Analyses of high spatial resolution datasets identify genes associated with multi-layered secondary cell wall thickening in Pinus bungeana

Abstract

Background and Aims

Secondary cell wall (SCW) thickening is a major cellular developmental stage determining wood structure and properties. Although the molecular regulation of cell wall deposition during tracheary element differentiation has been well established in primary growth systems, less is known about the gene regulatory processes involved in the multi-layered SCW thickening of mature trees.

Methods

Using third-generation [long-read single-molecule real-time (SMRT)] and second-generation [short-read sequencing by synthesis (SBS)] sequencing methods, we established a Pinus bungeana transcriptome resource with comprehensive functional and structural annotation for the first time. Using these approaches, we generated high spatial resolution datasets for the vascular cambium, xylem expansion regions, early SCW thickening, late SCW thickening and mature xylem tissues of 71-year-old Pinus bungeana trees.

Key Results

A total of 79 390 non-redundant transcripts, 31 808 long non-coding RNAs and 5147 transcription factors were annotated and quantified in different xylem tissues at all growth and differentiation stages. Furthermore, using this high spatial resolution dataset, we established a comprehensive transcriptomic profile and found that members of the NAC, WRKY, SUS, CESA and LAC gene families are major players in early SCW formation in tracheids, whereas members of the MYB and LBD transcription factor families are highly expressed during late SCW thickening.

Conclusions

Our results provide new molecular insights into the regulation of multi-layered SCW thickening in conifers. The high spatial resolution datasets provided can serve as important gene resources for improving softwoods.

INTRODUCTION

Wood is the most abundant plant biomass, an important carbon reservoir, indispensable to modern human existence. It is a natural renewable resource widely used in papermaking, construction, furniture, tools, biofuels and wood-based chemicals (Chen et al., 2020; Wang et al., 2023). Depending on its distinctive anatomical features, wood can be divided into softwoods, consisting mostly of conifers, or hardwoods, consisting mostly of angiosperm trees (Donaldson et al., 2017). To improve the processing and utilization of timber and to breed trees with tailored structure and desired properties, it is essential to understand the complex cellular processes by which is formed.

Wood formation is an ordered developmental process involving proliferation of vascular cambium, the differentiation of secondary tissues, cell expansion, secondary cell wall (SCW) thickening and programmed cell death (Mélanie et al., 2014). The cell types of wood and its volume fractions vary widely in different tree species. Hardwood consists mainly of vessels, fibres and parenchyma. In contrast, softwood features a simpler microstructure, and tracheids are the major cell types (≥90 %) of growth (Donaldson, 2019; Chen et al., 2020). Variations in the lumen diameter and cell wall thickness affect the function and properties of wood (Prendin et al., 2017). Tracheids with larger lumen diameters and thinner walls are largely responsible for water transport, whereas those with smaller lumen diameters and thicker walls have higher strength and density (Cuny et al., 2015; Beeckman, 2016). The typical structure of xylem SCW, comprising S₁, S₂ and S₃ layers, and the chemical organization, involving cellulose, hemicelluloses and lignin (Yin et al., 2011; Wang et al., 2023), are mainly based on tracheids of softwood. Thus, SCW thickening of xylem cells is a key developmental stage that determines wood properties and behaviour (Donaldson, 2019; Zhu and Li, 2021).

Molecular characterization of SCW thickening is fundamental to understanding how wall-thickened structures are formed and to suggest possible strategies by which the properties of wood might be improved by genetic modification of trees. Recently, numerous studies have revealed the molecular mechanisms underlying all the stages of SCW thickening in the xylem tissue of angiosperm trees and found that it is a highly regulated process that is strictly controlled by multi-layered transcriptional networks (Fromm, 2013; Eduardo et al., 2019; Zhu and Li, 2021). The top layer of regulatory networks is composed of wood-associated NAC domain (WND) transcription factors (TFs), which function as molecular switches that can activate or repress third-layer TFs and SCW biosynthesis genes by controlling the second layer of regulatory networks composed of MYB TFs (Akiyoshi et al., 2021; Ohtani et al., 2021). In addition, WNDs can directly regulate third-layer TFs and their downstream genes (Zhong et al., 2010; Ye and Zhong, 2015). However, owing to the large genome sizes and complex gene annotation of conifers (Nystedt et al., 2013; Niu et al., 2022), the molecular mechanisms that regulate SCW thickening in conifers have not been studied extensively. This is unfortunate because conifers are the largest timber tree species used for producing biofuels, chemicals and construction materials and play important economic and ecological roles (Terrett et al., 2019).

Previous research on Pinus abies involving high spatial resolution RNA-sequencing (RNA-Seq) (Giacomello et al., 2019) revealed that homologous genes of CesA stimulate SCW formation (Jokipii-Lukkari et al., 2017). In Pinus taeda, MYB1/8 was reported specifically to regulate SCW deposition (Bomal et al., 2008). Small RNA and degradome sequencing revealed that the targets of microRNA (novel_16) encoded MYB domain proteins and regulated SCW thickening in Pinus massoniana (Shen et al., 2021).

SCW thickening is a complex development process that involves the formation of S₁, S₂ and S₃ layers, multi-layered structures that differ in the orientation of cellulose microfibrils and chemical composition (Richter et al., 2011; Terrett et al., 2019). The S₁ and S₃ layers have a high microfibril angle, which prevents outwards swelling of the cell wall and shrinkage of the cell lumen, whereas the S₂ layer has a low microfibril angle and is not highly birefringent, as viewed through a polarizing microscope (Stamm, 1964; Reis and Vian, 2004; Peng et al., 2020; Kirui et al., 2022). In addition, the degree of lignification of these multi-layered structures increases sequentially across S₂, S₁ and S₃ layers (Donaldson et al., 2017). Given this complexity, diverse molecular processes are expected to be involved in the development of these multi-layered structures.

In this study, we selected three 71-year-old Pinus bungeana trees growing in natural field conditions, which is a species that has good wood development and utilization value (Jiang et al., 2010; Qi et al., 2021; Yang et al., 2022). However, to date, there are no genetic and functional genomic data available. Our goal was to identify the genes associated with multi-layered SCW thickening. A reference transcriptome resource with comprehensive functional and structural annotation of P. bungeana was produced, for the first time, by using second- and third-generation sequencing technologies. Based on this new resource, we generated high spatial resolution RNA sequencing data spanning the phloem (P), vascular cambium (CZ), zone of xylem expansion (EX), early SCWs (ES; S₁ layer formation), late SCWs (LS; S₂ and S₃ layer formation), mature xylem (MX) and the latewood of the previous year (LW). By combining these datasets with the anatomical data obtained from the corresponding tissue, we identified 21 co-expression modules and a thickness module (tM) using weighted correlation network analysis (WGCNA). Further comparative analysis of the high spatial resolution transcriptomic profile identified key transcriptional regulatory genes involved in ES and LS formation. The novel reference transcriptome resource presented here can serve as an important gene resource for further RNA-Seq analysis of P. bungeana and might provide phylogenetic insights into other coniferous species. The newly developed higher spatial resolution dataset will help to facilitate our understanding of the molecular mechanisms underlying multi-layered SCW thickening in P. bungeana.

MATERIALS AND METHODS

Plant materials and growth conditions

Samples of 71-year-old P. bungeana were obtained from a primary clonal seed orchard at the Chinese Academy of Forestry, Beijing, China (40.01°N, 116.25°E, 59 m above sea level). The cone, bark, phloem, cambium, xylem, needle and strobili tissues of P. bungeana were sampled from three independent trees on 9 June 2022, a time when new earlywood formation was ongoing and cambium cell activity was high. These samples were collected and stored at −80 °C, and they were used for PacBio long-read RNA sequencing.

Three wooden blocks containing bark, phloem, cambium and xylem tissues were collected at 1.5 m from the bottom of the tree trunk, a region without branches and resin canals. Each block was then cut into small pieces measuring 8 mm × 8 mm × 5 mm (tangential × longitudinal × radial); specific developmental zones from the P, CZ, EX, ES, LS, MX and LW were obtained by tangential cryosectioning. The samples for use in high spatial resolution RNA-Seq and analysis of anatomical features were stored at −80 and 4 °C, respectively. All samples were collected at noon.

Long-read single-molecule real-time sequencing and data processing

Total RNA was extracted using QIAzol Lysis Reagent (Qiagen). The RNA from the cone, bark, phloem, cambium, xylem, needle and strobili tissues of P. bungeana was mixed together in equal amounts. After synthesizing and constructing a complementary DNA (cDNA) library for PacBio long-read RNA-Seq, the mRNA was enriched using oligo (dT) magnetic beads. The mixed RNA sample was used to synthesize cDNA using the HiScript III 1st Strand cDNA Synthesis Kit. PCR amplification was performed using the KAPA HiFi HotStart PCR Kit, and a large PCR was selected to construct the SMRTbell library. Long-read single-molecule real-time (SMRT) sequencing was performed using a PacBio Sequel System (Annoroad, Beijing, China).

To obtain accurate and authentic sequencing information, raw reads were processed using the software SMRTLink v.5.0.1 (Wang et al., 2016). Initially, circular consensus sequence (CCS) data were generated from a subreads file in the conditions of full passes ≥ 3 and a Q-value of >0.9. CCS can be divided into full-length non-chimeras (FLNC), full-length chimeras (FLC), non-full-length (NFL) and short reads (SR). Polished consensus was obtained by iterative clustering and error correction and Quiver software polishing (error rate <1 %), and clean reads obtained from the same samples were used to correct low-quality isoforms by LoRDEC v.0.8 (Salmela and Rivals, 2014). Finally, after corrections and filtering, full-length transcripts were generated from these polished consensuses (Fig. 1).

Fig. 1.

Overview of Pinus bungeana long-read transcriptome sequencing. (A) Samples used for PacBio long-read RNA sequencing. Number 1–6 indicate needles (1), block [phloem, cambium and xylem] (2), strobili (3), bark (4), cone (5) and branch (6), cone (7) and branch (8) tissues of P. bungeana. (B) Length distribution of circular consensus sequencing (CCS) reads. (C) Data processing: 1 283 714 CCS reads were generated from subreads, then classified to full-length non-chimera (FLNC), full-length chimera (FLC), non-full-length (NFL) and short reads (SR). After correction, polishing and clustering, 81 239 full-length transcripts were generated. (D) Classification of CCS reads. (E) Unigene length distribution.

Functional and structural annotation of the full-length transcripts

To generate authentic annotation information, the software ANGEL was used to determine the open reading frames (ORFs), and ten public databases [NR, NT, Swiss-Prot, KOG, eggNOG, gene ontology (GO), TmHMM, Pfam, KEGG and SignalP] were used to annotate the gene functional annotation information. Additionally, TFs in P. bungeana were predicted using the PlantTFDB database. Long non-coding RNA (lncRNA) analysis was performed using the Coding Potential Calculator (CPC), Coding-NonCoding Identifying Tool (CNCI) and Coding Potential Assessing Tool (CPAT) databases. Simple sequence repeats (SSRs) were explored using ssr.pl software.

Sectioning and analysis of anatomical features

The samples used for investigating anatomical traits were first fixed for 12 h in formaldehyde (3.7 %)–acetic acid (5 %)–ethanol (50 %) fixative solution at 4 °C, then transferred to pure paraffin. Developing early wood was microtomed into 10-μm-thick sections using a HistoCore AUTOCUT (LECIA RM 2265). The cross-sections were then stained with Astra Blue and Basic Fuchsin solutions as previously described (Kraus et al., 1998), and the images were examined under a light microscope (OLYMPUS DP71).

The samples used for scanning electron microscopy observation were separated into smaller pieces and fixed in glutaraldehyde fixative solution (2.5 %) at 4 °C for 12 h. The cells were then transferred into PBS (0.1 mol L⁻¹), osmic acid (1 %) or different ethanol concentrations. After drying (Leica EM CPD300, Germany) and sputtering with gold–palladium, the samples were examined using a scanning electron microscope (ZEISS Gemini SEM 300, Germany) operating at 2.0 kV.

Determination of the thickness of the radial double wall and diameter of the radial lumen

The wood blocks were sampled from naturally growing trees. Three biological replicates were used for each experiment. Tissue fixation and embedding were performed as previously described. Sections of developing secondary xylem were cut into 10 μm thick using a HistoCore AUTOCUT (LECIA RM 2265) and observed by light microscopy (OLYMPUS DP71). Statistical software was used to quantify the thickness of the radial double wall and diameter of the radial lumen.

High spatial resolution transcriptome sequencing and analysis

The high spatial resolution sections (tangential, 25 µm thick) were produced using a cryo-ultramicrotome (Leica CM3050 S) and used for high spatial resolution RNA-Seq. Simultaneously, the cross-sections were cut and stained to ascertain the positions of the tissues at different developmental stages. Total RNA was extracted from P, CZ, EX, EX, LS, MX and LW tissue sections, and three biological replicates were performed for each tissue (Supplementary Data Table S1). The RNA quality and integrity were assessed using Aligent 2100, an RNA library was constructed, and raw data were analysed as previously described (Du et al., 2022). Finally, the libraries were sequenced on an Illumina HiSeq 2500 PE150 platform (Annoroad, Beijing, China).

To analyse the high spatial resolution RNA sequencing data further, adapters and lower QC reads were removed to obtain higher QC reads (Clean Reads) using fastp v.0.18.0 software. After alignment in the ribosome RNA database, clean reads that did not match were mapped to the reference transcriptome of P. bungeana using HISAT2 v.2.1.0 (Kim et al., 2015). The mapped reads of each sample were assembled by using StringTie v.1.3.1 (Pertea et al., 2016). Only perfectly mapped reads were used for in-depth analysis. The fragment per kilobase of transcript per million mapped reads (FPKM) values were calculated using the software RSEM v.1.3.1 (Li and Dewey, 2011).

The R package ‘WGCNA’ was used to identify correlations between high spatial resolution gene expression profiles and the traits of tracheids and to determine biologically significant modules. Genes with high connectivity might have important functions. The networks visualization was performed using Cytoscape v.3.3.0 software.

Phylogenetic tree building

All the NAC protein sequences in Populus tomentosa were downloaded from the JGI Phytozome v.13 (https://phytozome-next.jgi.doe.gov/). Phylogenic trees were constructed by MEGA v.7.0 software based on the neighbour-joining method. In the Phylogeny Test menu, we selected ‘Bootstrap method’ with a value of 1000. After selecting the default parameters from the dropdown menus, we completed construction of the phylogenetic tree. The phylogenetic tree was visualized using iTOL v.6 (https://itol.embl.de/itol.cgi).

Identification of differentially expressed genes and RT-qPCR analysis

We used feature Counts software and the R package ‘edgeR’ to identify the differentially expressed genes (DEGs). Genes were identified as DEGs when the fold-change ratio was higher than two and the adjusted P-value was <0.05 (Smyth, 2010; Liao et al., 2014). DEGs were clustered by the software Short Time-series Expression Miner (STEM) v.1.3.13 (Ernst and Bar-Joseph, 2006), setting the clustering algorithm (STEM clustering method), the maximum number of model profiles (30), fold-change ratio (2) and P-value (0.05). The DEGs were mapped to a GO annotation database of P. bungeana with a threshold of false discovery rate (FDR) < 0.05 (Ye et al., 2018). Real-time qPCR was performed as previously described (Guo et al., 2021). Each analysis was based on three replicates, and a relative quantification method was used. The specific primers used in this study are listed in the Supplementary Data (Table S2).

RESULTS

Overview of transcriptome sequencing and data processing in P. bungeana

Full-length transcriptomes of P. bungeana from different tissues (cone, bark, phloem, cambium, xylem, needle and strobili) were obtained using PacBio long-read RNA sequencing (Rhoads and Au, 2015;Boldogkői et al., 2019), generating ~178.56 Gb of raw sequencing data (Fig. 1A). After quality control, we obtained 1 283 714 polymerase reads and 112 574 744 subreads with N50 (The length of the contig for which 50% of all bases in the final contigs was longer than, or equal to, this value) values of 203 159 bp and 1696 bp, respectively (Supplementary Data Table S3). Based on the conditions that full passes were ≥ 3 and that Q-value was >0.9, 1 283 714 CCS reads were identified, with a mean read lengths of 1637 bp (Fig. 1B, C; Supplementary Data Fig. S1). The CCS reads consisted mainly of FLNC, FLC, NFL, and SR; 1 264 567 sequences were identified as FLNC reads (Fig. 1D). Based on iterative clustering and error correction and Quiver algorithms for clustering and polishing, 102 779 high-quality polished full-length consensus isoforms were obtained with an accuracy of >0.99 and mean lengths of 1692 bp (Fig. 1E; Supplementary Data Table S3). Short-read RNA-Seq generated 0.52–0.55 billion clean reads for each xylem tissue (Supplementary Data Table S4) to correct the polished consensus isoforms. Then, merging the redundant sequences of high-quality isoforms using the cd-hit software, 81 239 non-redundant transcripts were obtained and served as a reference transcriptome for subsequent analysis (Fig. 1C).

Functional and structural annotation of the reference transcriptome

To obtain a high-quality and complete functional annotation from the reference transcripts of P. bungeana, all the 81 239 non-redundant transcripts were aligned and annotated with NR (NCBI non-redundant protein sequences), NT (NCBI nucleotide sequence), Swiss-Prot (Protein knowledgebas), KOG (euKaryotic orthologous groups), eggNOG (evolutionary genealogy of genes: Non-supervised Orthologous Groups), GO, TmHMM (Prediction of transmembrane helices in proteins), Pfam (Protein families database of alignments and hidden Markov models), KEGG (Kyoto Encyclopedia of Genes and Genomes) and SignalP (Signal peptide and cleavage sites) databases. As a result, 97.72 % (79 390 transcripts) of the non-redundant transcripts obtained the functional annotation successfully (E-value < 1 × 10⁻⁵) (Fig. 2A; Supplementary Data Table S5). A total of 95.44 % (75 768 transcripts) of the annotated transcripts had the best hits for Pinaceae, whereas 48.64 % (38 613 transcripts) were unknown functional proteins (Supplementary Data Figs S2 and S3).

Fig. 2.

Functional annotation for the Pinus bungeana transcriptome. (A) Upset diagram of distinct intersections from ten functional annotation databases. Bars represent the numbers of non-redundant transcripts with significant hits from these databases. (B) Gene ontology (GO) classification of the non-redundant transcripts. The parallel axis indicates the number of transcripts.

According to the GO annotation, ‘cellular process’, ‘metabolic process’ and ‘single-organism process’ took the top three spots in the biological process category, ‘cell part’, ‘organelle’ and ‘organelle part’ were the most enriched GO terms of the cellular component category, and ‘binding’, ‘catalytic’ and ‘transporter’ had the largest number of transcripts in the molecular function category (Fig. 2B). In the KOG database, 62.65 % (49 793 transcripts) of transcripts were annotated and divided into 25 categories. The ‘general function prediction only’ (9740 transcripts), ‘posttranslational modification, protein turnover, chaperones’ (7087 transcripts), ‘signal transduction mechanisms’ (4266 transcripts), ‘carbohydrate transport and metabolism’ (3712 transcripts) and ‘translation, ribosomal structure and biogenesis’ (3232 transcripts) were the top five spots in the functional categories (Supplementary Data Fig. S4). The KEGG classification results showed that 42 pathways were annotated from 31 548 transcripts, and the top three pathways were ‘carbohydrate metabolism’, ‘signal transduction’ and ‘translation’ (Supplementary Data Fig. S5).

Using TransDecoder, 86 844 ORFs were predicted, and the number and length distribution of the complete coding protein sequences were calculated (Fig. 3A). A total of 31 808 putative non-redundant TFs distributed across 513 families were identified (Supplementary Data Fig. S6). The most abundant TF categories were bHLH (3505 transcripts), MYB (3026 transcripts), B3 (1970 transcripts), NAC (1806 transcripts) and WRKY (1570 transcripts).

Fig. 3.

Structural analysis for the Pinus bungeana transcriptome. (A) Length distribution of open reading frames (ORFs). (B) Venn diagram of long non-coding RNA transcripts from Coding Potential Assessing Tool (CPAT), Coding-NonCoding Identifying Tool (CNCI) and Coding Potential Calculator (CPC) analyses. (C) Distribution of simple sequence repeats (SSRs) in P. bungeana-expressed sequence tags (ESTs).

SSRs have a minimum repeat unit size of five (for tri- to hexa-nucleotides) or six (for dinucleotides). Based on our analysis of non-redundant transcripts, we identified 9198 tetra-nucleotide repeats (78.3 %) and 2549 tri-nucleotides repeats (21.7 %) (Fig. 3C). Additionally, the CPC, CNCI and CPAT databases were combined to distinguish lncRNA candidates. Analyses of the CPC, CNCI and CPAT databases revealed 11 718, 15 180 and 9873 lncRNA candidates, respectively. Finally, 5147 lncRNA transcripts were identified (Fig. 3B).

High spatial resolution RNA-Seq analysis for the P. bungeana xylem SCW thickening

To explore dynamic gene expression profiles during SCW thickening, blocks were sampled from the tree trunks of three independent P. bungeana trees. At the same time, according to the anatomical features of transections in the P. bungeana developing xylem, we generated high spatial resolution tangential cryogenic sections spanning the P, CZ, EX, ES, LS, MX and LW tissues (Fig. 4A) and used them for high spatial resolution RNA sequencing (Supplementary Data Table S1). Each xylem tissue sample retained >65 million clean reads for further analysis (Supplementary Data Table S4), and Pearson’s correlation was performed to evaluate the biological replicates. Although high spatial resolution RNA was obtained from orchard-grown P. bungeana trees, the relationship coefficients among the biological replicates were >0.8, except in the LW tissues.

Fig. 4.

High spatial resolution RNA-Seq analysis of the secondary cell wall thickening during Pinus bungeana xylem formation. (A) Representative microscopy images of transections of the P. bungeana xylem at different developmental stages. P, CZ, EX, ES, LS, MX and LW indicate phloem, vascular cambium, xylem expansion, early SCWs (S₁ layer formation), late SCWs (S₂ and S₃ layer formation), mature xylem and latewood of the previous year, respectively. Scale bars = 10 μm. (B) Temporally dynamic clusters of high spatial resolution RNA-Seq data across the different tissues. Grey and black lines indicate the expression dynamics and profile, respectively. The number of genes is shown in each cluster. (C) GO terms identified in clusters W1–W10 of P. bungeana. The depth of red colour represents −log₁₀(P-value).

Based on the expression patterns in which the fold-change was higher than two and the P-value was <0.05, 30 temporally dynamic clusters were identified, and only ten gene clusters (6164 genes, W1–W10 clusters) corresponded to the different developmental stages of SCW thickening (Fig. 4B). The gene expression trends at W5 and W10 were the same in ES and LS tissues, and the genes at W5 and W10 exhibited continuous upregulation and downregulation, respectively. GO analyses were performed to understand the molecular events associated with the expression clusters. The genes in W1–W10 were mainly categorized as being involved in cell differentiation, primary metabolism, growth and development, secondary metabolism, microstructure formation, cell–cell junctions, signal transduction, developmental maturation, programmed cell death and cell death (Fig. 4C). Most of these genes were associated with wood formation, and some genes in W1 and W4 were involved in formation of cell wall microstructure. Notably, genes in W1, W4 and W7 were concentrated in the cell death category, indicating that LW tissues grown in the previous year might remain transcriptionally active.

To determine the key TFs involved in SCW thickening of P. bungeana xylem, a total of 1410 TFs were identified in ten clusters. Important TFs in the multi-layered transcriptional networks of SCW thickening, such as NAC, MYB, WRKY, ERF and C3H, were found in all ten temporally dynamic clusters (Fig. 5A). Members of MYB, NAC and ERF TFs were significantly enriched in W2, W8 and W7, respectively. In addition, 18 PbNACs of the 140 TFs in W7 had the highest enrichment, indicating that NAC was the top layer of the regulatory network of SCW thickening, corresponding to its role in poplar. Hence, we also examined these 18 NAC-domain TFs in W7 and compared them with the 136 PtNACs (Fig. 5B). The results showed that these 18 PbNACs were distributed in four groups (B, I, K and L), which were involved in SCW thickening during formation of xylem in P. bungeana.

Fig. 5.

Transcription factors (TFs) involved in secondary cell wall thickening during formation of Pinus bungeana xylem. (A) Distribution of TF families in temporally dynamic clusters (W1–W10) in P. bungeana. The two heatmaps present TF families enriched in ten clusters and expression patterns of PbNACs in W7, respectively. Important TFs found in all ten of these temporally dynamic clusters are marked in red. (B) Phylogenetic tree of NAC domain TFs from P. bungeana and Populus tomentosa.

Dynamic variations in anatomical traits of the cell wall of tracheids during xylem formation in P. bungeana

The formation of P. bungeana wood follows a sequential cell-developmental process (Zhang et al., 2012; Kumar et al., 2016). Once the final size and shape of the tracheids are achieved, the thickness of the cell wall begins to increase (Dai et al., 2022). To evaluate the wood density and hydraulic features of P. bungeana, we examined the cell anatomical characteristics (double wall thickness and lumen diameter) of the tracheids in the wood using light microscopy and verified the results using scanning electron microscopy (Fig. 6A). The average thickness of the radial double wall increased from the CZ to the LS in the developing xylem, and that of the tangential double wall increased from the EX to the LS (Fig. 6B). In addition, the average diameter of the tangential lumen was expanded continuously in the developing xylem, whereas that of the radial lumen remained almost the same (Fig. 6B). The average thickness of the double wall increased rapidly from the ES, where the S₁ and S₂ layers were formed. These results suggest that cell wall thickness is determined mostly by the S₁ and S₂ layers in P. bungeana tracheids.

Fig. 6.

Anatomical analysis of the cell walls of tracheids during formation of xylem in Pinus bungeana. (A) Representative microscopy images of transections of the xylem at different developmental stages. Scale bars = 10 μm for light microscopy and 1 μm for scanning electron micrographs. (B) The thickness of radial/tangential cell wall and diameter of radial/tangential lumen in tracheids. Small open circles represent individual values. Means with the same letter are not significantly different at P < 0.05 according to the least significant difference (LSD) test. CZ, EX, ES, LS, MX and LW indicate vascular cambium, xylem expansion, early SCWs (S₁ layer formation), late SCWs (S₂ and S₃ layer formation), mature xylem and latewood of the previous year, respectively.

Spatiotemporal co-expression network of highly expressed genes involved in SCW thickening of P. bungeana

To evaluate the high spatial resolution dataset containing genes and traits and to identify the hub genes involved in SCW thickening, a total of 17 581 genes (FPKM ≥ 0.5, with annotation information) from 15 independent samples (three replicates, five xylem tissues covering CZ, EX, ES, LS and MX; Supplementary Data Table S1; Fig. S7) were analysed using WGCNA. All samples were clustered according to their expression patterns, and 21 modules were obtained using this clustering relationship (Fig. 7A). The similarity of genes in each module is presented at the network topology level (Fig. 7B). The thickness of the radial/tangential double wall and the diameter of the radial/tangential lumen were determined to analyse the relationships in the trait modules. We found that wM2, wM13 and wM16 were positively correlated with the cell wall thickness. In contrast, wM8, wM11, wM12 and wM19 were negatively correlated (Fig. 7C). The tM13 module was the most highly correlated with cell wall thickness, and its hub genes had high gene significance (|GS| ≥ 0.3) with anatomical characteristics and high module membership (|MM| ≥ 0.8) with module eigengenes (Fig. 7D; Supplementary Data Table S6). We constructed a gene regulatory network using the top 60 genes in the tM13 module (Fig. 7E). Significantly, some hub genes (members of the NAC, MYB, WRKY, ARF, SUS, LAC and CESA gene families) involved in SCW thickening were strongly co-expressed. These results indicate that several genes associated with SCW thickening might be regulated by TFs in the secondary xylem during early wood formation.

Fig. 7.

Weighted correlation network analysis of expression patterns of differentially expressed genes (DEGs) at different developmental stages of Pinus bungeana xylem. (A) Hierarchical cluster dendrogram of genes in co-expression modules. Different colours indicate different modules. (B) Network heatmap plot of genes in the co-expression modules. The progressively saturated red colours represent higher connectivity. (C) Correlations between module eigengenes and anatomical analysis for the thickness and diameter. The numbers indicate correlations and P-values for the module–trait associations based on Pearson’s correlation coefficients. Red and blue represent positive and negative correlation, respectively. (D) Correlation scatter plots of module membership vs. gene significance for the radial cell wall with tM13 (red), tangential cell walls with tM13 (blue) and tangential cell wall with tM1 (grey). (E) The gene regulatory network of the tM13 module genes, which was positively correlated with radial/tangential wall thickness. The hub gene families in the tM13 module are highlighted in purple. RW, TW, RL and TL indicate the radial cell wall, tangential cell wall, radial lumen and tangential lumen, respectively.

Identification of the core TFs that regulate formation of the multi-layered structure during the SCW thickening of tracheids

To explore the potential regulators involved in the formation of multi-layered structures (S₁, S₂ and S₃ layers) during tracheid SCW thickening, we compared the whole-transcriptome profiles of developing xylem tissues among EX, ES and LS. Three biological replicates were performed for each tissue, and these nine RNA libraries generated ≥0.69 billion raw reads; after filtering, ~0.68 billion clean reads were obtained (Supplementary Data Table S4). Sequences that could not be aligned to the reference transcriptome of P. bungeana were removed, and only those that were perfectly mapped were analysed further. When the fold-change ratio was higher than two and the adjusted P-value was <0.05, we successfully identified 2356 DEGs between the EX and ES (Supplementary Data Table S7) and 2157 DEGs in the ES and LS (Supplementary Data Table S8).

In comparison to EX, 1169 upregulated genes and 1187 downregulated genes were identified in ES (Fig. 8A). We then performed GO analysis to calculate the molecular processes of these DEGs. These results indicated that the upregulated DEGs were enriched in biological processes (GO: 0008150, P = 9.1 × 10^–1), cellular processes (GO: 0009987, P = 2.1 × 10^–2), metabolic processes (GO: 0008152, P = 1.9 × 10^–1), organic substance metabolic processes (GO: 0071704, P = 5.6 × 10^–3), single-organism processes (GO: 0044699, P = 2.0 × 10^–1), cellular metabolic processes (GO: 0044237, P = 1.0 × 10^–1) and primary metabolic processes (GO: 0044238, P = 1.2 × 10^–2). In comparison to the ES group, 998 upregulated genes and 1159 downregulated genes were identified in the LS group (Fig. 8B). The GO analysis in biological processes indicated that the upregulated DEGs were enriched in cellular processes (GO: 0009987, P = 9.0 × 10^–1), metabolic processes (GO: 0008152, P = 7.1 × 10^–1), response to stimulus (GO: 0050896, P = 1.1 × 10^–3), biological regulation (GO: 0065007, P = 4.8 × 10^–1), regulation of biological processes (GO: 0050789, P = 7.9 × 10^–1), developmental processes (GO: 0032502, P = 3.3 × 10^–1) and growth (GO: 0040007, P = 1.5 × 10^–1).

Fig. 8.

Identification of the hub genes associated with the formation of S₁ during Pinus bungeana SCW thickening of tracheids. (A) Volcano plot visualizing differentially expressed genes (DEGs) in EX and ES. (B) Volcano plot visualizing DEGs in ES and LS. (C) Identification of common genes between the tM13 module and DEGs (EX vs. ES) by overlapping them. The five hub genes in the tM13 module were also DEGs in EX and ES. (D) Identification of common genes between the tM13 module and DEGs (ES vs. LS) by overlapping them. The five hub genes in the tM13 module were also DEGs in ES and LS. EX, ES and LS indicate xylem expansion, early SCWs (S₁ layer formation) and late SCWs (S₂ and S₃ layer formation), respectively.

To identify the core genes associated with formation of the multi-layered structure, we overlapped genes in the tM13 module and DEGs between EX and ES using a Venn diagram and found that NAC45, WRKY1, LAC9, SUS2 and CESA8 mainly contributed to ES formation (Fig. 8C; Supplementary Data Fig. S8). Interestingly, NAC45 and WRKY1 were also found among the DEGs between ES and LS, and NAC73, MYB8 and LBD1 appeared in these DEGs and the tM13 module (Fig. 8D; Supplementary Data Fig. S8). These results show that the NAC-domain TFs are key players in the P. bungeana SCW thickening of the tracheids.

DISCUSSION

Here, we present full-length and high spatial resolution analyses of the transcriptome, covering comprehensive growth and differentiation stages of the developing earlywood of P. bungeana. Based on the association analysis between anatomical traits and transcriptomes at different xylem developmental stages, we found that 17 581 genes were clustered into 21 modules, and the gene expression pattern in the tM13 module was associated with SCW thickening in P. bungeana (Fig. 7).

In our study, we used the second- and third-generation joint sequencing techniques to generate a P. bungeana reference transcriptome dataset for the first time, which provides transcriptome resources for investigating the anatomical traits associated with SCW thickening in P. bungeana xylem. Merging anatomical traits with the aid of RNA-Seq methods, such as high spatial resolution transcriptomic analyses (Sundell et al., 2017; Liu et al., 2020) and WGCNA, has provided an excellent methodology for establishing the entire high spatial resolution database in the sequential developmental process of multi-layered structures (Wang et al., 2022). Besides the 21 clusters identified above, the associated analysis of other anatomical traits obtained from the same tissue could reveal the modes of accumulation in the new gene regulatory networks. Thus, gene expression information is now available at a higher resolution for further trait–gene association analysis and provides a richer transcriptome resource for further studies of gene regulatory networks during earlywood formation in P. bungeana.

Over the past 20 years, wood formation studies on conifer trees have largely been limited to genomic information and growth cycles, which makes it difficult effectively to identify genes that control a trait, because they might be expressed in juvenile wood but not in mature wood. Because the high spatial resolution RNA-Seq method covers the entire growth and differentiation stages of secondary xylem formation (Supplementary Data Table S9), it can be used independently to determine new regulatory networks in mature trees growing in natural conditions (Jokipii-Lukkari et al., 2017) or in combination with other trait data. Using the high spatial resolution transcriptome dataset, we constructed a comprehensive gene expression profile across the vascular cambium and developing xylem during the earlywood growth of P. bungeana, yielding novel insights, including significant differences in gene expression between EX and SCW tissues (Fig. 8; Supplementary Data Tables S7 and S8). In addition, many TFs, such as NAC, MYB, WRKY, LBD and LAC, were associated with SCW thickening (Fig. 7). High spatial resolution RNA-Seq datasets were obtained from a tangential cryogenic section (cryosection) series covering the whole xylem development zone, which captured nearly real spatial information. Integrating spatial transcriptomic data with trait data (Rao et al., 2021) provides a valuable transcriptome resource for exploring the gene regulation networks in earlywood formation in P. bungeana.

Using the high spatial resolution RNA-Seq datasets, we confirmed the previously known TFs and found new gene regulation networks involved in the consecutive developmental process of SCW thickening (Supplementary Data Table S10). Previous studies in Populus have demonstrated that WNDs can regulate the promoters of MYB3 and MYB20 (Zhong et al., 2011); MYB21 and MYB20 (the homologous genes of AtMYB46 and AtMYB83) were also activated by WNDs, which regulated the cell wall biosynthesis (McCarthy et al., 2010; Zhong et al., 2013). Moreover, PdMYB199 (Tang et al., 2020), PttMYB21a (Karpinska et al., 2004), PdMYB221 (Tang et al., 2015), PtoMYB156 (Yang et al., 2017) and PtrMYB189 (Jiao et al., 2019) have been identified in Populus as negative factors in SCW thickening. In this study, we found that PbNAC45 and PbWRKY1 regulated genes involved in the entire developmental stage of SCW thickening. PbNAC45 also activated the expression of PbCESA8, PbSUS2 and PbLAC9 to regulate ES formation, whereas PbNAC73, PbMYB8 and PbLBD1 contributed specifically to LS formation (Fig. 8). These findings provide theoretical support for improving the wood structure and properties, especially for the multi-layered structures.

In summary, we initially established a high-quality reference transcriptome resource with a comprehensive functional and structural annotation of P. bungeana and combined it with high spatial resolution RNA-Seq datasets spanning comprehensive growth and differentiation stages in developing earlywood of P. bungeana. We adopted two approaches to analyse the transcriptome data: (1) association analysis of genes with anatomical traits (WGCNA); and (2) comparative analysis of DEGs at different developmental stages. Using these approaches, we found that PbNAC45 and PbWRKY1 were involved in the entire developmental stage of SCW thickening and also regulated the individual stages of the consecutive developmental processes of multi-layered structures. Furthermore, we identified specific regulatory genes involved in ES (PbCESA8, PbSUS2 and PbLAC9) and LS formation (PbNAC73, PbMYB8 and PbLBD1). These results regarding the molecular regulation of complex cell development processes, such as xylem formation and the multi-layered SCW thickening, in P. bungeana will help to improve the properties and behaviour of wood.

Conclusion

The stem differentiation system realistically reflects the entire process of xylem SCW thickening in mature trees, particularly in the formation of a multi-layered structure. In our study, we explored an excellent transcriptome resource that can serve as a reference transcriptome for future RNA-Seq experiments in P. bungeana and provide important genetic information for the transcriptome analysis of different xylem tissues. Further analysis of a high spatial resolution dataset produced a useful genetic resource to screen the genes associated with multi-layered SCW thickening in P. bungeana. Using these transcriptome resources, we identified new genes involved in wood formation that provided new insights into the different developmental stages of xylem formation in P. bungeana. In addition, the use of this analytical strategy in other conifer species will significantly aid in comprehending the molecular regulation of multi-layered SCW thickening in softwoods.

SUPPLEMENTARY DATA

Supplementary data are available at Annals of Botany online and consist of the following.

Figure S1: quality of circular consensus sequencing (CCS) reads. Figure S2: phylogram of the 12 Pinopsida, 4 Magnoliopsida and other species. Figure S3: summary of BLASTx matches to the annotated transcripts. Number of best BLASTx matches of transcripts grouped by genus. Figure S4: EuKaryotic orthologous groups (KOG) functional classification of Pinus bungeana unigenes. Figure S5: Kyoto Encyclopedia of Genes and Genomes (KEGG) classification of Pinus bungeana unigenes. Figure S6: distribution of transcription factor families. Figure S7: global transcriptome analysis in different developmental stages of Pinus bungeana xylem. (A) Representative transverse section stained with Astra Blue and Basic Fuchsin from the stem of tree 1. (B) The number of transcription factor (TF) genes and non-TF protein-coding (mRNA) genes expressed in different developmental stages. (C) Proportions of expressed genes with low levels (FPKM < 2), medium levels (2 ≤ FPKM < 10) and high levels (FPKM ≥ 10) in different developmental stages. (D) Heatmaps of co-expressed upregulated (red) and downregulated (green) differentially expressed genes (DEGs) (FPKM > 0.5, with annotation information) involved in earlywood formation. Three replicates and six xylem tissues covering the CZ, EX, ES, LS, MX and LW. Figure S8: validation of hub genes in the transcriptional level. qRT-PCR (left vertical axis) and RNA-Seq (right vertical axis). Error bars represent the s.e.m. of three independent experiments. CZ, EX, ES, LS and MX indicate the vascular cambium, xylem expansion, early SCWs (S₁ layer formation), late SCWs (S₂ and S₃ layer formation) and mature xylem, respectively. Table S1: cryosections in different tissue types, and information of pooling for RNA-Seq analysis. Table S2: oligos used as primers in the experiment. Table S3: summary of the transcriptome data for Pinus bungeana using PacBio Iso-Seq. Table S4: summary of reads sequenced from Pinus bungeana stem using Illumina RNA-Seq. Table S5: summary of annotations on unigene from Pinus bungeana against public databases. Table S6: hub genes of the tM13 module, which was positively correlated with cell wall thickness. Table S7: 2356 differentially expressed genes (DEGs) were identified between EX and ES. Table S8: 2157 differentially expressed genes (DEGs) were identified between ES and LS. Table S9: the expression levels of all Pinus bungeana genes in various tissues. Table S10: the cell wall-related genes and their expression in various samples. Table S11: summary of abbreviations.

FUNDING

This work was supported by National Natural Science Foundation of China (grant number 32071691) and the Department of Science and Technology, National Forestry and Grassland Administration (2020132601).

AUTHOR CONTRIBUTIONS

Y.G., J.G. and Y.Y. designed the research; Y.G., L.C.J., J.W., L.M., Y.L. and Y.Z. performed the research and discussion; Y.G. analysed the data; Y.G. and Y.Y. wrote the article.

DATA AVAILABILITY

The RNA-Seq data were available at GSA BioProject with accession number PRJCA016348 (https://ngdc.cncb.ac.cn/gsub/submit/bioproject/subPRO024298).

ABBREVIATIONS

All the abbreviations are listed in the Supplementary Data (Table S11).

CONFLICT OF INTEREST

The authors declare no conflict of interest.