing proteins. Under Pi sufficiency, the concentration of cellular inositol pyrophosphate InsP₈ is high, which brings SPX1 and PHR1 together to form a complex that prevents PHR1 from binding to P1BS elements on the promoters of the PSI genes; under Pi deficiency, the level of InsP₈ is decreased, releasing PHR1 from the complex, enabling it to bind to the promoters of PSI genes and to thereby increase gene transcription (Wild et al., 2016; Dong et al., 2019; Ried et al., 2021). The transcription of the PHR1 gene and the accumulation of the PHR1 protein, however, is not responsive to the change of Pi availability in the environment (Rubio et al., 2001).

In Arabidopsis, PHR1 belongs to a 15-member protein family that can be divided into two clades (Rubio et al., 2001) (Figure 1A). Besides sharing a common MYB and coiled-coil domain, proteins in Clade A contain an extra N-terminal part, and proteins in Clade B contain an extra C-terminal part (Figure 1B). PHR1-LIKE1 (PHL1) is a close relative of PHR1 in Clade A. The mutation of PHL1 has little effect on the expression of PSI genes; in the phr1phl1 double mutant, however, the expression of some PSI genes is further reduced compared to that in the phr1 single mutant, indicating a functional redundancy between PHR1 and PHL1 (Bustos et al., 2010). The double mutation in PHR1 and PHL1 does not completely abolish the transcription of Pi starvation-responsive (PSR) genes (Bustos et al., 2010), suggesting that other members in this PHR1 family also function in regulating transcriptional responses to Pi starvation. Consistent with this inference, our recent work indicated that three other members of the PHR1 family—PHL2, PHL3, and PHL4 are also involved in regulating transcriptional responses of plants to Pi starvation (Sun et al., 2016; Wang et al., 2018). The functional orthologs of PHR1 have been found in other plant species, such as in rice (Oryza sativa) (Zhou et al., 2008), Brassica napus (Ren et al., 2012), common bean (Phaseolus vulgaris L.) (Valdés-López et al., 2008), wheat (Triticum aestivum) (Wang et al., 2013), soybean (Glycine max) (Xue et al., 2017), barley (Hordeum vulgare) (Sega et al., 2020), and tomato (Solanum lycopersicum) (Zhang et al., 2021), suggesting that PHR1 is an evolutionarily conserved and agriculturally important regulator of the responses of plants to Pi starvation.

Figure 1.

Phylogenetic tree of Arabidopsis MYB-CC family members and tissue-specific expression patterns of PHR1 and PHL1. A, The phylogenetic tree was generated by MEGA 7 software. Multiple alignment was conducted by an online service, Clustal Omega (https://www.ebi.ac.uk/Tools/msa/clustalo/). A maximum likelihood (ML) algorithm and full-length protein sequences were used to construct the tree. B, Diagram showing the amino acid sequences of MYB-CC family members. The sequences with high similarities are shown in red. One unit in the scale (between two long vertical lines) represents 20 amino acids. The diagram was generated using an online service, COBALT (https://www.ncbi.nlm.nih.gov/tools/cobalt/). Diamonds indicate the genes that were investigated in this study. C–N, The expression patterns of PHR1::GUS (C–H) and PHL1::GUS (I–N).

Although the functions of PHR1 family proteins in regulating transcriptional responses to Pi starvation have been well recognized, how different members of the PHR1 family execute such regulation remains largely unknown. To elucidate the functions of each member in the PHR1 family in regulating PSR gene expression, we generated a group of Arabidopsis mutants with various combinations of mutations in PHR1, PHL1, PHL2, and PHL3. We then performed comparative phenotypic and transcriptomic analyses of these mutants. Our results show that PHR1/PHL1 and PHL2/PHL3 act as two distinct modules in regulating plant development and transcriptional responses to Pi starvation.

Results

The expression patterns of PHR1 and PHL1

As the first step to understand the functions of these four members of the PHR1 family (hereafter, they are collectively referred to as PHR proteins or PHR genes), we investigated their tissue-specific expression patterns and responsiveness to Pi starvation. The 2-kb DNA fragments upstream of the start codon of PHR1 and PHL1 were fused to a GUS reporter gene. These two gene constructs were transformed into wild-type (WT) Arabidopsis plants. More than 20 independent transgenic lines were generated for each construct, and the GUS expression patterns of a representative line are shown in Figure 1. PHR1::GUS was ubiquitously expressed in all tissues examined, which included roots, leaves, stems, flowers, and siliques (Figure 1, C–H), whereas PHL1::GUS was expressed in a vascular tissue-preferred manner (Figure 1, I–N). Overall, the expression of PHL1::GUS was much lower than that of PHR1::GUS.

We then compared the GUS expression pattern of 8-day-old seedlings grown on Pi-sufficient (+Pi) and Pi-deficient (−Pi) media (Supplemental Figure 1). Consistent with previous studies (Rubio et al., 2001; Bustos et al., 2010), the expression of PHR1 and PHL1 was not responsive to Pi starvation.

For PHL2, we failed to amplify the 2-kb DNA sequence upstream of its transcription start site. For PHL3, the 2-kb sequence upstream of its transcription start site could not drive GUS gene expression in the transgenic plants, indicating that the regulatory elements that control PHL3 gene expression lie beyond its 2-kb upstream sequence.

Generation of multiple PHR gene mutants

To determine the functions of the four PHR transcription factors in plant development and responses to Pi deficiency, we generated double, triple, and quadruple mutants using the T-DNA insertion line of each gene through genetic crosses. The positions of the T-DNA in phr1 (SALK_067629), phl1 (SAIL_731_B09), phl2 (SALK_114420), and phl3 (SALK_010040C) are shown in Supplemental Figure 2A. RT-qPCR analyses using the primers after T-DNA insertion sites indicated that all of these lines were null alleles (Supplemental Figure 2B). Researchers previously reported that a Ds transposon insertion line of PHL3 was unable to be fertilized (Pagnussat et al., 2005). This developmental defect, however, was not observed in our T-DNA PHL3 line. Therefore, we speculated that the defect in the phl3 line with the Ds insertion was not caused by the functional disruption of PHL3 but was caused by other unknown mutations in the genome. To provide more supporting evidence for this inference, we used the CRISPR/Cas9 gene-editing technique to create two additional alleles of PHL3 in the WT background which were hypothetically null alleles because of the introduction of a premature stop codon in PHL3 (Supplemental Figure 3A). The PHL3 T-DNA line and the two CRISPR-generated phl3 alleles had similar growth characteristics (Supplemental Figure 3B). We, therefore, considered this T-DNA line to be functionally equivalent to the two CRISPR-generated phl3 lines.

PHR1/PHL1 and PHL2/PHL3 separately regulate plant development

We first grew the WT and all of the mutants in the soil to observe their growth phenotypes. None of the single mutants showed obvious defects in growth and development (Figure 2A). When phl1, but not phl2 or phl3, was combined with phr1, the double mutant displayed early senescence in rosette leaves (Figure 2B and Supplemental Figure 4, A and B). The triple mutant phr1phl1phl2 did not show further phenotypic changes compared with phr1phl1. However, the quadruple mutant phr1phl1phl2phl3 exhibited strong growth retardation (Figure 2C). To determine which combination of the mutations caused growth retardation, we compared the growth phenotype of the double mutants phr1phl3, phl1phl3, and phl2phl3. The results showed that phl2phl3 had a similar growth reduction as the quadruple mutant (Figure 2D). When phl2 was crossed to a CRISPR-generated phl3 alleles, the resultant double mutants showed the same growth retardation phenotype (Supplemental Figure 4C), confirming that the growth defects observed in phl2phl3 were due to the combined mutations of PHL2 and PHL3. Interestingly, the early senescence phenotype of the double and triple mutants was suppressed in the quadruple mutant (Figure 2C and Supplemental Figure 4B), suggesting an antagonistic interaction between PHR1/PHL1 and PHL2/PHL3.

Figure 2.

Growth phenotypes of phr mutants in soil. A, The WT and five single phr mutants. B, The WT, phr1, and three double mutants. C, The WT, phr1, phr1phl1, and the triple and quadruple mutants. D, The WT, phr1, and three double mutants. All plants were 35-day-old when photographed. r1, l1, l2, and l3 are short for phr1, phl1, phl2, and phl3 in double, triple, and quadruple mutants. Bars = 5 cm.

To determine the function of PHRs in root growth, we germinated the seeds of the WT and the mutants on 1/2 MS medium in Petri dishes. All of the mutants showed similar primary root (PR) length at 8 days after germination (DAG), except the mutants phl2phl3 and phr1phl1phl2phl3 whose PR lengths were about 50% of the WT (Figure 3A). Together, the above results suggested that PHR1 and PHL1 act as one functional module, and that PHL2 and PHL3 act as another functional module in regulating plant growth. This is consistent with their positions in the phylogenetic tree of the PHR1 family (Figure 1A).

Figure 3.

Primary root length and cellular Pi content of the WT and various mutants grown under Pi sufficiency (+Pi) and Pi deficiency (−Pi). A, Morphology of 8-day-old seedlings of various mutants grown under +Pi and −Pi conditions. Bar = 1 cm. The scale bar refers to all images in (A). B and C, The cellular Pi contents of shoots and roots of 10-day-old seedlings grown under +Pi (B) and −Pi (C). Agar was used as the gelling agent for measurement of cellular Pi content. The experiments were repeated three times, and the representative results are shown. The boxplots contain the first and third quartiles, split by the median; Tukey whiskers go to the highest or lowest point. One-way ANOVA and Tukey's multiple comparisons test were used for statistical analysis, P < 0.05.

PHR1/PHL1 and PHL2/PHL3 separately regulate plant responses to Pi starvation

Next, we analyzed the developmental and physiological responses of the phr mutants to Pi starvation. Seeds of the WT and the mutants were directly germinated on Pi-sufficient (+Pi) and Pi-deficient (−Pi) media. We compared the root growth responses to Pi starvation between the WT and various mutants at 8 DAG. On the −Pi medium, the PR length of WT seedlings was about 60–70% shorter on −Pi medium than that on +Pi medium (Figure 3A). PR growth of phr1, but not of phl1, phl2, or phl3, was further reduced compared with the WT on −Pi medium. This was consistent with that previously reported by Liu et al. (2017). The PR length of phr1phl1 was even shorter than that of phr1, suggesting that PHL1 functions redundantly with PHR1 in sustaining PR growth under Pi deficiency. The triple mutant phr1phl1phl2 and quadruple mutant phr1phl1phl2phl3 did not exhibit a further reduction of PR growth on −Pi medium. On −Pi medium, the PR length of phl2phl3 was similar to that of the WT. Given that on the +Pi medium, the PR length of phl2phl3 was already only half of that of the WT, the above results indicated that phl2phl3 was relatively insensitive to PSI inhibition of PR growth. The quantitative comparisons of PR lengths of all genotypes on −Pi and +Pi media are shown in Supplemental Figure 5.

The root hair length of WT seedlings was about three-fold greater on the −Pi medium than on the +Pi medium (Supplemental Figure 6). The mutation of PHR1 partially reduced the stimulation of root hair growth by Pi deficiency. The phr1-containing triple and quadruple mutants, but not the double mutants, showed a further reduction in root hair length. These results indicated that PHR1, PHL2, and PHL3 redundantly regulated PSI root hair growth.

We then analyzed the root-associated acid phosphatase (APase) activity using a histochemical staining method (see “Materials and Methods”). On the −Pi medium, phr1 had a weaker APase activity than the WT, whereas phl1, phl2, and phl3 single mutants had similar APase activity as the WT (Supplemental Figure 7). The double mutant phr1phl1 exhibited a lower APase activity compared with phr1. The triple mutant phr1phl1phl2 and the quadruple mutant phr1phl1phl2phl3, however, did not show a further reduction in APase activity. These results indicated that PHR1 plays a major role while PHL1 plays a minor and redundant role with PHR1 in regulating the root-associated APase activity induced by Pi deficiency. The contributions of PHL2 and PHL3 in regulating this Pi response were not substantial.

Finally, we investigated the roles of the four PHR proteins in regulating Pi homeostasis by analyzing the cellular Pi contents in all mutants at 8 DAG. Under Pi sufficiency, the Pi content in shoots was about 30% lower in phr1 than in the WT whereas the Pi contents in phl1, phl2, and phl3 shoots were similar to that in WT shoots (Figure 3B). The double mutants phr1phl1 showed a further decrease in shoot Pi content. The cellular Pi content in phr1phl1phl2 and phr1phl1phl2phl3 shoots was similar to that in phr1phl1 shoots. In roots, the Pi content of all mutants did not significantly differ from that of the WT. Under Pi deficiency, Pi contents in shoots were similar to that of the WT for all of the mutants; however, the Pi content in roots was about two- to three-fold higher for phr1phl1, phr1phl1phl2, and phr1phl1phl2phl3 than for the WT (Figure 3C). This might be due to the reduced translocation of Pi from roots to shoots. We also measured the total P contents in all mutants. The total P contents in all mutants were closely associated with their cellular Pi levels (Supplemental Figure 8). These results indicated that PHR1 plays a major role in regulating Pi homeostasis under both Pi sufficiency and deficiency; PHL1, in contrast, has a minor but redundant function with PHR1 in regulating root Pi homeostasis, and PHL2 and PHL3 alone do not seem to have a substantial contribution in maintaining Pi homeostasis.

Based on these results, we concluded that the PHR1/PHL1 module has a large effect on PSI inhibition of PR growth, on the activity of root-associated APases induced by Pi starvation, and on the maintenance of Pi homeostasis; in contrast, the effects of the PHL2/PHL3 module on these Pi responses were not substantial.

PHR1/PHL1 and PHL2/PHL3 do not physically interact with each other

Early research showed that PHR1 and PHL1 could form a dimer in vitro (Bustos et al., 2010). Using luciferase complementation imaging (LCI) assays in the leaves of Nicotiana tabacum leaves, we also found that PHL2 could directly interact with PHL3 but not with PHR1 (Sun et al., 2016). Here, LCI assays indicated that PHR1, PHL1, PHL2, and PHL3 all could interact with themselves (Supplemental Figure 9A). However, PHR1 and PHL1 could not interact with PHL2 or PHL3 (Supplemental Figure 9B). These four PHR transcription factors all shared above 80% of identity in both MYB and coiled-coil domains, but PHR1 and PHL1 contain an extra N-terminal segment, and PHL2 and PHL3 contain an extra C-terminal segment (Figure 1B). The above results suggest that although the coiled-coil domain is generally believed to be involved in protein–protein interactions, it seemed that the N-terminal parts of PHR1/PHL1 and that the C-terminal parts of PHL2/PHL3 determine the specificity of interactions. The molecular features of these four PHR proteins might provide a structural basis for the distinct functions of PHR1/PHL1 and PHL2/PHL3 modules in regulating plant development and physiological responses to Pi starvation.

Genome-wide identification of PSR genes

To understand the role of the four PHR proteins in regulating the transcription of PSR genes, we performed RNA-seq experiments using the WT and various mutants. We focused on roots because they act as the frontline when plants are facing Pi deficiency. Total RNAs were extracted from the roots of 8-day-old seedlings grown under Pi sufficiency and deficiency. All RNA samples were subjected to sequencing using next generation sequencing technology. Three biological replicates were used for all genotypes.

We first used Log₂FC ≥ 1 (or ≤−1) and FDR (false discovery rate) <0.05 as the cutoff to identify the PSI and PSS (phosphate starvation-suppressed) genes of the WT. With the above criteria, 1,595 PSI genes and 2,646 PSS genes were identified for further analyses (Supplemental Table 1A). To validate the results of the RNA-seq experiments, we performed RT-qPCR analysis of five PSI marker genes using the same batches of RNAs. The RT-qPCR results are shown in Supplemental Figure 10 along with those from the RNA-seq experiments. The data from the RNA-seq and RT-qPCR experiments were subjected to a Spearmen's correlation analysis. The correlation coefficients between the RNA-seq data and the RT-qPCR data for the five PSI genes were 0.90 (miRNA399), 0.85 (AtPAP10), 0.79 (Pht1;1), 0.77 (IPS1), and 0.70 (Pht1;4), respectively, indicating that the expression profiles of these five PSI marker genes from the RNA-seq experiment were largely consistent with those from the RT-qPCR assays.

We then performed GO (Gene Ontology) analyses for the PSI and PSS genes identified in the WT. The PSI genes were mainly enriched in the GO terms of response to Pi starvation, Pi transport, Pi homeostasis, galactolipid metabolism, and phospholipid metabolism. This was not surprising because these genes are known to be involved in plant responses to Pi starvation. The other PSI gene-enriched GO terms included those involved in the responses to other abiotic and biotic stresses, indicating that long-term exposure to Pi starvation triggered transcriptional responses to multiple stresses (Supplemental Figure 11A).

The PSS gene-enriched GO terms could be generally classified into four categories: (1) photosynthesis-related processes; (2) ribosome biogenesis; (3) responses to abiotic stress; and (4) hormone pathways (cytokinin and ABA). (Supplemental Figure 11B) Our laboratory previously reported that the suppression of photosynthetic gene expression was required for sustained root growth under Pi deficiency in illuminated Petri dishes (Kang et al., 2014). This is because the aberrant photosynthesis that occurs when roots are illuminated in Petri dishes could generate a large amount of reactive oxygen species, which would inhibit root growth. The advantage of reducing ribosome biogenesis might be the conservation of Pi, because the formation of ribosomes requires a large amount of Pi (Veneklaas et al., 2012).

Because most PSR gene-enriched GO terms identified in this work have been previously reported by other researchers (Misson et al., 2005; Bustos et al., 2010; Osorio et al., 2019), those terms are not presented in detail in this manuscript.

Distinct roles of the four PHR proteins in regulating PSR gene expression

We then compared the expression levels of the PSR genes between the WT and various phr mutants under Pi starvation. The expression level of each PSR gene of the Pi-deficient mutant was compared with that of the Pi-sufficient WT. The fold-changes obtained by comparing the Pi-deficient mutants with the Pi-sufficient WT for each PSR gene were presented in a heatmap (Figure 4A and Supplemental Table 2). The heatmap indicated that the expression of many PSR genes was substantially altered in phr1, phl2, and phl3, but the expression profile of phl1 was similar to that of the WT. The expression profiles of phl2 and phl3 were very similar, especially for PSS genes, but were clearly different from the expression profile of phr1. The changes in the expression of many PSR genes relative to the WT were greater in phr1phl1 than in phr1. A principal component analysis (PCA) was also performed using the RNA-seq data (Supplemental Figure 12). The results showed that under Pi deficiency, phl1 was tightly clustered with the WT, indicating that mutation in PHL1 has little effect on the expression of PSR genes. In contrast, phl2 and phl3 were closely associated with each other and were clearly distinguished from phr1 and phl1. Furthermore, phr1phl1phl2 and phr1phl1phl2phl3 were associated with phr1phl1, indicating that PHR1 and PHL1 together play a dominant role in regulating the expression of PSR genes.

Figure 4.

Summary of the differentially expressed genes in the WT and various phr mutants. A, Heatmap showing the expression levels of the PSR genes in the WT and various phr mutants. The value (Log₂FC) for each PSR gene in the heatmap was calculated by comparing the expression of Pi-deficient plants of each genotype to the expression of the Pi-sufficient WT. The diagram was generated using an online software, Morpheus (https://software.broadinstitute.org/morpheus/). B, Classifications of the mutation-affected PSR genes and the mutation-affected ancillary genes. C, Histograms showing the number of the six groups of genes in each mutant defined in (B). D, A table showing the number of the six groups of the genes in each mutant defined in (B). PSI, Pi starvation-induced; PSS, Pi starvation-suppressed.

When comparing the expression level of a given PSR gene between the WT and the phr mutants under Pi deficiency, we arbitrarily defined a two-fold difference with FDR < 0.05 as significant. With these two criteria, we found that the expression of 17.6% (281/1,595), 0.8% (13/1,595), 22.4% (358/1,595), and 19.1% (304/1,595) of the PSI genes and 11.9% (316/2,646), 0.8% (21/2,646), 48.4% (1280/2,646), and 42.2% (1,117/2,646) of the PSS genes was significantly altered relative to the WT in phr1, phl1, phl2, and phl3, respectively (Figure 4, C and D). Among these, the effect of the PHL1 mutation on the expression of PSR genes was almost negligible in the context of the entire genome. The number of the PSR genes whose expression was significantly affected in the multiple mutants increased progressively with the increase in the number of PHR mutations (Figure 4D).

In this report, all PSR genes whose expression levels were significantly altered in the mutants relative to the WT are referred to as mutation-affected PSR genes (Figure 4B). For these mutation-affected genes, we defined the PSI genes whose expression was significantly higher than that of the PSI genes of the WT as “PSI-up genes” and the PSI genes whose expression was significantly lower than the PSI genes of the WT as “PSI-down genes” (Figure 4B). Similarly, we defined the PSS genes whose expression was significantly higher than the PSS genes of the WT as “PSS-up genes” and the PSS genes whose expression was significantly lower than the PSS genes of the WT as “PSS-down genes”. These definitions indicate that the expression of PSI-down and PSS-up genes is positively regulated by PHR proteins, and that the expression of PSI-up and PSS-down genes is negatively regulated by PHR proteins. It follows that a given PHR protein could function as a positive or a negative regulator of PSR gene expression depending on which targets it acts on (Supplemental Table 1, B–J).

Next, we performed GO analysis of the mutation-affected PSR genes. Because the number of the PSR genes affected by PHL1 was too small for GO analysis (Supplemental Table 1C), we performed GO analyses of mutation-affected PSR genes only for phr1, phl2, and phl3. In phr1, the PSI-up genes were enriched in the GO in terms of response to the bacterium, toxin metabolic process, and response to fungus (Figure 5A and Supplemental Table 1B). In phl2 and phl3, the PSI-up genes shared several GO enrichments, including those in response to oxidative stress and the JA signaling pathway (Figures 5, C, and E). Because oxidative stress response and the JA pathway are both involved in defense responses, one shared function of PHR1, PHL2, and PHL3 might be to suppress the expression of defense-related genes under Pi starvation. The PSI-down genes in phr1 were mainly enriched in the GO terms of cellular response to Pi starvation, Pi transport, and galactolipid biosynthetic process (Figure 5B). The PSI-down genes in phl2 and phl3 were moderately enriched in the GO terms of cellular response to Pi starvation and gene silencing by miRNA (Figures 5, D, and F). These miRNAs, including miRNA169C, miRNA399B/D, miRNA827A, and miRNA156A/C (Supplemental Table 1, D, and E), have been demonstrated to be involved in Pi signaling (Bari et al., 2006; Hsieh et al., 2009; Kant et al., 2011). These results supported that, like PHR1, PHL2 and PHL3 also participate in regulating genes that control plant responses to Pi starvation.

Figure 5.

Lists of the GO terms of the mutation-affected PSR genes in the four phr single mutants. A, C, and E, the top GO terms of the phr1-, phl2-, and phl3-affected PSI-up genes, respectively. B, D, and F, the top GO terms of the phr1-, phl2-, and phl3-affected PSI-down genes, respectively. G–I, the top GO terms of the phr1-, phl2-, and phl3-affected PSS-up genes, respectively. The number of genes used for GO analyses in each group is indicated in parentheses. PSI, Pi starvation-induced; PSS, Pi starvation-suppressed.

Because the number of PSS-down genes in all single mutants was too low for GO analysis, GO analyses were only performed for the PSS-up genes. The PSS-up genes in phr1 were highly enriched in photosynthesis, light harvesting in photosystem I, protein-chromophore linkage, and photosynthetic electron transport chain (Figure 5G). A strong similarity of the GO enrichment of PSS-up genes was observed between phl2 and phl3, which were highly enriched in translation and ribosome biogenesis (Figure 5, H and I).

Another highly enriched GO term of PSS-up genes in phl2 was related to photosynthesis, which was also detected in phr1 but not in phl3. Taken together, these results indicate that PHR1 and PHL2 together inhibited the expression of many photosynthesis-related genes, whereas PHL2 and PHL3 together suppressed a large number of genes (near 200) involved in protein translation.

The above analyses of expression profiles of all of the phr mutants once again indicate that PHR1 and PHL1 function as one module and that PHL2 and PHL3 function as another module in regulating transcriptional responses to Pi starvation. Notably, a strong similarity of the enriched GO term of ribosome biogenesis was also observed between phl2 and phl3 in their downregulated genes under Pi sufficiency (Supplemental Figure 13). The latter suggested that the protein translation capacity in phl2 and phl3 was significantly reduced. This might explain the retarded growth of phl2phl3 under normal growth conditions.

PHR1 regulates the highly responsive PSR genes, while PHL2 and PHL3 regulate the moderately and weakly responsive PSR genes

Although our phenotypic analyses indicated that the effect of PHR1 on plant responses to Pi starvation was much stronger than that of PHL2 or PHL3 (Figures 2 and 3), the above transcriptomic analyses showed that the numbers of mutation-affected PSI genes among phr1, phl2, and phl3 were similar. To find the causes for this inconsistency, we analyzed the types of the PSR genes whose expression was affected by the mutation of each PHR gene. To perform such analyses, we divided all PSR genes into six groups according to the degree of their induction or suppression by Pi starvation. Groups 1 and 2 included genes with a high degree of induction (Log₂FC > 5) or suppression (Log₂FC < −5). Groups 3 and 4 included genes with a medium degree of induction (2 < Log₂FC ≤ 5) or suppression (−5 ≤ Log₂FC < −2). Groups 5 and 6 included genes with a low degree of induction (1 ≤ Log₂FC ≤ 2) or suppression (−2 ≤ Log₂FC ≤ −1). The GO annotation of these six groups of PSR genes is shown in Supplemental Figure 14. Group 1 mainly included genes involved in cellular responses to Pi starvation and gene silencing by miRNA. Group 2 included genes involved in photosynthesis and responding to light stimulus. Group 3 included genes responding to stresses, such as wounding, salt, oxidative stress, Pi starvation, as well as pathogenesis-related compounds, like jasmonic acid and chitin. Group 4 also included the genes responding to light stimulus and photosynthesis-related process. Group 5 included genes involved in glucosinolate biosynthesis and amino acid transport, and Group 6 included genes involved in ribosome biogenesis and also some genes responding to light intensity. We then determined the number of the mutation-affected PSR genes in these six groups in each phr mutant. As shown in Figure 6A, the mutation of PHR1 affected more Group 1 genes than the mutation of PHL2 or PHL3, whereas the mutation of PHL2 and PHL3 affected more genes (especially PSS genes) in Group 4–6 than the mutation of PHR1.

Figure 6.

Different types of PSR genes regulated by the four PHR proteins. A, A table indicating the number of PSR genes by type whose expression was affected in each phr mutant. B, Heatmap showing the relative expression levels of 135 highly responsive PSI genes in the WT and four phr mutants. The value (Log₂FC) for each PSR gene in the heatmap was calculated by comparing those of Pi-deficient plants of each genotype to those of Pi-sufficient WT. C, Heatmap showing the relative expression of 34 genes known to be or putatively to be involved in Pi homeostasis in the WT and various phr mutants under Pi deficiency. The TPM (transcripts per million) values used in the heatmaps have been standardized by min–max standardization. No., number of genes; PSI, Pi starvation-induced; PSS, Pi starvation-suppressed.

We then drew a heatmap using the expression levels of Group 1 genes in the WT and in four single phr mutants. The heatmap showed that PHR1 not only affected more PSI genes in Group 1 than PHL2 and PHL3, but also had stronger effects on the magnitude of induction of these genes than PHL2 and PHL3 (Figure 6B). We also used a heatmap to compare the effects of mutation of each PHR gene on the expression of some PSI genes known to be or putatively to be involved in Pi signaling, Pi transport, Pi remobilization, and phospholipid remodeling. Consistent with the previous results, this heatmap indicated that a single PHR1 mutation substantially impaired the induction of these genes (Bustos et al., 2010), while the mutation of the other three PHR genes had a much weaker effect (Figure 6C). This might explain why a single PHR1 mutation had much stronger effects than a single PHL2 or PHL3 mutation on plant responses to Pi starvation, although the number of the PSI genes regulated by these three PHR proteins was similar.

The common and specific targets of PHR proteins

We next identified the common and unique targets of these four PHR transcription factors. The targets identified using our experimental methods included genes that are both directly and indirectly regulated by the PHR proteins. Here, we arbitrarily defined the common targets as those genes whose expression was significantly affected by the mutation of at least two PHR genes. A Venn diagram was drawn using the mutation-affected PSI and PSS genes in four single phr mutants. There were five PSI genes whose expression was altered in all four phr single mutants (Figure 7A, area “a”). Except for miR399D, the expression of the other four genes was <4 (TPM value) even under Pi starvation (Supplemental Table 3A), suggesting that these four genes are not important for regulating Pi responses. We next focused on the 44 PSI genes commonly affected in phr1, phl2, and phl3 (Figure 7A, area “b”). K-means clustering had divided these genes into three groups with different expression patterns (Figure 7E). The 19 Group 1 genes were all downregulated in these three phr mutants relative to the WT; these genes included IPS1, miR827A, miR399B, and miR156A, which are well-known for their functions in Pi signaling. For 11 Group 2 genes, all of them were upregulated in phr1; however, seven genes were upregulated and four genes were downregulated in phl2 and phl3 relative to the WT. Four genes (PAD3, WRKY51, NAC042, TI1) in Group 2 were involved in the defense response. The 14 Group 3 genes were all upregulated in phl2 and phl3, and 10 of them were downregulated in phr1 relative to the WT. Some of them were involved in the regulation of redox homeostasis (Supplemental Table 3A). The expression patterns of the PSI genes in Groups 2 and 3 indicated an antagonistic interaction between PHR1 and PHL2 or between PHR1 and PHL3 in regulating these PSI genes. This added another layer of complexity to the regulation of the transcriptional response to Pi starvation.

Figure 7.

Common and unique PSR and ancillary genes regulated by four PHR proteins. A–D, Venn diagram showing the common and unique PSI (Pi starvation-induced) genes (A), PSS (Pi starvation-suppressed) genes (B), ancillary-up genes (C), and ancillary-down genes (D) whose expression was affected in four individual phr single mutants. The numbers of the genes affected in each mutant are indicated in the parentheses. The letters in the Venn diagrams represent a specific group of the genes. E, Heatmap showing the 44 PSI (in A, part “b”) genes whose expression was commonly affected in phr1, phl2, and phl3. F, Heatmap showing the 206 PSI (in A, part “c”) genes whose expression was commonly affected in phl2 and phl3. E and F, K-means clustering was used to divide these commonly affected genes into three (E) and two (F) groups. The TPM (transcripts per million) values used in the heatmaps have been standardized.

We also found a large number (206) of the PSI genes whose expression was commonly affected in phl2 and phl3 but not in phr1 (Figure 7A, area “c”). K-means clustering of these 206 genes resulted in two groups, one of which was significantly upregulated (63 genes of Group 1), while another was downregulated (143 genes of Group 2) compared with the WT (Figure 7F). Group 1 contained genes involved in response to salt stress and ABA- and JA-related functions. Group 2 contained several genes that encode miRNAs, such as miR822A, miR319B, miR156C, miR159B, miR839A, miR846A, miR857A, miR169C, and miR853A (Supplemental Table 3A). It would be useful to determine the functions of these PHL2- and PHL3-specifically regulated miRNAs in plant responses to Pi starvation.

A Venn diagram similar to the one created for the PSI genes was drawn for the commonly affected PSS genes (Figure 7B). Five genes commonly affected in all four phr single mutants were identified (Figure 7B, area “a”). These five genes encoded: (1) DRM2 (AT2G33830), a negative regulator of local and systemic acquired resistance; (2) a glycine-rich protein (AT3G20470); (3) a polyketide cyclase/dehydrase (AT1G70880); (4) a BTB and TAZ domain protein (AT5G63160); and (5) LHY, a putative transcription factor involved in circadian rhythm (AT1G01060). Among these genes, except LHY whose expression was further suppressed compared with that of the WT, the degree of suppression was reduced in all of the phr mutants. The functions of these five PSS genes in Pi responses are not known.

For the 82 PSS genes commonly affected in phr1, phl2, and phl3 (Figure 7B, area “b”), GO analyses showed that they were weakly enriched in photosynthesis and lipid transport. A large number of PSS genes (949) were commonly affected in phl2 and phl3 (Figure 7B, area “c”). These genes were highly enriched in the GO terms of ribosome biogenesis, translation, plastid translation, cellular response to oxidative stress, and others. The large number of PSI and PSS genes that are commonly regulated by PHL2 and PHL3 provides additional evidence that these two proteins function as one module that regulates transcriptional responses to Pi starvation.

We then examined the PSI and PSS genes whose expression was specifically affected by the individual PHR proteins, i.e. to identify the unique targets of the PHR proteins. We found 197 PSI genes that were specifically regulated by PHR1 (Figure 7A, area “d”). These PSI genes mainly belong to the functional categories of cellular response to Pi starvation, Pi transport, galactolipid biosynthetic process, and defense response to the bacterium. This meant that PHR1 alone regulated many genes involved in adaptive responses to Pi starvation. Eighty-five PHL2-specifically regulated PSI genes were weakly enriched in response to chitin (Figure 7A, area “e”), and 33 PHL3-specifically regulated PSI genes were weakly enriched in carboxylic acid transport (Figure 7A, area “f”). In terms of specifically affected PSS genes in phr mutants, 166 phr1-specifically affected PSS genes were enriched in photosynthesis (Figure 7B, area “d”), and 186 phl2-specifically affected PSS genes were enriched in photosynthesis, chloroplast organization, and response to light stimulus (Figure 7B, area “e”). There was no significant GO enrichment for the 65 phl3-specifically affected PSS genes (Figure 7B, area “f”).

The above results further indicated that PHR1 has a greater role than PHL1, PHL2, or PHL3 in regulating plant transcriptional response to Pi deficiency, because many PHR1-specifically regulated genes are highly enriched in the GO term of response to Pi starvation.

Interactions between PHR1 and the three PHLs in regulating PSI gene expression

The analyses of PHR mutation-affected genes indicated that all four PHR proteins participate in regulating PSR gene expression (Figure 4). Although PHL1 itself had only a small role in regulating PSR gene expression, it had a much larger effect when it acted with PHR1 in terms of both the number of genes affected and the magnitude of the effect on expression (Figures 4, A and D). By comparing the expression patterns of mutation-affected PSI genes among the two single mutants phr1 and phl (here, phl represents any of phl1, phl2, or phl3) and their corresponding double mutant (phr1phl), we found that there were some interactions between PHR1 and the three individual PHL proteins.

Here, we used PHR1 and PHL1 as an example to illustrate how these two proteins interact to regulate the expression of PSI genes. First, there were 439 PSI genes whose expression was not affected in phr1 or in phl1 but was affected in phr1phl1 (area “a” in Figure 8A), indicating a synergistic interaction between PHR1 and PHL1. K-means clustering divided these genes into two groups (Figure 8B and Supplemental Table 4A). The expression of one group (398 genes, 91% of the total) was downregulated in phr1phl1 compared with that in the WT; in contrast, the expression of another group (41 genes, 9% of the total) was upregulated in phr1phl1. These results indicated that PHR1 and PHL1 together could function as a promoter of one set of PSI genes (here, the Group (1) genes) and as a suppressor of another set of PSI genes (here, the Group (2) genes) (Figure 8B).

Figure 8.

Interactive effects of PHR1 and a PHL protein on the expression of PSI (Pi starvation-induced) genes. A, Venn diagram of the PSI genes whose expression was affected in phr1, phl1, and phr1phl1. B and C, Heatmaps showing the relative expression levels of the PSI genes in part “a” (B) and part “c” (C) in the WT, phr1, phl1, and phr1phl1. In (B) and (C), the histograms next to the heatmap show the different types of interactive effects of PHR1 and PHL1 on the expression of PSI genes. The same displays for the interactive effects of PHR1 and PHL2, and of PHR1 and PHL3 on PSI gene expression are shown in (D–F) and (G–I), respectively. The TPM (transcripts per million) values used in the heatmap have been standardized.

Another type of interaction between PHR1 and PHL1 was on those genes whose expression was affected in both phr1 and phr1phl1 but not in phl1 (area “c” in Figure 8A), or in both phl1 and phr1phl but not in phr1 (area “d” in Figure 8A). In addition to the above criteria, there should be a significant difference in the expression levels of the genes affected in phr1 and phr1phl or in phl and phr1phl. In these cases, the cutoff for the significant difference was arbitrarily set at Log₂FC ≥ 1 (or ≤ −1) and FDR < 0.05. With these criteria, we found that among the 196 genes affected in both phr1 and phr1phl1 but not in phl1 (part “c” in Figure 8A), 68 genes (Group (1) in Figure 8C) were downregulated in phr1 and were further downregulated in phr1phl1, indicating a synergistic interaction between PHR1 and PHL1. The other 16 genes (Group (2) in Figure 8C) were upregulated in phr1, but this upregulation was significantly reduced in phr1phl1, indicating an antagonistic interaction between PHR1 and PHL1 (Figure 8C and Supplemental Table 4A). In two genes affected in both phl1 and phr1phl1 but not in phr1 (Figure 8A, area “d”), the expression of one (JMJD5) was upregulated in phl1 and further upregulated in phr1phl1.

In summary, among the 1,595 PSI genes observed in the WT, 646 were commonly regulated by PHR1 and PHL1 (there were 64 PSI-up genes and 582 PSI-down genes in phr1phl1, Figure 4D). In these 646 commonly regulated genes, 78.5% (439 in Figure 8B and 68 in Figure 8C) were synergistically regulated and 2.5% (16 in Figure 8C) were antagonistically regulated by PHR1 and PHL1, respectively.

Similar analyses were performed for the interactions between PHR1 and PHL2, and between PHR1 and PHL3 in regulating PSI gene expression. The number and identities of the genes in the three above-mentioned classes are displayed in Figure 8, D–I and Supplemental Tables 4, B and C. Together, the results showed that the interactions between PHR1 and PHL1 were stronger than the interactions between PHR1 and PHL2 or PHR1 and PHL3.

Ancillary PSR genes regulated by PHR proteins

We also found that under Pi deficiency, the expression of some genes was not altered in the WT but was increased or decreased in the mutants. We designated these as ancillary genes, which were further divided into “ancillary-up” genes (upregulated in the mutant) and “ancillary-down” genes (downregulated in the mutant) genes (Figure 4). We found 370; 11; 1,979; and 1,676 ancillary-up genes and 99; 9; 1,185; and 838 ancillary-down genes in phr1, phl1, phl2, and phl3, respectively (Supplemental Table 5). Among the 370 ancillary-up genes in phr1, 272 were specifically regulated by PHR1 and were highly enriched in the GO terms of defense response to bacterium, response to chitin, response to salicylic acid, and protein phosphorylation (Supplemental Figure 15A). This suggested that PHR1 functions as a repressor of the defense response under Pi starvation (Figure 7C, part “d”). phl2 and phl3 shared 1,435 ancillary-up genes (Figure 7C, part “c”), which were highly enriched in the GO terms of ribosome biogenesis (Supplemental Figure 15, B, and 15C). phr1, phl2, and phl3 shared 54 ancillary-up genes, which were enriched in GO terms of defense response and response to biotic stimulus, indicating that PHL2 and PHL3 also regulate defense response genes but to a lesser extent than PHR1. That 451 phl2-specifically affected ancillary-up genes were moderately enriched in response to water deprivation (Figure 7C, part “e”) and that the 174 phl3-specifically affected ancillary-up genes were not enriched in any GO term (Figure 7C, part “f”), indicating that there were some subtle functional differences between PHL2 and PHL3 in regulating transcriptional responses to Pi starvation.

phr1 and phl1 had 99 and 9 ancillary-down genes, respectively, but the phr1phl1 double mutant had a large increase in the number of ancillary-down genes (861). These genes in the phr1phl1 double mutant were enriched in the GO terms of suberin biosynthetic process, phenylpropanoid biosynthesis, and lipid metabolism (Supplemental Figure 15D), indicating that PHR1 and PHL1 function together to prevent a decrease in secondary metabolism under Pi deficiency. phl2 and phl3 shared 721 ancillary-down genes (Figure 7D, part “c”), which were highly enriched in rRNA modification (Supplemental Figure 15, E and F), again suggesting that PHL2 and PHL3 regulate genes related to protein translation. These results further indicate that PHL2 and PHL3 have similar functions in regulating PSR genes and that the PHL2/PHL3 module largely controls the expression of genes involved in protein translation.

Discussion

The functions of PHR1 family proteins in regulating plant transcriptional responses have been well documented in a broad range of plant species. In Arabidopsis, five members of this family, namely PHR1 and PHL1–PHL4, have been characterized for their functions in regulating Pi responses (Rubio et al., 2001; Bustos et al., 2010; Sun et al., 2016; Wang et al., 2018). To date, however, there has been no comprehensive study of how each PHR member functions in regulating Pi responses and transcription of PSR genes at the genomic level. In this research, we first conducted comparative phenotypic analyses to investigate how PHR1, PHL1, PHL2, and PHL3 regulate plant development as well as plant responses to Pi starvation; we then performed comparative genomic analyses to elucidate how these four PHR proteins control PSR gene transcription.

Our phenotypic analyses of the WT and various phr mutants indicated that PHR1 and PHL1 form one functional module, and that PHL2 and PHL3 form another functional module in regulating plant development and responses to Pi starvation. These two distinct modules can be clearly differentiated based on the phenotypes of the phr mutants. The growth characteristics of all single phr mutants did not differ from those of the WT. The double mutant of phr1phl1, however, exhibited early leaf senescence, while the double mutant of phl2phl3 showed retarded growth of both shoots and roots (Figures 2 and 3, and Supplemental Figure 4). These results indicated that PHR1 and PHL1 act redundantly to control leaf senescence, whereas PHL2 and PHL3 act redundantly to regulate root and shoot growth.

PHR1 and PHL1 also function together to regulate developmental and physiological responses to Pi starvation, as shown by three lines of evidence. First, PHR1/PHL1 functions in maintaining PR growth under Pi starvation. Functional disruption of PHR1 caused plants to be more sensitive to PSI inhibition of PR growth, and the PHR1/PHL1 double mutations further increased such sensitivity (Figure 3A and Supplemental Figure 5). In contrast, the phl2phl3 double mutant was less sensitive than the WT to PSI inhibition of PR growth. Second, mutation of PHR1 caused a substantial reduction of root-associated APase activity, and mutation of both PHR1 and PHL1 further increased the severity of this defect (Supplemental Figure 7); mutation in PHL2 or PHL3, however, did not have an obvious effect on the induction of root-associated APase activity by Pi starvation. The root-associated APase activity was not further decreased even in the quadruple mutant phr1phl1phl2phl3 compared with that of phr1phl1, indicating that PHL2 and PHL3 do not play a substantially role in this Pi response. Third, phr1 and phr1phl1 have a significant decrease in Pi contents in shoots under Pi sufficiency, but an increased accumulation of Pi in roots under Pi deficiency. Such defects were not observed in phl2 or phl3 (Figure 3, B and C).

In regulating the expression of PSR genes at the genomic level, PHR1/PHL1 and PHL2/PHL3 also function as two distinct modules. This is quite evident in the heatmap that compared the expression profiles of PSR genes among the WT and various phr mutants (Figure 4A). This inference was also supported by our PCA of the RNA-seq data (Supplemental Figure 12B). When acting alone, PHR1 had a strong effect, PHL2 and PHL3 had similar but mild effects, and PHL1 had little effect on the expression of PSR genes (Figure 4 and Supplemental Table 1). However, when combined with PHR1, PHL1 had much stronger effects than PHL2 or PHL3 in regulating PSR gene expression. Also, the expression of a large number of PSI genes (206) and PSS genes (949) was commonly affected in phl2 and phl3. Furthermore, phl2 and phl3 share 1,435 ancillary-up genes and 721 ancillary-down genes. The different gene expression patterns controlled by these four PHRs are probably due to their unique protein structures, i.e. PHR1 and PHL1 contain a similar extra N-terminal segment, and PHL2 and PHL3 contain a similar extra C-terminal segment besides their shared MYB and coiled-coil domains (Figure 1B). Such structural differences may restrict the direct protein–protein interactions to those between PHR1 and PHL1 and between PHL2 and PHL3 and may not permit direct interactions between these two different sets of proteins (Supplemental Figure 9).

Interestingly, in the regulation of the transcription of PSR genes, PHR1 has a dominant role in the PHR1/PHL1 module, but PHL2 and PHL3 have almost equal roles in the PHL2/PHL3 module. This might be explained by the expression patterns of PHR1 and PHL1. PHR1 is ubiquitously expressed in all tissues, but PHL1 is predominantly expressed in vascular tissue, and the expression of PHR1 is much stronger than that of PHL1 in the entire plant (Figure 1).

We also found that mutation of PHR1 had much stronger effects on highly responsive PSI genes than mutation of PHL2 and PHL3 in terms of both the number of genes affected and the magnitude of gene expression changes (Figure 6). These highly responsive PSI genes include the key components of Pi signaling, such as miRNA399s, IPS1, SPX1, and SPX3, three Pi transporters, three APases, and one ribonuclease (Figure 6B). We also examined the effects of mutations of PHR genes on the expression of some genes known or putatively known to be involved in regulating Pi homeostasis (Figure 6C). The results showed that the induction of these genes by Pi starvation is more impaired in phr1 than in the other three phr mutants. These results also separate the function of PHR1 from that of PHL2/PHL3 in regulating PSI gene transcription and are consistent with our findings that phr1 displayed more obvious defects in response to Pi starvation than the other three phr mutants.

Our detailed analyses of the transcriptomic changes in the various phr mutants identified the common and specific targets for each PHR protein. The common targets are mainly related to Pi signaling, transport, and recycling, while the specific targets participate in more diverse biological processes (Supplemental Table 3). We also found that a given PHR protein or a combination of two PHR proteins can serve not only as a positive but also as a negative regulator of PSR gene expression depending on the target. It will be useful to determine the functions of these negatively regulated target genes in plant Pi responses.

In this research, although the results showed that PHR1/PHL1 and PHL2/PHL3 modules operate largely independently, some interactions do occur between them. Researchers previously reported that PHR1 and PHL1 could synergistically regulate the expression of PSR genes (Bustos et al., 2010), indicating an interaction between PHR1 and PHL1. In the current research, we found that such interactions also exist between PHR1 and PHL2, and between PHR1 and PHL3 (Supplemental Figure 9). Furthermore, we found that the interactions between PHR1 and the other three PHR proteins can not only be synergistic as previously shown for PHR1 and PHL1 (Bustos et al., 2010), but can also be antagonistic in regulating a given PSR gene (Figure 8, C, F, and I). Another example of the antagonistic interaction between PHR1/PHL1 and PHL2/PHL3 involves leaf senescence (Figure 2 and Supplemental Figure 4).

Finally, we point out that the common and unique targets identified for the four PHR proteins in this research include both direct and indirect targets. To further elucidate how these PHR proteins interact to regulate the transcription of PSR genes, it will be important to distinguish between direct and indirect targets using ChIP-seq combined with other approaches, such as electrophoresis mobility shift assays. A consensus binding sequence (termed the P1BS element) on the promoters of some PSI genes has previously been identified for PHR1 (Rubio et al., 2001). We showed that PHL2 and PHL3 could also bind to two regions on the promoter of the AtPAP10 gene, one region that contains P1BS-like elements, and another region that does not contain any sequence related to the P1BS elements (Sun et al., 2016). In the future, we intend to use ChIP-seq and other experimental approaches to identify new binding motifs on the promoters of PSR genes for these four PHR proteins. We would also like to determine whether PHR proteins bind to the different core sequences on the promoters of PSR genes, whether the promoters of the PSR genes that are commonly targeted by more than one PHR protein have a unique sequence architecture, and whether the promoters of the PSR genes that are regulated in different ways (i.e. positively versus negatively and synergistically versus antagonistically) contain different core sequences. Answering these questions will increase our understanding of how PHR proteins interact to regulate the transcriptional responses of plants to Pi starvation.

Materials and methods

Plant materials and growth conditions

All Arabidopsis (A. thaliana) plants used in this study were in the Colombia-0 (Col-0) ecotype background. The T-DNA insertional lines SALK_067629 (phr1), SAIL_731_B09 (phl1), SALK_114420C (phl2), and SALK_010040C (phl3) were obtained from the Arabidopsis Biological Resource Center (ABRC). Two additional phl3 alleles were generated using the CRISPR/Cas9-based genome editing method. The CRISPR lines that possessed heritable mutations of PHL3 without Cas9 T-DNA were used for further study. The double, triple, and quadruple mutants were generated through genetic cross. The primers used for genotyping of T-DNA lines and CRISPR lines are listed in Supplemental Table 6.

Surface-sterilized seeds were sown on agar or agarose plates containing the +Pi or −Pi medium. The +Pi medium contained half-strength Murashige & Skoog basal salts (Caisson Labs, MSP01-01190008), half-strength Murashige & Skoog vitamin powder (1000×) (Phyto Technology Laboratories, M533), 1.0% (w/v) sucrose, 0.5% (w/v) MES, and 1.2% (w/v) agar (Sigma–Aldrich, catalog no. A1296), or 0.8% (w/v) agarose (Biowest, catalog no. 111860). In the −Pi medium, half-strength Murashige & Skoog without phosphate (Caisson Labs, MSP11-05160009) was used to replace Murashige & Skoog basal salts. In this study, agarose was used as the gelling agent in the medium unless otherwise indicated. The residual P concentration in agarose was below the detection sensitivity of the inductively coupled plasma optical emission spectrometer (ICP-OES). The pH of the medium was adjusted to 5.8 with NaOH. After seeds were stratified for 2 days at 4°C, the medium plates were placed vertically in a growth room with a photoperiod of 16 h light and 8 h dark at 22–24°C. The light intensity was 100 μmol m⁻² s⁻¹.

The Nicotiana benthamiana seeds were sown directly in the soil. After about 10 days, vigorous seedlings were transplanted into new pots, and the seedlings were grown under the same lighting conditions as Arabidopsis.

Vector construction and plant transformation

The additional mutant alleles of PHL3 were generated according to Yan et al. (2015). In brief, the target site for CRISPR/Cas9 editing in PHL3 was determined by an online software (http://crispr.mit.edu/). The target site was located in the first exon to ensure an introduction of premature stop codon. The synthesized DNA fragment containing the target site of PHL3 was inserted between two Bsa I restriction enzyme sites in the vector AtU6-26-sgRNA-SK. The fragment between the Nhe I and Spe I site of AtU6-26-sgRNA-SK was cloned into the Spe I restriction enzyme site in pCAMBIA1300-pYAO:Cas9.

For the expression pattern analysis of PHR1 and PHL1, the 2-kb promoter sequences were amplified from genomic DNA by PCR with Xba I and Xma I restriction enzyme sites added to the ends of the PCR products. The amplified fragments were cloned into Xba I and Xma I sites before the GUS reporter gene in the plant transformation vector pBI101.

For LCI assays, the coding sequences of PHR1, PHL1, PHL2, and PHL3 were inserted into the vectors pCAMBIA-nLUC and pCAMBIA-cLUC (Chen et al., 2008; Zhao and Zhou, 2020), respectively, to generate PHR1/PHL1/PHL2/PHL3-nLUC and cLUC-PHR1/PHL1/PHL2/PHL3 constructs.

The primers used for vector construction are listed in Supplemental Table 6. All constructs were mobilized into Agrobacterium tumefaciens strain GV3101. The Agrobacterium strains were then transformed into Arabidopsis plants via the flower dip method (Clough and Bent, 1998) or were used for transient expression in the leaves of N. benthamiana (Batoko et al., 2000). The constructs of PHR1::GUS and PHL1::GUS carried the kanamycin-resistance gene as a selectable marker for plant transformation. The stable transgenic lines were selected using antibiotic-containing media.

Histochemical analysis of GUS activity

Histochemical analysis of GUS activity was carried out as described by Jefferson et al. (1987). Plants tissues were incubated at 37°C overnight in GUS staining buffer (2 mM X-Gluc in 50 mM sodium phosphate buffer, pH 7.2) containing 0.1% Triton X-100, 2 mM K₄Fe(CN)₆, 2 mM K₃Fe(CN)₆, and 10 mM EDTA. The stained seedlings were sequentially transferred to 50%, 70%, and 100% (v/v) ethanol to remove chlorophyll. The stained materials were photographed with a camera attached to a stereomicroscope (Olympus SZ61).

Analysis of root-associated APase activity

Histochemical staining of root-associated APase activity was performed as described by Wang et al. (2011). The roots of seedlings that had grown vertically for 7–8 days were overlaid with a 0.5% agar solution containing 0.01% BCIP (5-bromo-4-chloro-3-indolyl-phosphate) at 23°C. After 16 h, the roots were photographed with a camera attached to a stereomicroscope (Olympus SZ61).

Quantification of cellular Pi and total P contents

Cellular Pi content were measured as described by Wang et al. (2011). Specifically, the weighed fresh shoot and root tissues were submerged in 1 ml of 1% glacial acetate and were alternatively freeze–thawed 10 times in liquid nitrogen and in a water bath maintained at 65°C. A 100-μl volume of the extract was mixed with 200 μl of ddH₂O and 700 μl of Pi reaction buffer containing a mixture of [0.48% NH₄MoO₄, 2.85% (v/v) H₂SO₄] and [10% (w/v) ascorbic acid] in a ratio of 6:1. The reaction was allowed to proceed at 37°C for 1 h. The cellular Pi content was determined at A₈₂₀ according to a prepared standard curve and was expressed as μmol/g FW.

To determine the total P content, about 50 mg of fresh tissues was oven-dried at 500°C for 3 h and flamed to ash. The ashes were dissolved in 100 µl of 30% (v/v) HCl and 10% (v/v) HNO₃. About 10 µl of the dissolved sample was mixed with 290 µl of ddH₂O and 700 µl of Pi reaction buffer, and the Pi was quantified by above method. The total P contents of plant tissues were determined and expressed as Pi contents extracted from flamed ashes.

LCI assays

The Agrobacterium strain harboring different constructs were grown to a cell density of OD₆₀₀ = 0.5 and were harvested and resuspended in infiltration buffer (10 mM MES, 0.2 mM acetosyringone, and 10 mM MgCl₂) to the same concentration. The Agrobacterium strains were infiltrated into the leaves of 4-week-old N. benthamiana. Forty-eight hours later, the leaves were painted uniformly with luciferin (100 mM) dissolved in 0.1% Triton X-100. After a 5-min exposure in the dark, the luminescence images were captured with an Andor iXon CCD camera (Andor Technology).

Reverse transcription quantitative PCR (Rt-qPCR) analyses

Total RNAs of 8-day-old seedlings were extracted using the Magen HiPure Plant RNA Mini Kit. A 2-μg quantity of RNA was reversely transcribed to cDNA using M-MLV reverse transcriptase (Takara). RT-qPCR analyses were carried out using EvaGreen 2 × qPCR MasterMix (ABM) on a Bio-Rad CFX96 real-time PCR detection system. The ACTIN 2 gene (AT3G18780) was used as an internal control, and the relative expression level of each gene was calculated by the 2^−ΔΔCt method (Livak and Schmittgen, 2001). The primers used for RT-qPCR analysis are listed in Supplemental Table 6.

Transcriptomic analyses

Total RNAs were isolated from roots of 8-day-old seedlings of the WT and various mutants grown under Pi sufficiency and deficiency using the RNeasy plant mini kit (Qiagen, catalog: 74904). Three biological replicates were used for each genotype. All of the mutants used were T-DNA lines, except that the phl3 single mutant and phl3 in phr1phl1phl2phl3 were CRISPR/Cas9 knockout lines. mRNAs were purified from total RNAs using Oligo d(T) beads and were reversely transcribed into cDNAs. The cDNAs were fragmented, and sequencing libraries were prepared by end repairing, “A” tail addition, and adaptor ligation. The quality of the libraries was analyzed by the Agilent 2100 Bioanalyzer and ABI StepOnePlus Real-Time PCR System. The libraries that passed the quality control were sequenced using the Illumina platform by BIONOVA Co. (Beijing). The clean sequencing data were aligned into the Arabidopsis TAIR 10 genome by Tophat software. The Q30 values of all RNA-seq samples were greater than 90%, and the ratios of mapped rates to clean reads were above 90%. Quantification of transcripts and comparative analyses among the mutants were performed with EdgeR software using TPM (transcripts per million). Differentially expressed genes were selected using EdgeR software. GO analysis was performed with two online softwares, agrigo v2 (Tian et al., 2017) and metascape (Zhou et al., 2019). The Venn diagrams were drawn by Venny 2.1.0 (Oliveros, 2007-2015). The heatmaps were drawn by an online software, Morpheus (https://software.broadinstitute.org/morpheus/). The PCA was conducted by TBtools software (Chen et al., 2020).

Statistical analysis

Significant differences between treatments and genotypes were determined using ANOVAs with P < 0.05 followed by Tukey's multiple comparisons tests (GraphPad prism v6.01) or Student's t-tests with P < 0.05 (Excel). At least three independent biological replicates were included in each analysis.

Accession numbers

Sequence data for the genes characterized in this article can be found in the Arabidopsis Genome Initiative or GenBank/EMBL databases under the following accession numbers: AT4G28610 (PHR1), AT5G29000 (PHL1), AT3G24120 (PHL2), AT4G13640 (PHL3), AT2G34202 (MIR399D), AT3G09922 (IPS1), AT5G43350 (Pht1;1), AT2G38940 (Pht1;4), AT2G16430 (AtPAP10). The raw data of Illumina reads are available at the National Center for Biotechnology Information Sequence Read Archive browser (http://ncbi.nlm.nih.gov/sra) under the accession no. GSE217158.

Supplemental data

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

Supplemental Figure S1. Transcriptional responses of PHR1::GUS and PHL1::GUS to Pi starvation.

Supplemental Figure S2. Relative expression of the four PHR genes in their corresponding mutants.

Supplemental Figure S3. Growth phenotypes of three phl3 mutant alleles.

Supplemental Figure S4. Leaf senescence phenotypes of the phr mutants.

Supplemental Figure S5. Quantitative analysis of primary root growth of the WT and various mutants in Figure 3A.

Supplemental Figure S6. Root hair phenotypes of 8-day-old seedlings of the WT and various phr mutants grown under Pi-sufficient and Pi-deficient conditions.

Supplemental Figure S7. Root surface-associated APase activity of 8-day-old seedlings of the WT and various mutants grown under Pi starvation conditions.

Supplemental Figure S8. The total P content of the WT and various mutants grown under Pi sufficiency and Pi deficiency.

Supplemental Figure S9. Physical interaction among PHR proteins.

Supplemental Figure S10. Expression of five PSI marker genes in the WT and various mutants as determined by RT-qPCR and RNA-seq.

Supplemental Figure S11. List of the representative PSR gene-enriched GO terms.

Supplemental Figure S12. The PCA of the relationship among the phr mutants under Pi starvation.

Supplemental Figure S13. GO annotation of differentially expressed genes in phl2 and phl3 under normal growth condition.

Supplemental Figure S14. GO annotation of the six groups of PSR genes defined according to the degree of their induction or suppression triggered by Pi starvation.

Supplemental Figure S15. GO annotation of mutation-affected ancillary genes in phr mutants.

Supplemental Table S1. Genome-wide identification of PSR genes and effects of PHR mutations on their expression.

Supplemental Table S2. Expression values of the PSR genes used to generate the heatmap in Figure 4A.

Supplemental Table S3. Common and unique PSR and ancillary genes.

Supplemental Table S4. Interaction of PHR1 and PHL1, PHR1 and PHL2, or PHR1 and PHL3 in regulating PSI genes.

Supplemental Table S5. Ancillary genes of various phr mutants.

Supplemental Table S6. Sequences of the primers used in this study.

Funding

This work was supported by funds from the National Natural Science Foundation of China (grant no. 32070298) and the Ministry of Science and Technology of China (grant no. 2016YFD0100700).