Introduction

Phosphorus (P) is an essential macronutrient for plant growth and development, involving the composition of membrane phospholipids and nucleic acids and metabolic roles such as energy storage and transfer (Rouached et al. 2010; Veneklaas et al. 2012). However, P exists primarily in soils as hardly soluble phosphate compounds combined with the mineral elements, such as aluminum phosphate (AlPO4, AlP) and the aluminum salt of phosphoric acid in acidic acid soils, and it is difficult for plants to absorb (Pradhan et al. 2017). Therefore, P is a limiting nutrient for plants due to its low availability of P in soils (Augusto et al. 2017). Certainly, in P-poor environments, plants have developed general strategies for obtaining and benefiting from P, including exudation of compounds (Ryan et al. 2001; Shen et al. 2003), root structural specializations (Hu et al. 2010; Williamson et al. 2001), and mycorrhizal symbioses (Smith and Smith 2011; Wan et al.2018), which all can lead to an increase in the bioavailability of AlPO4 in acidic soils to plants. However, the molecular mechanism of plant response to hardly soluble AlPO4 is poorly understood. Various sequencing technologies have recently provided critical information about gene expression changes in some plant species in response to P limitation. Transcriptome analysis of barley revealed that many genes were significantly upregulated or downregulated in response to low P stress. Furthermore, differentially expressed genes (DEGs) were discovered to be primarily involved in P metabolism, sucrose synthesis, phospholipid degradation, hydrolysis of phosphoric enzymes, phosphorylation/dephosphorylation, and post-transcriptional regulation. (Ren et al. 2018). Furthermore, when treated with low phosphorus, DEGs are enriched in carbohydrate metabolic processes, oxidation–reduction processes, biosynthetic processes, and the tricarboxylic acid cycle in oat roots (Chao et al. 2017). To gain a better understanding of these processes, DEGs were studied under low P stress in other crops, including Zea mays L.(Du et al. 2016), Oryza sativa L. (Deng et al. 2018), and Glycine max Linn (Liu et al. 2020), using transcriptome analysis. The above results of studies show that transcriptome can provide more information on the gene regulation related to low P adaptation for plants.

Artemisia annua is an annual herb appreciated for producing artemisinin, a sesquiterpene molecule used to treat fever and malaria (Baraldi et al. 2008; Ma et al. 2007; Wani et al. 2021; Wani et al. 2022). It is widely distributed in most Chinese areas, especially in southwest China (Zhang et al. 2017). It adapts well to different soil types and has no specific nutritional requirements; however, P and potassium (K) supply stimulate its growth even in small quantities (Aftab et al. 2014; Müller and Brandes 1997). For example, Todeschini et al. (2022) showed that P nutrition affected A. annua plant biomass production, and its lowest level led to the highest artemisinin concentration. Therefore, optimizing P supply to A. annua is essential for maximizing dry matter production and/or artemisinin yield (Todeschini et al. 2022). Many studies have explored the ability of crops to acquire P from various hardly soluble forms (Sharma et al. 2013; Giles et al. 2014; Lambers 2022; Lee et al. 2012; Li et al. 2015; Pearse et al. 2007). Currently, minimal effort has been made to understand the adaptive strategies in A. annua against Plimit. A controlled greenhouse experiment was conducted to learn more about the genetic behavior of A. annua in response to the availability of a hardly soluble P source, AlP, to simulate low P availability in acid soil, with the soluble phosphorus form, KP, serving as a control. We then explored the A. annua growth and genetics response to the hardly soluble P source, AlP, via second-generation sequencing analysis. This study could deepen our understanding of the genetic variation of A. annua under low phosphorus availability and suggest strategies to improve its P-use efficiency and the production of biomass and artemisinin with less fertilizer application.

Material and methods

Plant materials

Artemisia annua seeds were collected from the planting area in the Guangxi Medicinal Botanical Garden scientific research base in Nanning, China (108°23′ E, 22°51′ N). The seeds of A. annua were sowed into a plastic container (30 cm × 20 cm × 8 cm, length × width × height, respectively) filled with washed and sterilized river sand and then rinsed with distilled water until the sand was wet, every 2 days during germination. The seedings were supplied with 200 ml of pH 6.5 half-strength Hoagland nutrient solution weekly. All the seedings of similar size with two cotyledons were on standby for the two phosphorus treatments.

Different phosphorus treatments

There were two P sources used in this study: a hardly soluble P source: aluminum phosphate (AlPO4)/(AlP), the water solubility of which is only 1.89 × 10−9 g/100 ml at 20 °C, and a soluble P source: monopotassium phosphate (KH2PO4)/(KP), the water solubility of which is 22.6 g/100 ml at 20 °C. The AlP was the low-P-availability treatment group, and the KP was the control group; each group was replicated five times. River sand was used as a cultivation substrate in this experiment. All the river sand before the experiment was sieved through a 2-mm mesh and cleaned to remove nutrients with running water and then autoclave sterilized for 30 min. The flower pot used for holding the river sand was a height of 12 cm and a diameter of 4 cm. P supplements were added in powder form, 118.06 mg AlPO4 and 131.83 mg KH2PO4, respectively, mixed with the treated river sand, to ensure that each flower pot contained 30 mg P content. After the transplantation of seedlings, all flower pots were put in the greenhouse, with illumination intensity 300 µmol m−2 s−1 during the day, temperature 26 ± 2 °C, and relative humidity 62 ± 2%. All the flower pots’ positions were changed randomly to avoid the influence of environmental differences. Each flowerpot was supplemented with enough distilled water every 2 days, and a 5-ml Hoagland nutrient solution (0.5 × , without P) was added each week. The P treatment was sustained for 3 months (Pearseet al. 2007; Wanet al. 2018).

Plant growth parameters and sample collection

After the different P source treatments, the plant height, stem diameter, leaf number, and leaf area were measured before seedlings were harvested. The leaf area was taken with a digital photo and then calculated with ImageJ software (National Institutes of Health, USA). Three fresh leaves of every flowerpot were collected and wrapped with aluminum foil, immediately frozen in liquid nitrogen for 3–5 min, and then stored in a − 80 ℃ until RNA-Seq and real-time quantitative PCR (RT-qPCR). And then, the final biomass of roots, stems, and leaves were counted after drying at 60 ℃ for 72 h.

Isolation of Artemisinin and Analysis by LC–MS

The leaves were dried at 60 ℃ and prepared for artemisinin determination. Samples of 0.2 g of dried A. annua leaves were extracted using 25 ml petroleum ether (boiling point 30–60) for 40 min with ultrasonic waves, filtrated, transferred to 100 ml evaporating dish, and dried at 40 ℃; then, the evaporating dish was rinsed with methanol repeatedly. All solutions were kept in a volumetric flask at a constant volume of 10 ml (Stringham et al. 2018). The sample extracts were analyzed using an LC–ESI–MS/MS system (HPLC, EXPEC 5210 system1). The analytical conditions were as follows: HPLC column, Waters ACQUITY UPLC BEH C18 (1.7 μm 2.1 × 150 mm); solvent system, water (0.1% acetic acid): acetonitrile; gradient program, 80:20 V/V at 0 min, 5:95 V/V at 4.0 min, 5:95 V/V at 6 min, 80:20 V/V at 6.1 min, 80:20 V/V at 9 min; flow rate, 0.3 ml /min; temperature, 40 ℃; and injection volume, 2 µl. LIT and triple quadrupole (QQQ) scans were acquired on a triple quadrupole linear ion trap mass spectrometer EXPEC 5210 LC/MS/MS system equipped with an ESI. The ESI source operation parameters were as follows: ion source, turbo spray; source temperature, 105 ℃; capillary voltage, (IS) 4800 V; source offset voltage, 700 V; desolvation temperature, 495 ℃; cone gas flow, 72 l/h; and desolvation gas flow300 l/h. Instrument tuning and mass calibration were performed with 10 and 100 µmol/l polypropylene glycol solutions in QQQ and LIT modes, respectively. QQQ scans were acquired as MRM experiments with collision gas (nitrogen) set to 5 psi. DP and CE for individual MRM transitions were done with further DP and CE optimization. A specific set of MRM transitions were monitored for each period according to the metabolites eluted within this period.

RNA extraction and transcriptome sequencing

The RNAprep Pure Plant Kit (TIANGEN, Beijing, China) was used to extract total RNA from frozen samples. The purity of the RNA was determined using the manufacturer’s protocols and a KaiaoK5500 Spectrophotometer (Kaiao, Beijing, China). An RNA Nano 6000 Assay Kit of the Bioanalyzer 2100 system was used to assess RNA concentration and integrity (Agilent Technologies, CA, USA). RNA degradation was monitored on agarose gels. Three replicates of each treatment were deemed high quality and used to build transcriptome libraries. The cDNA construction library refers to the following reagents: oligo (dT) magnetic beads for enriched mRNA from total RNA, divalent cations under elevated temperature in NEB Next First Strand Synthesis Reaction Buffer (5 ×) for fragmentation, random hexamer primer and RNase H for synthesizing the First-strand cDNA, and DNA polymerase I, buffer, dNTPs, and RNase H for synthesizing second-strand cDNA. QiaQuick PCR kits and elution with EB buffer were used to purify the library fragments, and then, the terminal repair, A-tailing, and adapter were implemented. The aimed products were retrieved, PCR was performed, and the library was completed. The library quality was evaluated using the Step One Plus Real-Time PCR Agilent Bioanalyzer 2100 systems. The BioNovo Gene Technology Co., Ltd. (Suzhou, China) sequenced six libraries using an Illumina HiSeq 2500.

qRT-PCR analysis

Total RNA extraction, reverse transcription, and qPCR were operated as described previously (El-Sappah et al. 2021). β-actin mRNA was used as an internal control; all primers, including β-actin as a reference gene, were designed with the Primer 5.0 software (Leišová-Svobodová et al. 2020; Ahmed et al. 2023). The gene-specific primers used for qPCR are listed in Table S1. Gene expression was relatively quantified using the MM Ct method, as described by Livak and Schmittgen (El-Sappah et al. 2023). Each treatment had three replications, and the experiment was performed thrice. The mean and standard errors (SEs) were presented using ORIGIN 8.6 (Singh et al. 2021).

Statistical analysis

The one-way ANOVA was performed using the SPSS Statistics 19.0 software (IBM Corp, Armonk, NY, USA), and the Duncan test was used to determine significant differences (P < 0.05). The data were presented as mean ± SD (standard deviation). The sample was sequenced on the machine, and the software in the sequencing platform was transformed to generate the raw data (RAW Data) of FASTQ, which is the off-machine data. The raw data of each sample was used for statistical analysis. The connector sequence of 3′ ends was removed with Cutadapt, using the HISAT2 software to compare the filtered reads to the reference genome. The read distribution compared to the genome was statistically divided into CDS (coding region), intron (intron), and so on. HTSeq statistics, with three statistical schemes, was used to compare the Read Count value on each gene to the original expression of the gene. RSeQC was used to analyze expression saturation; DESeq was used to analyze the difference in gene expression and to screen the expression of different genes. GO enrichment analysis was performed using top-GO.

Results

Growth parameter assessment and artemisinin concentration

Plant growth parameters, including the plant height, the number of leaves, the stem diameter, and the leaf area, were significantly reduced under the AlP compared to the KP (Fig. 1A-E). Moreover, the leaf biomass and the root biomass had a slight reduction, but the stem biomass recorded a significant reduction under the AlP, compared to the control, KP, which led to a significant reduction in the total biomass of A. annua (Fig. 1F). Conversely, LC–MS analysis revealed not only no decrease but also a slight increase in artemisinin concentration under the AlP, compared to the KP (Fig. 1G).

Fig. 1
figure 1

Plant morphology, growth parameter assessment, and artemisinin concentration in response to the hardly soluble phosphorus source (AlP) compared to control (KP). A Plant height under the AlP, B plant height under the KP, C leaf area under the AlP, D leaf area under the AlP, E morphological traits, F biomass measurements, and G artemisinin concentration (mg kg.−1). Different letters (a and b) indicated significant differences between the AlP and the KP (P < 0.05, Student’s t-test)

Full size image

Transcriptome sequencing and the DEGs

The cDNA libraries were constructed for generating transcriptome sequences using A. annua leaves of the KPas, a control sample and the AlP as a hardly soluble Psource. Sequence libraries were prepared on the Illumina NextSeq 500 platform from 100 × to 120 × depth. Raw data from paired-end sequencing-by-synthesis generated 44,281,438, and 45,146,916 bp reads from the KP and the AlP, respectively. For all samples, the maximum read length was 6,910,596,600 bp. Following quality control and raw read data processing, 40,888,780 and 41,961,136 reads from the KP and the AlP, respectively, were retained for further assembly. Filtered reads were assembled, and transcripts were generated using Trinity at a hash size of 25. As a result of assembly, 39,274 transcripts of the KP vs the AlP were obtained in treatment comparatives with average transcript lengths of 283.58 bp for the KP vs the AlP (Table S2). The assembled transcript from different replicates showed variation in its numbers, possibly due to variable P absorbed by the plants or noises caused by technology at some point during the sequencing process. The distribution pattern of these transcripts is presented in Fig. 2.

Fig. 2
figure 2

Overview of the RNA-seq data and distribution of DEGs. A Principal component analysis (PCA) of the RNA-seq output. The PCA plot is calculated based on the transcriptome-wide profiles of gene expression. Distances between samples reveal differences in the transcriptome profiles between the samples. B Volcano plots display differentially expressed transcripts. Each dot represented a DEG; dots above the red line displayed the significant DEGs (P < 0.05)

Full size image

The principal component analysis (PCA) was used to assess variability between RNA-seq experiments. The PCA results revealed a strong correlation between the three replicates at two different treatments (Fig. 2A). A total of 762 DEGs were identified through a comparison of control to treatment (the KP vs the AlP) (323 upregulated, 439 downregulated). To identify common transcripts in the DGE data, transcripts exclusive to the low P availability treatment and the control treatment and downregulated and upregulated transcripts were analyzed for overlap (Fig. 2B).

Identification of the GO and KEGG enrichment analysis

All DEGs were assigned 210 GO terms (P < 0.05), divided into three categories: molecular function, cellular component, and biological process. The top 20 enriched GO terms were shown in Fig. 3 and Fig. S1, with the largest two terms being “protein localization (GO:0,072,662)” and “enzyme inhibitor activity (GO:0,004,857)” from the “biological process” and “molecular function” categories, respectively (Table S3). Furthermore, the MAPK signaling pathway, plant-pathogen interaction, and aminoacyl-tRNA biogenesis subcategories of the “molecular function” category were significantly enriched in more than 50 DEGs.

Fig. 3
figure 3

GO enrichment analysis of all the DEGs between the AlP and the KP. The top 20 enriched GO terms were presented. The horizontal axis represented the rich factor, while the vertical axis represented the GO terms. Number: DEG number; P adjust: adjusted P value

Full size image

To describe enriched biological pathways, a KEGG pathway enrichment analysis was performed (Fig. 4; Fig. S2; Table S4). The DEGs under the KP, compared to the AlP, were most significantly enriched in “catalytic activity,” “cellular metabolic process,” and “ion binding” categories. Moreover, this comparison was significantly involved in “glyoxylate and dicarboxylate metabolism,” “RNA degradation,” “glycine, serine and threonine metabolism,” and “proteasome” (Fig. 4).

Fig. 4
figure 4

The KEGG enrichment analysis of DEGs between the AlP and the KP. The most enriched KEGG pathways were presented. The horizontal axis represented the rich factor, while the vertical axis represented the pathway names. Number: DEG number; P adjust: adjusted P value

Full size image

Analysis of DEGs involved in hormone biosynthesis and signal transduction

Several genes involved in phytohormone, tryptophan, carotenoid, and phenylalanine acid signaling were differentially expressed under AlP treatment, according to RNA-seq and RT-PCR analyses (Fig. 5B, C). For example, in the brassinosteroid and carotenoid acid metabolic pathways, DEGs 3 and 2, respectively, were differentially expressed. In the tryptophan biosynthesis pathway, the expression levels of Gretchen Hagen 3 (GH3, CTI12_AA417000) and small auxin-up RNA (SAUR, CTI12_AA600200) were upregulated (Fig. 5B, C).

Fig. 5
figure 5

A Plant hormone signal transduction KEGG pathway in response to the hardly soluble phosphorus source (AlP). The red shapes represented upregulated genes under the AlP; the blue shapes represented downregulated genes under the AlP; the shapes marked with yellow represented both the AlP and the KP upregulated genes. B Validation of RNA sequencing results by quantitative real-time PCR (qRT-PCR) of selected genes. C Heat maps showing variations in the expression of genes involved in hormone signal transduction biosynthesis under the AlP and the KP

Full size image

The cytokinin response 1 (CRE1, CTI12_AA197330) gene involved in cytokinin (zeatin) signal transduction was significantly downregulated in response to AlP stress.

In brassinosteroid signal transduction, the BRI1-associated receptor kinase 1 BAK1, (CTI12_AA100010), BSK (CTI12_AA091650), and CYCD3 (CTI12_AA210800) were upregulated, whereas BSK (CTI12_AA380960) was downregulated. Only PR1 (CTI12_AA324860) showed upregulation in response to the AlP treatment in phenylalanine metabolism.

Analysis of DEGs involved in the artemisinin biosynthesis pathway

Our transcriptome analysis revealed many DEGs, such as the artemisinin biosynthetic genes. Seven essential artemisinin biosynthesis-related structural genes were examined, and their expression was investigated further using RT-PCR (Fig. 6). There are two independent pathways that lead to isopentenyl diphosphate (IPP) in the synthesis of artemisinin: the mevalonate (MVA) pathway and the methylerythritol phosphate (MEP) pathway in the cytosol and plastid, respectively (Fig. 6A).

Fig. 6
figure 6

A Artemisinin backbone biosynthesis KEGG in response to the hardly soluble phosphorus source (AlP). The red shapes represented upregulated genes under the AlP; the blue shapes represented downregulated genes under the AlP; the shapes marked with yellow represented both the AlP and the KP upregulated genes. B Validation of RNA sequencing results by quantitative real-time PCR (qRT-PCR) of selected genes. C Heat maps showing variations in the expression of genes involved in artemisinin biosynthesis under the AlP and the KP

Full size image

None of the associated genes in the MEP pathway displayed altered expression in response to the AlP treatment. In the MVA pathway, only the acetyl-CoA acetyltransferase (ACAT, CTI12_AA520360) was upregulated, whereas the 3-hydroxy-3-methylglutaryl-CoAreductase (HMGR) was downregulated. When IPP and dimethylallyl diphosphate (DMAPP) are condensed to farnesyl diphosphate (FPP) via a farnesyl diphosphate synthase (FPS) catalyzed reaction, two distinct pathways are initiated, with the FPS (CTI12_AA302700) demonstrating upregulation. In the steps from farnesyl diphosphate to artemisinin formation, many genes, such as the cytochrome P450 monooxygenase (CYP) gene family, showed differential expression under the AlP treatment. Our transcriptomic findings, consistent with our expression findings, show amorphadiene monooxygenase (CYP71AV1, CTI12 AA566140), which has previously been shown to play a role in the artemisinin synthesis pathway, is being upregulated. The aldehyde dehydrogenase (ALDH, CTI12_AA008900) gene catalyzes dihydroartemisinic aldehyde conversion to dihydroartemisinic acid was upregulated (Fig. 6B, C).

Identification of TFs related to AlP treatment

To better understand A. annua’s transcriptional regulation mechanisms under different P form treatments, 30 TFs changed dramatically in response to the AlP (Fig. 7A). The 17 differentially expressed TF gene families were classified; 12 upregulated TF genes (Fig. 7B; table S5) and 18 downregulated TF genes (Fig. 7C; table S6) were obtained using the Plant TFDB database.

Fig. 7
figure 7

TF family percent distribution.) The upregulated TFs in response to different phosphorus treatments, and B the downregulated TFs in response to different phosphorus treatments

Full size image

MYB, which contained three members, constituted the most prominent upregulated family among these transcription families. Two families, ERF and WRKY, were the largest among the downregulated TFs, with ten members (Fig. 7).

Discussion

After nitrogen (N), P is the second most crucial nutrient for plants (Amarasinghe et al. 2022; Kvakić et al. 2020). Many biological structures and functions of plants depend on the participation of P, such as the creation of nucleic acids (DNA and RNA), photosynthesis, glycolysis, respiration, membrane formation and stability, and enzyme activation and inactivation (Malhotra et al. 2018; Shen et al. 2011; Todeschini et al.2022; Vance et al.2003). Therefore, the P availability in the growth medium has been shown to have important effects on plant growth and development in numerous studies (Malhotra et al. 2018; Todeschini et al. 2022). P can be found in soil as mineral salts or organic compounds (Cordovil et al. 2020); however, most are hardly soluble (Miller et al. 2010). Therefore, our study compared the effect of AlPO4, as a hardly soluble P source, with the control KH2PO4, as a water-soluble P source. A. annua also showed a significant reduction in growth parameters, such as the plant height, the number of leaves, the stem diameter, and the leaf area (Fig. 1A–E), which also led to a reduction of total biomass under the AlP, compared to the KP (Fig. 1F). Interestingly, the artemisinin concentration was not reduced but significantly increased under the AlP, compared to the KP which was similar to Todeschini et al. (2022) study that showed the inverse relationship between P level and artemisinin concentration. However, plant growth and secondary metabolism responses to P availability are quite complex. For example, optimizing A. annua’s P and boron supply is critical for increasing dry matter production and/or artemisinin yield (Lulie et al. 2017). The yield of faba beans can be increased up to a certain level of P fertilizer application but exceeding that level decreases yield. Furthermore, Kebede et al. (2018) and Singh (2000) found no significant increase in oil content as P concentration increased. Again, limitations in the published data describing A. annua responses to P (Davies et al. 2011; Liu et al. 2003) indicated that artemisinin concentration decreased when P application was greater than 200 mg l−1 (KH2PO4). This decrease in artemisinin production occurred at a P concentration similar to that at which no further increase in plant growth was observed, and the amount of artemisinin per plant decreased dramatically (Liu et al. 2003). Furthermore, the increase in artesiminin concentration under stress, despite the reduction in plant growth parameters, may be due to artesmine’s expected essential role in elevating the harmful effect of stress and participating in plant cell protection. Many previous studies support our hypothesis because they discovered that artemisinin content (except in severe drought) and biosynthetic pathway genes are generally linked (Qureshi, et al. 2005; Yadav, et al. 2017; Vashisth, et al. 2018).

Hence, we performed transcriptomic analysis to gain more knowledge about the genetic behavior under the hardly soluble P source and more knowledge about which genes were responsible for these findings. cDNA libraries were used to generate transcriptome sequences. Using the leaves of the A. annua plant, 762 DEGs were identified in the comparison between the AlP vs. The KP (323 upregulated, 439 downregulated). The results of the GO enrichment analysis revealed that the two most enriched GO terms were “protein localization” and “enzyme inhibitor activity” (Fig. 3). Furthermore, the subcategories “catalytic activity,” “cellular metabolic process,” and “ion binding” were significantly enriched in the 1459 DEGs (Table S2). Based on KEGG pathway enrichment analysis, “glyoxylate and dicarboxylate metabolism,” “RNA degradation,” “glycine, serine and threonine metabolism,” and “proteasome pathways” were significantly enriched under the AlP compared to the KP (Fig. 4; Table S2). These findings suggested that hardly soluble P source could have an adverse effect on the regulation of artemisinin accumulation in A. annua. The AlP treatment also increased the expression of some structural genes involved in artemisinin biosynthesis.

Previous studies have shown that endogenous hormones are necessary for plant growth and development (Pacifici et al. 2015). In RNA sequencing data and RT-PCR analysis (Fig. 5B, C), it was discovered that the AlP treatment varied the expression of many genes involved in tryptophan, carotenoid, and phenylalanine acid pathways. The differentially expressed 3 and 2 DEGs were in the brassinosteroid and carotenoid acid metabolic pathways, respectively. These DEGs are widely thought to control plant growth and stress adaptation (Shi et al. 2020). Tafvizi et al. (2009) and Chen et al. (2007) previously investigated 14 and 10 differentially expressed genes (DEGs) that regulate cotton plant growth in the cytokinin (zeatin) and GA production pathways, respectively. The expression levels of GH3 and SAUR were upregulated in the tryptophan biosynthesis pathway, which promoted cell expansion and might be one of the causes of the maintaining artemisinin concentration under the AlP. According to earlier research, auxin can quickly and briefly increase the expression of three gene families, the SAUR family, the GH3 family, and the Aux/IAA family, which regulate plant development and growth. The auxin response factors (ARFs), controlling most SAUR, GH3 and Aux genes, activate or repress the expression of target genes (Woodward and Bartel 2005). In addition, overexpression of GH3-8 causes abnormal plant morphology as well as slowed growth and development in rice (Ding et al. 2008). On the other hand, the SAURs are the most common family of early auxin response genes, and they play a crucial role in regulating plant growth and development via hormonal and environmental cues (Ren and Gray 2015). Conversely, in our investigation, a factor that decreased cell division and shoot initiation was the downregulation of CRE1. As demonstrated by Laffont et al. (2015) in Medicago truncatula, the cytokinin CRE1 pathway influences root development and tolerance to abiotic and biotic environmental challenges in addition to being necessary for symbiotic nodule organogenesis. In our study, in brassinosteroid signal transduction, both BAK1, BSK (CTI12_AA091650), and CYCD3 were upregulated, whereas the BSK (CTI12_AA380960) was downregulated, which would be expected to have a significant effect on cell division and plant length. Although it has already been established that BSKs and BAK1 are both substrates of the BRI1 kinase, there is evidence to suggest that they have different functions in brassinosteroid signaling (Tang et al. 2008). Moreover, BSK is a crucial family of receptor-like cytoplasmic kinases (RLCK) in the first step of BR signal transduction, activating downstream phosphatase BSU1 (Kim et al. 2009).

Regarding artemisinin biosynthesis, the sesquiterpene route involves numerous enzymatic steps to produce artemisinin (Xie et al. 2016). In the artemisinin syntheses pathway, two distinct mechanisms, the MEP pathway in the plastid and the MVA pathway in the cytosol, are used to generate isopentenyl diphosphate, as shown in Fig. 6A (Vranová et al. 2013). Genes implicated in the MVA and MEP pathways have been described in numerous plants. Our transcriptome analysis uncovered many differentially expressed genes that were identified as producing artemisinin. In our investigation, the MVA pathway had all of the DEGs, whereas the MEP pathway did not exhibit appreciable changes in its gene expression in response to the AlP treatment. In the MVA pathway, the ACAT was upregulated, which produced isopentyl diphosphate (Xieet al.2016), whereas both HMGR and GGPS were downregulated. It showed upregulation of FPS but no change in ADS expression in our study. FPS overexpression increased artemisinin production (Han et al. 2006; Banyai et al. 2010), confirming the role of FPS and substrate availability in the regulation of artemisinin biosynthesis (Ikram and Simonsen 2017; Simonsen et al. 2013).

The CYP71AV1 gene, which is required for the two oxidation steps of artemisinin biosynthesis, amorpha4,11-diene, and artemisinic alcohol, was upregulated (Teoh et al. 2006; Ikram and Simonsen 2017). Furthermore, our study recorded the upregulation of ALDH. ALDH1 is used in yeast and plants for the metabolic engineering of artemisinin precursors (Xie et al.2016). Paddon et al. (2013) successfully produced artemisinic acid on an industrial scale by incorporating ALDH1 into engineered yeast strains. Zhang et al. (2011) found that ALDH1 was overexpressed in tobacco plants. Transgenic plants could synthesize dihydroartemisinic alcohol even if neither artemisinic nor dihydroartemisinic acid was found (Zhanget al.2011). Additionally, according to previous transcriptional investigations, ALDH1 expression in A. annua is directly linked to the generation of artemisinin (Dilshad et al. 2015; Xiang et al. 2015), showing that it participates in the biosynthetic process.

Despite the fact that HMGR was downregulated, artemisinin production increased in the presence of Alp, in contrast to previous studies that showed HMGR upregulation leads to an increase in artemisinin production. Despite the downregulation of HMGR, artemisinin levels increased due to the upregulation of most artemisinin-related genes (ACAT, FPS, CYP71AV1, and ALDH1).

On the contrary, it has been postulated that TFs play significant roles in the transcriptional control of gene expression through their binding to DNA regulatory elements (Hou et al. 2019; Mathelier et al. 2016). In addition, there is strong evidence that TFs have a role in phosphate homeostasis (Castrillo et al. 2013; Nilsson et al. 2007; Secco et al. 2012; Wang et al. 2009). Our study identified 31 TF families, such as GATA, NIN-like, C2H2, GRF, MYB-related, ERF, and HSF, and showed differential expression under the AlP (Fig. 7A, B). These transcription factors, under our treatment, play significant roles in alleviating P starvation, increasing phosphate acquisition, ROS homeostasis, root system establishment, and artemisinin biosynthesis regulation. Many TFs, including one MYB gene (CTI12_AA463820) and one MYB-related gene (CTI12_AA434110), showed upregulated expression after exposure to the AlP. In addition, downregulation was revealed in two MYB-related genes (CTI12_AA271060 and CTI12_AA271110) and two MYB genes (CTI12_AA572640 and CTI12_AA340370), which may be crucial for enhancing phosphate uptake, activating responses to P shortages, and root architecture. An earlier investigation in rice found that the OsMYB2P-1 gene controls downstream genes to repress or activate responses to P shortage and influence root architecture (Dai et al. 2012). Additionally, OsMYB4P overexpression might trigger the expression of several Pht genes and boost phosphate uptake (Yang et al. 2014). Additionally, it has been discovered that the MYB TFs, AaMYB1, AaMIXTA1, and AaTAR2 are crucial for increasing trichome initialization and artemisinin accumulation (Matías-Hernández et al. 2017; Shi et al. 2018; Zhou et al. 2020). Our findings also indicated that two WRKY TF members showed downregulation. Our results demonstrate the anticipated functions of these TF families under phosphate deprivation circumstances. AaWRKY1, the first isolated and characterized A. annua transcription factor, regulates artemisinin biosynthesis (Shen et al. 2016). A previous study (Dai et al. 2016) indicated that WRKY74 modifies rice’s susceptibility to phosphate deprivation. C2H2 displayed differential expression in our work, with just one TF upregulated and one TFs downregulated, both of which may be significantly involved in P starvation. TaZAT8, a C2H2-ZFP-type TF gene in wheat, is crucial in mediating wheat tolerance to a lack of P by controlling P uptake, ROS homeostasis, and the development of the root system (Ding et al. 2016). Additionally, our research showed that BHLH was downregulated, which may impact the final accumulation of artemisinin. Previous research has demonstrated the positive regulation of artemisinin production by AaORA and AabHLH1 (Ji et al. 2014; Lu et al. 2013). Five members of the ERF family were downregulated in response to AlP. As a result, ERF may be crucial in coping with P stress and may result in a large accumulation of artemisinin. ADS and CYP71AV1 were both favorably regulated by AaERF1 and AaERF2 simultaneously, which helped plants produce artemisinin and artemisinic acid (Yu et al. 2012). Overall, our results will help to understand how hardly soluble P fertilizer influences the transcriptional regulation of A. annua L. about artemisinin and plant hormone production in these conditions.