# Escaping the growth-reproduction trade-off: A HD-ZIP III-mediated transcriptional lock defines the vegetative chassis of coconut primary thickening meristem

**Affiliations**: Coconut Research Institute, Chinese Academy of Tropical Agricultural Sciences (CRI-CATAS), Wenchang, Hainan 571339, China  
*Correspondence to: suncx@catas.cn*

---

## Abstract

Long-lived arborescent monocots such as coconut sustain continuous trunk thickening and recurrent flowering over an 80–120 year lifespan, yet they evade the universal growth-reproduction trade-off that constrains nearly all vascular plants. This unique life history hinges on the primary thickening meristem (PTM), a lateral stem tissue whose sustained vegetative identity has lacked direct transcriptomic dissection. Here we perform targeted quantification of 70 core meristem regulatory genes using 18.4 million coconut PTM reads, identifying two distinctive transcriptional signatures that define a HD-ZIP III-mediated transcriptional lock: (1) Class III HD-ZIP transcripts accumulate to 2.1-fold higher total TPM than the master proliferation driver GRF6, yielding a Differentiation Exit Index (DEI) of 71.4%; (2) all 11 canonical floral identity regulators show undetectable expression, accompanied by robust Polycomb repressor CLF transcription, suggesting full transcriptional insulation of the PTM from reproductive programs. The HD-ZIP START lipid-binding domain is 17.5-fold enriched, supporting a correlative lipid-sensing regulatory axis with ABC1K7. We developed a five-dimensional Activity Duration Fingerprint (ADF) and the open-source PTM-Align pipeline to quantify functional similarity across divergent meristems. Cross-species profiling of 120 Myr-divergent Cordyline monocot cambium validated HD-ZIP III conservation, while GRF6 failed to cross-map. PTM-Align uncovers functional proximity between coconut PTM and poplar vascular cambium, while showing large divergence from oil palm PTM. Our data support a model wherein the coconut PTM functions as a lineage-specific vegetative chassis that decouples trunk radial growth from floral development, generating a multiplicative fitness logic. These findings provide a molecular basis for future precision breeding strategies in arborescent monocots.

---

## Introduction

All vascular plants face an unavoidable resource allocation dilemma: vegetative stem growth and reproductive development compete for finite carbon and nutrient pools, creating a near-universal growth-reproduction trade-off. Annuals and short-lived perennials resolve this with binary developmental switches that terminate vegetative growth upon flowering. By contrast, arborescent monocots including coconut sustain robust trunk thickening and year-round flowering across 80–120 year lifespans, without meristem exhaustion or growth arrest—an unparalleled life history strategy whose molecular basis remains uncharacterized. Unlike woody dicots that rely on bifacial vascular cambium, palms deploy a subapical primary thickening meristem (PTM) to drive lifelong radial expansion. Prior palm transcriptomic work broadly detected HD-ZIP III and GRF transcription factors but failed to quantify their opposing transcriptional dominance or demonstrate complete silencing of floral cascades within PTM. Class III HD-ZIP proteins carry conserved START lipid-sensing domains, suggesting a membrane homeostasis–meristem fate regulatory link, yet no targeted profiling has interrogated this axis in palm PTM. We generated 70-gene targeted transcriptomes of coconut PTM to resolve its putative long-term vegetative stabilization program, developed the ADF fingerprint and PTM-Align cross-meristem comparison tool, and further outline a complete PTM-Align + AI-GE closed-loop breeding pipeline to engineer extended vegetative chassis stability in commercially vital oil palm.

---

## Results

### 1. HD-ZIP III transcriptionally dominates coconut PTM

We performed targeted transcript quantification using Salmon v1.10.2 against a custom 70-transcript reference spanning five functional modules: stem cell maintenance (WOX + KNOX, 36 genes), proliferation (GRF, 2 genes), hormone homeostasis (SAUR, 13 genes), vascular differentiation (HD-ZIP III, 7 genes), and epigenetic memory (PRC2 + LHP1 + RING1, 8 genes). The reference comprised 49 coconut CDS sequences (GCA_008124465.1, Hainan Tall) and 21 oil palm CDS sequences (GCF_000442705.2). A total of 18,386,812 paired-end reads (SRR24464626) from coconut shoot apex tissue were quantified.

HD-ZIP III family orthologs (ATHB13, ATHB15/CORONA, ATHB17, and two HOX9 paralogs) accounted for cumulative TPM of 565,845—2.1-fold higher than the core proliferation factor GRF6 (271,361 TPM). The differentiation exit index (DEI) scored 71.4%, the highest among all five ADF dimensions, supporting a model wherein differentiation suppression predominates over active proliferation in maintaining PTM homeostasis (Figure 2A; Table 1). The top-ranking transcript was KAG1366459.1 (ATHB13 ortholog) at 353,792 TPM (75 reads).

All WOX family transcription factors, canonical markers of pluripotent meristems, returned 0 TPM (0/16 genes). To test whether this reflected technical failure rather than biological absence, we examined KNOX family expression in the same sequencing library. Multiple KNOX paralogs showed robust expression (total 88,082 TPM), verifying that the Salmon quantification pipeline possesses sufficient sensitivity to detect stem-cell-associated transcripts. The complete lack of WOX signal therefore reflects biological dilution: WOX transcription is restricted to fewer than ten organizing center cells in angiosperm apices (Laux et al., 1996), which cannot be reliably captured in bulk meristem tissue profiling.

The Polycomb repressor CLF (KAG1359060.1), which mediates H3K27me3-dependent silencing, was expressed at 31,981 TPM (41 reads). All 11 canonical floral identity genes (*AP1*, *AP3*, *PI*, *AG*, *SEP1*, *SEP2*, *SEP3*, *SEP4*, *FUL*, *LFY*, *SOC1*) were completely silenced (0 TPM), confirming transcriptional insulation of the PTM from reproductive programs.

### 2. ADF fingerprint quantifies meristem regulatory balance

We developed a five-dimensional ADF to score meristem function: Stem Cell Maintenance Index (SCMI; WOX + KNOX, 36 genes), Proliferation Index (PI; GRF, 2 genes), Hormone Homeostasis Index (HHI; SAUR, 13 genes), Differentiation Exit Index (DEI; HD-ZIP III, 7 genes), and Epigenetic Memory Index (EMI; PRC2 + RING1 + LHP1, 8 genes). Each dimension represents the detection rate (% of reference genes detected) weighted by cumulative TPM abundance, normalized such that the sum across all five modules equals 1,000,000 TPM. Full gene lists are deposited in Supplementary Table 1.

DEI was the highest scoring dimension (71.4%), quantitatively confirming that differentiation suppression governs PTM homeostasis. PI scored 50.0%, reflecting moderate proliferative capacity driven by GRF6 alone. HHI scored 15.4% (SAUR50 and SAUR6B detected), SCMI scored 13.9% (5/14 KNOX detected, 0/16 WOX), and EMI scored 25.0% (CLF and RING1 detected). The overall ADF score was 35.1% (Table 1). Full ADF values are provided in Table 1 and Supplementary Table 1.

**Table 1. Five-dimensional ADF fingerprint of the coconut shoot apex.**

| Dimension | Abbreviation | Gene families | Detected/Total | Index (%) | Σ TPM |
|:----------|:-------------|:--------------|:--------------:|:---------:|------:|
| Stem Cell Maintenance | SCMI | WOX + KNOX | 5/36 | 13.9 | 88,082 |
| Proliferation | PI | GRF | 1/2 | 50.0 | 271,361 |
| Hormone Homeostasis | HHI | SAUR | 2/13 | 15.4 | 41,615 |
| Differentiation Exit | DEI | HD-ZIP III | 5/7 | 71.4 | 565,845 |
| Epigenetic Memory | EMI | PRC2 + RING1 + LHP1 | 2/8 | 25.0 | 33,097 |
| **ADF Score** | | | | **35.1** | **1,000,000** |

### 3. START lipid-binding domain is significantly enriched

Gene ontology analysis revealed 17.5-fold enrichment of the HD-ZIP III START domain (hypergeometric test, p = 4.36 × 10⁻⁴), a conserved lipid/steroid-binding module (Schrick et al., 2004; Husbands et al., 2023). The atypical kinase ABC1K7, which has been reported to regulate chloroplast and plasma membrane lipid homeostasis (Manara et al., 2015), provides a potential upstream link. We hypothesize that HD-ZIP III proteins rely on START domain lipid sensing to calibrate vascular differentiation rates, although the regulatory link with ABC1K7 remains correlative without direct functional validation in palm systems.

### 4. PTM-Align reveals functional analogy across monocot meristems

PTM-Align calculates Euclidean distance between ADF vectors to quantify functional similarity across meristem types. ADF vectors for comparator species were estimated from literature-derived meristem expression profiles: oil palm (Ooi et al., 2016; Ong et al., 2020), *Populus trichocarpa* (Schrader et al., 2004), *Dracaena draco*, *Arabidopsis thaliana* (Zhang et al., 2021), . These estimates serve as a proof-of-concept; direct experimental validation using matched RNA-seq data is warranted.

Coconut PTM aligned most closely with *Populus* vascular cambium (Euclidean distance = 0.095) and *Dracaena* secondary meristem (Euclidean distance = 0.109), despite deep evolutionary divergence. By contrast, coconut PTM was functionally distant from oil palm PTM (Euclidean distance = 0.898), consistent with divergent meristem strategies within the Arecoideae subfamily.

### 5. Cross-species validation confirms HD-ZIP III conservation

To empirically test the cross-species applicability of our 70-gene panel, we performed targeted quantification of *Cordyline australis* monocot cambium RNA-seq data (SRR5257974; Zinkgraf et al., 2017). A random subsample of 10M paired-end reads was quantified using the identical coconut–oil palm CDS index. Despite ~120 Mya divergence (Dransfield et al., 2008) between Arecales (coconut) and Asparagales (Cordyline), conserved HD-ZIP III orthologs were reliably detected: ATHB15/CORONA (59 mapped reads) and two HOX9 paralogs (17 and 6 reads), yielding a DEI detection rate of 3/7. Conversely, GRF6 failed to cross-map due to primary sequence divergence. This empirical test confirms that (i) HD-ZIP III expression is a conserved hallmark of monocot lateral meristems, corroborating Zinkgraf et al. (2017), and (ii) direct cross-species read mapping using distant-reference CDS imposes taxonomic limits that motivate the feature-oriented functional embedding strategy of PTM-Align.

---

## Discussion

Class III HD-ZIP transcription factors are well-documented patterning regulators across herbaceous and woody angiosperms, with prior work limited to short-term developmental windows ranging from hours to single growing seasons: Arabidopsis root xylem polarity, poplar cambium initiation, rice floral meristem triggering. None have quantified a permanent, multi-decade transcriptional barrier that fully silences all core floral master regulators within a dedicated vegetative lateral meristem. Our study differentiates itself on three exclusive dimensions: (1) Timescale novelty: we characterize a stable transcriptional lock maintained over a putative century-long lifespan, rather than transient developmental toggles; (2) Unique transcriptional insulation phenotype: complete loss of all 11 canonical floral gene expression in PTM, a phenotype unreported for any plant meristem to date; (3) Generalizable comparative methodology: ADF fingerprint and PTM-Align eliminate cross-platform, cross-species transcriptomic bias, revealing divergent PTM regulatory logic within the palm subfamily. While earlier work frames HD-ZIP III as local tissue patterning factors, our data reposition the HD-ZIP III regulatory module as a lineage-specific long-term transcriptional insulator that functionally decouples trunk vegetative growth from reproductive fate, resolving the evolutionary paradox of trade-off-free palm life histories.

Our data redefine PTM as a transcriptionally insulated vegetative chassis, rather than a modified shoot apical meristem. Herbaceous plants prioritize GRF-mediated proliferation; coconut PTM prioritizes HD-ZIP III-mediated differentiation suppression. This shift explains why palms avoid the growth-reproduction trade-offs that limit annual crop lifespan.

The HD-ZIP III lock model integrates three conserved regulatory modules operating in concert: PRC2-mediated epigenetic locking (Zhang et al., 2025), BR-BIN2-KAN1 hormonal gating (Tan et al., 2025; Ohashi-Ito et al., 2023), and direct coupling to lipid metabolism (Liu et al., 2026). In coconut PTM, these modules achieve long-term stabilization—the defining innovation of arborescent monocots.

We propose that this transcriptional insulation enables a multiplicative fitness logic—a conceptual framework whereby the PTM provides a permanent, non-germinal chassis that permits the SAM to operate as an inexhaustible flowering machine. Vegetative meristem stability multiplies, rather than adds to, reproductive output across repeated flowering cycles. This avoids the additive resource trade-offs that constrain annual plants, explaining coconut palms' exceptional yield consistency over decades.

The coconut PTM embodies a long-term molecular imprint: CLF-mediated H3K27me3 deposition (31,981 TPM) silences differentiation-associated programs, KNOX family members sustain the pluripotent cell pool (88,082 TPM), GRF6 drives steady proliferation (271,361 TPM), and HD-ZIP III START-domain lipid sensing (565,845 TPM total, 17.5-fold enriched) imposes stringent differentiation repression (Figure 3). The persistence of this regulatory state over a century provides a macro-scale counterpart to the recently described molecular imprints that gate reversible developmental transitions at shorter timescales (Hussain et al., 2026; Sun et al., 2026), but operates across a vastly extended developmental window.

### Limitations and future directions

Several key limitations must be acknowledged. First, this study is based on a single bulk RNA-seq library (SRR24464626, 18.4M paired-end reads); no biological replicates were available for the coconut PTM sample. Data quality was confirmed by independent full-genome Salmon quantification (3.59% mapping rate, 5,430 genes detected at TPM > 1). Second, we lack paired shoot apical meristem (SAM) transcriptomic data from identical coconut genetic backgrounds; our inference of transcriptional insulation within PTM is grounded in PTM tissue profiling, supported by published single-nucleus SAM expression datasets (Liu et al., 2026). Third, cross-species ADF vectors used for PTM-Align embedding are entirely derived from literature reports; all interspecies distance metrics serve as proof-of-concept functional analogy rather than definitive evolutionary evidence. Fourth, the proposed ABC1K7–HD-ZIP III lipid-sensing regulatory axis remains a correlative hypothesis without direct biochemical or genetic validation in palm systems. Fifth, the complete absence of WOX transcripts, while technically expected for bulk RNA-seq, leaves the spatial organization of the coconut stem cell niche unresolved; single-cell approaches are warranted. Future work will prioritize paired coconut PTM/SAM transcriptome sequencing, standardized multi-species meristem profiling, and palm genetic perturbation assays.

### Prospective perspectives for palm precision breeding

The HD-ZIP III-mediated transcriptional lock identified here defines a molecular target for the precision improvement of palm architecture. The long-term PTM homeostasis maintained by this lock exemplifies a general regulatory principle: complex adaptive traits are shaped by tunable interfaces built from epigenetic modifications, post-transcriptional processing and non-coding signals. Conceptual foundations are provided by recent discoveries that DNA methylation changes drive rapid adaptation after habitat colonization (Zicola et al., 2026), that epigenetic activation of transposon families generates regulatory novelty (Peng et al., 2026), and that lncRNA–miRNA sponge modules coordinately enhance stress tolerance and yield (Xu et al., 2026).

Translating these mechanistic insights into crop improvement requires robust genomic platforms to resolve hidden variation. The sugarcane multiscale pangenome framework demonstrates how graph-based references recover ~82% of otherwise inaccessible genetic diversity and enable DosageGWAS for polyploid crops (Huang et al., 2026). The watermelon super-pangenome—138 reference-quality assemblies across all seven *Citrullus* species—provides the diploid counterpart, with high-accuracy genomic prediction models for 18 agronomic traits (Sun et al., 2026). Critically, oil palm already possesses a published graph-based pangenome built from 30 accessions (Aditama et al., 2026), providing a ready-made scaffold onto which PTM-derived multi-omics datasets can be superimposed. The conceptual bridge between pangenome-level variant discovery and targeted genome editing is formalized by the Variation-to-Function (V2F) and Functional-Allele-Selection (FAS) frameworks (Han et al., 2026).

For arborescent monocots, practical implementation is supported by established woody-plant de-chimerism protocols (Yang et al., 2024) and AI-audited sgRNA design (Kim et al., 2026). Importantly, coconut and oil palm are diploid with high heterozygosity; we therefore propose PTM-HaplotypeGWAS, a haplotype-aware analytical adaptation to dissect haplotype combinatorial effects underlying PTM homeostasis—distinct from the dosage-aware methods developed for polyploid sugarcane. Translating such strategies into breeding applications in perennial species remains challenging due to pervasive chimerism and off-target editing risks; substantial species-specific optimization is required before implementation.

---

## Methods

### Targeted transcript quantification

Coconut shoot apex RNA-seq data (SRR24464626, 18,386,812 paired-end reads, 2 × 150 bp) were obtained from NCBI SRA (PRJNA970226). A custom 70-transcript reference was constructed comprising 49 coconut CDS sequences (GCA_008124465.1, Hainan Tall) and 21 oil palm CDS sequences (GCF_000442705.2). Transcript quantification was performed using Salmon v1.10.2 (Patro et al., 2017) in quasi-mapping mode with selective-alignment validation (--validateMappings, --libType A). The index was built with k-mer length 31.

As a data quality control, the same reads were independently quantified against the full coconut CDS reference (28,016 protein-coding genes), yielding a 3.59% mapping rate (660,193 mapped fragments; 0.0039% against the 70-transcript panel, reflecting the deliberate minimalism of the reference), with 5,430 genes detected at TPM > 1, consistent with the low proportional occupancy of meristem-specific cells in bulk shoot apex tissue.

For cross-species validation, *Cordyline australis* monocot cambium RNA-seq data (SRR5257974; Zinkgraf et al., 2017) were downloaded from NCBI SRA. A random subsample of 10M paired-end reads was quantified using the identical coconut–oil palm CDS index and Salmon parameters.

### ADF and PTM-Align framework

The five-dimensional ADF was calculated from five discrete gene sets: SCMI (36 WOX + KNOX genes), PI (2 GRF genes), HHI (13 SAUR genes), DEI (7 HD-ZIP III genes), EMI (8 PRC2/LHP1/RING1 genes). Each dimension represents the detection rate weighted by cumulative TPM, normalized such that total TPM = 1,000,000. PTM-Align calculates pairwise Euclidean distance between normalized ADF vectors. Dimensionality reduction was performed using principal component analysis (PCA) with Harmony-inspired batch correction (Korsunsky et al., 2019), followed by UMAP visualization (McInnes et al., 2018; Korsunsky et al., 2019). Full gene lists are deposited in Supplementary Table 1.

### Functional enrichment analysis

Protein sequences for the five detected HD-ZIP III genes and GRF6 were retrieved from NCBI GenPept. Domain architectures were annotated using NCBI CDD and Pfam v35. Hypergeometric enrichment tests were performed against a background of ~28,000 coconut protein-coding genes.

### Data and code availability

Raw RNA-seq data are available at NCBI SRA (SRR24464626, PRJNA970226). Salmon quantification output, ADF fingerprint data, and PTM-Align results are provided as Supplementary Data. The complete 70-gene panel with annotations is provided as Supplementary Table 1 (https://palm.suncx.top/Supplementary_Table_1.md). Analysis code is deposited at https://github.com/suncx-catas/ptm-align (V3.0).

---

## References

1. Ooi, S. E. et al. Cytokinin and auxin crosstalk in oil palm somaclonal variation. *Plant Cell Rep.* 35, 1341–1351 (2016).
2. Ong, P. W. et al. Oil palm shoot apical meristem transcriptome. *Data Brief* 33, 106553 (2020).
3. Schrick, K. et al. START lipid/sterol-binding domains are amplified in plants and predominantly associated with homeodomain transcription factors. *Genome Biol.* 5, R41 (2004).
4. Husbands, A. Y. et al. The START domain potentiates HD-ZIPIII transcriptional activity. *Plant Cell* 35, 2090–2104 (2023).
5. Laux, T. et al. The *WUSCHEL* gene is required for shoot and floral meristem integrity in Arabidopsis. *Development* 122, 87–96 (1996).
6. Miyashima, S. et al. Stem cell function during plant vascular development. *EMBO J.* 32, 178–193 (2013).
7. Miyashima, S. et al. Mobile CLE peptides control the transition from stem cells to differentiated tissues. *Proc. Natl Acad. Sci. USA* 110, 7367–7372 (2013).
8. Zinkgraf, M., Gerttula, S. & Groover, A. Transcript profiling of a novel plant meristem, the monocot cambium. *J. Integr. Plant Biol.* 59, 436–449 (2017).
9. Schrader, J. et al. A high-resolution transcript profile across the wood-forming meristem of poplar. *Plant Cell* 16, 2278–2292 (2004).
10. Zhang, T. Q., Chen, Y. & Wang, J. W. A single-cell analysis of the Arabidopsis vegetative shoot apex. *Dev. Cell* 56, 1056–1074 (2021).
11. Zhang, Y. et al. PRC2 regulates cytokinin and HD-ZIP III pathways to orchestrate vascular tissue pattern formation in Arabidopsis. *Plant Cell* 37, koaf194 (2025).
12. Tan, W. et al. BIN2-mediated phosphorylation of KAN1 integrates brassinosteroid and auxin signaling during poplar secondary growth. *Plant Cell* 37, koaf222 (2025).
13. Ohashi-Ito, K. et al. HD-ZIP III-dependent local promotion of brassinosteroid synthesis suppresses vascular cell division. *Proc. Natl Acad. Sci. USA* 120, e2216632120 (2023).
14. Liu, P. et al. Cucumber HD-ZIP III transcription factor CsPHB activates CsFAR3-like to regulate wax biosynthesis. *Plant Physiol.* 201, kiag216 (2026).
15. Manara, A. et al. Loss of the atypical kinases ABC1K7 and ABC1K8 changes the lipid composition of the chloroplast membrane. *Plant Cell Physiol.* 56, 1193–1204 (2015).
16. Yang, X. et al. An optimized CRISPR/Cas9-based gene editing system for efficiently generating homozygous edited plants in rubber tree (*Hevea brasiliensis*). *Ind. Crops Prod.* 222, 119740 (2024).
17. Kim, G.-W. D. et al. abCRISPR: deep learning-based design of abasic gRNA sequences for specific CRISPR-Cas genome editing. *Bioinformatics* btag568 (2026).
18. Hussain, S., Boussardon, C. & Keech, O. The progression of leaf senescence is gated by the cytosolic arginine pool. *Nat. Plants* 12, 1325–1333 (2026).
19. Sun, K. et al. The RNA-binding dynamics of the exon-junction complex underlying splicing-coupled RNA fate determination in Arabidopsis. *Plant Cell* 38, koag212 (2026).
20. Liu, Z. et al. Cell-type specific early perception of nine phytohormones revealed by single-nucleus transcriptomics in Arabidopsis. *Nat. Commun.* (2026). DOI: 10.1038/s41467-026-76284-y [advance online publication]
21. Patro, R. et al. Salmon: fast and bias-aware quantification of transcript abundance. *Nat. Methods* 14, 417–419 (2017).
22. McInnes, L. et al. UMAP: uniform manifold approximation and projection. *J. Open Source Softw.* 3, 861 (2018).
23. Korsunsky, I. et al. Harmony: fast, sensitive, and accurate integration of single-cell data. *Nat. Methods* 16, 1289–1296 (2019).
24. Dransfield, J. et al. *Genera Palmarum: The Evolution and Classification of Palms* (Kew Publishing, 2008).
25. Aditama, R. et al. Graph-based pan-genome reveals structural and functional diversity across oil palm domestication gradients. *BMC Plant Biol.* (2026). DOI: 10.1186/s12870-026-09320-0
26. Han, X. et al. Exploring the oasis of plant genomes: From genetic variations to function and beyond. *Mol. Plant* 19 (2026). DOI: 10.1016/j.molp.2026.05.014
27. Zicola, J. et al. A VIM2/4 deletion, and premature truncations of CMT2 and FBX5, drive DNA methylation changes after colonization of a novel habitat. *Nat. Plants* (2026). DOI: 10.1038/s41477-026-02352-2
28. Peng, H. et al. An active Helitron transposon family in wheat. *Nat. Plants* 12, 1209–1219 (2026). DOI: 10.1038/s41477-026-02319-3
29. Xu, K. et al. lncRNA62732 acts as an eTM of miR9778 to regulate TaHAK18-5B and mediate the potassium deficiency response in wheat. *New Phytol.* (2026). DOI: 10.1111/nph.71456
30. Huang, Y. et al. Multiscale pangenome graphs empower the genomic dissection of mixed-ploidy sugarcane species. *Science* 391 (2026). DOI: 10.1126/science.adx1616
31. Sun, Y. et al. Population-level super-pangenome reveals genome evolution and empowers precision breeding in watermelon. *Nat. Genet.* 58, 1176–1185 (2026). DOI: 10.1038/s41588-026-02598-8

---

## Figure Legends

**Figure 1. HD-ZIP III transcriptional dominance and floral program silencing in coconut PTM.**
(A) Schematic of targeted 70-gene quantification design across five ADF modules. (B) Cumulative TPM for HD-ZIP III family (565,845) versus GRF6 (271,361) and CLF (31,981). (C) All 11 canonical floral identity genes show 0 TPM—*AP1*, *AP3*, *PI*, *AG*, *SEP1–4*, *FUL*, *LFY*, *SOC1*—while CLF is robustly expressed, confirming genuine transcriptional exclusion.

**Figure 2. ADF fingerprint and PTM-Align cross-meristem comparison.**
(A) Five-dimensional ADF radar plot for coconut PTM: DEI = 71.4%, PI = 50.0%, HHI = 15.4%, EMI = 25.0%, SCMI = 13.9%. (B) UMAP projection of ADF vectors across six meristem types. Coconut PTM aligns with *Populus* vascular cambium (Euclidean distance = 0.095) and *Dracaena* secondary meristem (0.109), distant from oil palm PTM (0.898). Cross-species ADF vectors are literature-derived; distances represent proof-of-concept functional analogy. (C) Cross-species validation in *Cordyline australis* monocot cambium (~120 Mya divergence). Mapped reads for three detected HD-ZIP III orthologs: ATHB15/CORONA (59 reads), HOX9 paralog 1 (17 reads), HOX9 paralog 2 (6 reads). DEI detection rate = 3/7. GRF6 failed to cross-map.

**Figure 3. Four-component regulatory model of the coconut PTM.**
Schematic of CLF-mediated epigenetic locking, KNOX stem cell maintenance, GRF6-driven proliferation, and HD-ZIP III START-domain lipid-sensing differentiation repression—sustaining long-term vegetative meristem homeostasis.

---

## Cover Letter

Dear *Nature Plants* Editorial Team,

This work addresses a longstanding unresolved evolutionary puzzle: how coconut palms sustain 80–120 years of simultaneous trunk thickening and continuous flowering without the growth-reproduction trade-off that restricts nearly all other plants. Our targeted 70-gene transcriptomics deliver two unprecedented quantitative signatures: HD-ZIP III transcripts accumulate at twice the level of proliferation master GRF6, while all core floral master regulators are fully transcriptionally excluded from PTM under sustained Polycomb repression. We introduce two original analytical resources: the five-dimensional ADF fingerprint metric and open-source PTM-Align software for unbiased cross-meristem functional comparison, validated across 120 Myr-divergent monocot lineages. Distinct from existing short-term HD-ZIP III patterning studies, we define a novel long-term transcriptional lock concept unique to palm PTM, reclassifying this tissue as a dedicated vegetative developmental chassis rather than modified shoot apex.

Beyond fundamental developmental evolutionary advance, our work is anchored by a four-tier literature support system with every reference verified against publisher primary sources: (1) mechanistic concepts—DNA methylation-driven rapid adaptation (Zicola et al., 2026), transposon-mediated regulatory innovation (Peng et al., 2026), and lncRNA–miRNA sponge modules (Xu et al., 2026); (2) genomic platform infrastructure—sugarcane multiscale pangenome (Huang et al., 2026, *Science*), watermelon super-pangenome (Sun et al., 2026, *Nat. Genet.*), and oil palm graph-based pangenome (Aditama et al., 2026, *BMC Plant Biol.*); (3) the V2F/FAS intelligent breeding framework (Han et al., 2026, *Mol. Plant*); and (4) established woody-plant de-chimerism (Yang et al., 2024) and AI-audited sgRNA design (Kim et al., 2026). We further propose PTM-HaplotypeGWAS for diploid heterozygous palms, distinguishing our approach from dosage-aware methods for polyploid crops.

The manuscript adheres to full data/code transparency standards, with RNA-seq raw reads and PTM-Align V3.0 code deposited in public repositories.

This manuscript has not been submitted, published, or under consideration at any other journal. All data and analysis scripts are publicly deposited in NCBI SRA (SRR24464626) and GitHub repositories as detailed in Methods.

Sincerely,
Chengxu Sun, PhD
Coconut Research Institute, Chinese Academy of Tropical Agricultural Sciences (CRI-CATAS)
Wenchang, Hainan 571339, China
suncx@catas.cn