Non-destructive quantification of shoot apical meristem homeostasis for prediction of plant architecture and biomass using robot-based 3D imaging and photosynthesis measurements

Abstract

Shoot apical meristem (SAM) homeostasis integrates environmental and genetic cues to regulate growth dynamics that drive biomass accumulation and crop yield; however, no robust, non-destructive, quantitative proxy has been established for modeling or monitoring SAM-homeostasis-associated dynamics. Here, we developed a novel robot-based 3D imaging system and a custom pot-chamber gas exchange system to non-destructively measure plant occupation volume (POV) and whole-plant photosynthetic rate in wild-type Arabidopsis plants and nine mutants with disrupted SAM homeostasis. We demonstrate that POV robustly captures 3D plant architecture, whereas whole-plant photosynthetic rate serves as a superior proxy for optimal growth dynamics and final biomass associated with SAM homeostasis, outperforming conventional traits such as leaf number, leaf size, total leaf area, and rosette diameter. The strong positive correlations among POV, whole plant photosynthesis, and biomass accumulation establish a powerful new framework for quantitative studies of SAM homeostasis and data-driven evaluation of plant architecture.

This study introduces a robotic 3D imaging system and a custom pot-chamber gas exchange system for nondestructive quantification of plant occupation volume (POV) and whole-plant photosynthetic rate (Ap) in Arabidopsis. It shows that POV and Ap outperform conventional traits as proxies for growth dynamics and biomass accumulation. The strong correlations among POV, Ap, and biomass establish a quantitative framework for non-destructive quantification of shoot apical meristem homeostasis and prediction of plant architecture.

Introduction

Photosynthesis provides the primary energy and carbon source for plant growth and development, converting light energy into chemical energy and enabling carbon assimilation and biomass accumulation (Blankenship, 2002; Kromdijk et al., 2016). A bidirectional positive feedback loop couples photosynthesis and growth: increased leaf area and improved leaf physiology enhance whole-canopy CO₂ fixation and photosynthetic capacity (Kromdijk et al., 2016), and photosynthetically fixed carbon supplies substrates for organ initiation and sustained biomass accumulation (Evans, 2013). This reciprocal interaction creates a tight functional linkage between canopy photosynthesis and whole-plant growth (Wang et al., 2018; Chang et al., 2019; Yang et al., 2023; Ge et al., 2024).

Extensive research in Arabidopsis has shown that the shoot apical meristem (SAM), the primary stem-cell niche, maintains homeostasis through coordinated cell proliferation in the central zone, differentiation in the peripheral zone, and signaling maintenance in the organizing center (Ian, 1972). The SAM regulates the initiation, number, size, and spatial arrangement of plant organs, thereby determining canopy structure and reproductive yield (Laux et al., 1996; Schoof et al., 2000b; Su et al., 2020; Lopes et al., 2021). This SAM activity establishes a fundamental developmental connection among canopy architecture, light interception efficiency, photosynthetic performance, and biomass accumulation (Wang et al., 2018; Chang et al., 2019; Yang et al., 2023; Ge et al., 2024). Consequently, plant architecture, which integrates leaf number, morphology, and spatial arrangement, serves as a critical bridge between SAM-driven development and whole-plant function.

At the molecular level, the canonical CLAVATA (CLV)–WUSCHEL (WUS) feedback network constitutes the core regulatory network for SAM homeostasis. The CLV3/EMBRYO SURROUNDING REGION (CLE) peptide CLV3, secreted by stem cells of the central zone (Clark et al., 1996; Müller et al., 2006), interacts with the receptor kinase CLV1 and the CLV2/CORYNE (CRN) complex to repress WUS expression in the organizing center and thereby limit stem-cell proliferation (Fletcher et al., 1999; Jeong et al., 1999; Schoof et al., 2000a; Guo et al., 2010). Additional receptor-like kinases, including the BARELY ANY MERISTEM (BAM) family members BAM1, BAM3, and RECEPTOR-LIKE PROTEIN KINASE 2 (RPK2), promote differentiation in the peripheral zone through partially redundant, spatially restricted signaling (DeYoung et al., 2006; Mizuno et al., 2007; Guo et al., 2010). Although conserved across crops, this pathway exhibits evolutionary divergence; rice contains multiple stage-specific CLV3 homologs (Suzaki et al., 2006; Kinoshita et al., 2007), whereas maize uses the FASCIATED EAR (FEA3)–FON2-LIKE CLE PROTEIN1 (FCP1) module to relay differentiation signals from peripheral tissues to the SAM (Bommert et al., 2013). These variations highlight the pathway’s complexity and offer diverse molecular targets for optimizing crop architecture and yield.

Despite extensive characterization of the CLV–WUS pathway, the mechanisms by which SAM homeostasis dynamically shapes 3D plant architecture and canopy function remain poorly understood, owing primarily to methodological limitations. Traditional techniques, such as manual leaf counting, destructive biomass sampling, and single-leaf gas exchange measurements, fail to capture dynamic, organ-level 3D traits (Pouteau and Albertini, 2009). Existing canopy measurement systems (e.g., canopy photosynthesis and transpiration measurement system) (Song et al., 2017) and 3D models (Zheng et al., 2008; Song et al., 2013; Chang et al., 2019) are poorly suited for compact model species like Arabidopsis because of (1) sensor–plant scale mismatch, (2) low throughput for organ-resolved traits, and (3) dependence on destructive or low-frequency measurements (Chang et al., 2019; Elnashef et al., 2019). Traditional functional–structural models also fail to resolve fine-scale architectural features of compact canopies (España et al., 1999; Jiang et al., 2016), and integrative indices that link 3D architecture to canopy performance, such as the maize canopy occupation volume index (Liu et al., 2021), lack counterparts in small-stature species, further impeding mechanistic studies of how SAM homeostasis translates into whole-plant photosynthetic function.

To address these challenges, we analyzed mutants of nine core CLV–WUS pathway genes (wus, clv1, clv2, clv3, bam1, bam3, rpk2, cik123, and cik234) and pursued two integrated objectives: (1) development of non-destructive, Arabidopsis-specific phenotyping tools capable of resolving dynamic 3D architectural traits at the organ level and (2) establishment of quantitative links among SAM homeostasis, 3D plant architecture, and whole-plant photosynthesis. We demonstrate that measurements obtained with a non-destructive plant phenotyping system, including plant occupation volume (POV) and whole-plant photosynthetic rate, serve as robust, integrative indicators for quantitative investigation of SAM homeostasis and its effects on plant architecture.

Results

A non-destructive phenotyping platform links SAM-driven architecture to plant photosynthesis

To overcome the scale and throughput limitations of existing crop-oriented systems, we developed two non-destructive tools tailored to Arabidopsis: (1) a custom pot-chamber gas exchange system for measuring whole-plant photosynthetic rate (Ap; Figure 1A and Supplemental Figures 1A–1C) and (2) an automated, robot-assisted 3D imaging system for quantifying organ-level architectural traits, specifically POV (Figure 1B). The pot chamber, which features a minimized volume to match flow-rate requirements for stable CO₂ concentration, includes a water-filled trough for an airtight seal (Figure 1A and Supplemental Figure 1C). Equipped with a portable gas exchange system, the chamber measures CO₂ concentration using an integrated infrared gas analyzer under light-emitting diode (LED) lighting (Figure 1A and Supplemental Figure 1B). The robotic 3D imaging system consists of a robotic arm, a camera, and a rotation stage and was designed to automatically capture multi-view images for 3D point cloud reconstruction, thereby enabling quantification of POV and refined architectural traits (Figure 1B). By integrating Ap and POV, this dual-module platform provides an optimized proxy for evaluating plant growth dynamics in diverse genetic backgrounds (Figure 1C).

Figure 1.

Non-destructive phenotyping platform for measurement of plant photosynthesis and architecture in Arabidopsis.

The workflow for the non-destructive phenotyping platform comprises two modules.

(A) A customized pot-chamber-based system measures whole-plant photosynthesis, with effective chamber volume (V) quantified from the measured gas stabilization time (t_stab) and molar flow rate (flow) using the formula $V=\frac{t\text{stab}\ast flow}{K}$. This system integrates a specialized measurement device with a portable photosynthesis system (LI-6400) for real-time monitoring of photosynthetic traits under controlled environmental conditions.

(B) An automated robotic platform enables high-throughput 3D phenotyping via a robotic arm for multi-view image acquisition, followed by a standardized 3D point cloud pipeline to derive plant occupation volume (POV) and extract architectural features, facilitating the quantitative assessment of spatial morphological development.

(C) Integrating POV with the whole-plant CO₂ assimilation rate (Ap) provides an optimized proxy for evaluating growth dynamics across diverse genetic backgrounds. Chamber volume: $V=\frac{t\text{stab}\ast flow}{K}$, where t_stab is the stabilization time, flow is the molar flow rate, and K is a system-specific correction factor. POV: $\text{POV}={\sum }_1^{n}Vi$, where Vi represents the volume of each voxel occupied by the plant 3D point cloud. Whole-plant CO₂ assimilation rate: $Ap=\frac{F\ast \left(C_e-C_O\right)}{100}$, where F is the molar flow rate, and C_e and C_O are the CO₂ mole fractions entering and leaving the chamber, respectively.

We grew wild-type Columbia-0 (Col) and mutants of nine core CLV–WUS pathway genes (wus, clv1, clv2, clv3, bam1, bam3, rpk2, cik123, and cik234) to evaluate the informativeness of plant-level traits across regulatory perturbations of the CLV–WUS pathway in Arabidopsis. The wus mutant showed severe growth impairment, with reduced aboveground biomass, whole-plant photosynthetic rate, and POV compared with Col (Figures 2A–2C and Supplemental Figures 1D and 1E). From 2 to 21 days after transplanting (DAT), wus produced fewer green leaves and exhibited a substantially smaller leaf length, leaf width, and rosette diameter compared with Col (Figures 2D and 2E). Notably, leaf-level photosynthetic rate (per unit area; Supplemental Figure 1F) and chlorophyll fluorescence parameters were indistinguishable between wus and Col (Figures 2F and 2G), indicating that growth defects in wus arise primarily from architectural constraints rather than impaired leaf physiology.

Figure 2.

Impaired SAM homeostasis causes severe growth defects in the wus mutant.

(A) Representative images and dry biomass of wild-type Col and wus. Scale bar: 7.5 cm.

(B and C) Plant photosynthetic rate (Ap) and POV in Col and wus.

(D) Temporal dynamics of leaf number in Col and wus across different growth stages.

(E) Statistical analysis of rosette diameter and leaf length and width in Col and wus.

(F) Statistical analysis of leaf-level photosynthesis (A) in Col and wus.

(G) Statistical analysis of chlorophyll fluorescence parameters (Fv/Fm, YII, and NPQ) in Col and wus across different growth stages.

(H and I) Correlations of Ap with biomass (n = 15, H) and POV with Ap in Col and wus (n = 15, I).

(J) Correlations of leaf traits and rosette diameter with POV in Col and wus (n = 15).

Data represent ≥ 3 plants. DAT, days after transplanting. r, Pearson correlation coefficient. Asterisks indicate significant differences determined by two-tailed t-tests (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001). Data are presented as mean ± SD, with dashed lines representing the error bars.

Correlation analyses confirmed the functional relationships among traits in Col and wus. Ap was strongly correlated with final biomass (Pearson correlation coefficient r = 0.58, p < 0.0231; Figure 2H), confirming that Ap is a reliable proxy for growth. POV exhibited an even tighter correlation with Ap (r = 0.83, p < 0.001; Figure 2I) and was strongly associated with all measured architectural parameters (green leaf number: r = 0.92; leaf length: r = 0.88; leaf width: r = 0.59; rosette diameter: r = 0.91; all p < 0.05; Figure 2J), establishing POV as the primary structural determinant of whole-plant photosynthetic performance.

Together, these results demonstrate that SAM dysfunction in wus restricts POV through architectural defects, limiting whole-plant photosynthesis and biomass accumulation without compromising intrinsic leaf photosynthetic capacity. Our integrated platform, combining precise Ap measurements, which directly predict biomass, with high-resolution 3D quantification of POV and organ traits, provides a robust, non-destructive framework for monitoring SAM-driven growth dynamics and linking developmental genetics to whole-plant physiological performance.

Plant photosynthesis and POV reliably detect SAM dysregulation in mutants with altered leaf initiation

To evaluate the informativeness of plant-level traits under conditions of disrupted leaf initiation rates, we examined other loss-of-function mutants in the core CLV–WUS signaling pathway, clv3 and clv1 (Figure 3A and Supplemental Figure 2A and 2B). The clv3 mutant produced significantly more leaves than Col during later vegetative growth and consequently had more green leaves (Figure 3B) but exhibited substantially reduced Ap and biomass (Figure 3C and 3D). Although the new leaves of the clv3 mutant eventually exceeded those of Col in length, width, and area (Figure 3E), persistent reductions in rosette diameter and in the length and width of the longest rosette leaf, which are key traits for light interception and resource utilization (Figure 3F), caused severe rosette dwarfism and markedly lower POV (Figure 3G and Supplemental Figure 2C). For the clv3 mutant, leaf-level photosynthetic rate and chlorophyll fluorescence parameters were indistinguishable from those of Col (Figures 3H and 3I and Supplemental Tables 1–3), indicating that the decline in Ap arose primarily from architectural constraints on POV rather than from photosynthetic dysfunction.

Figure 3.

Disrupted SAM homeostasis impairs leaf initiation rate in clv3 and clv1 mutants.

(A–I) Phenotypic and physiological traits of Col and the clv3 mutant.

(A) Plant morphology images. Scale bar: 7.5 cm.

(B) Dynamics of leaf emergence at different growth stages and statistical analysis of green leaf number.

(C) Statistical analysis of Ap at various growth stages.

(D) Statistical analysis of biomass.

(E) Analysis of leaf length, width, and area of all leaves according to emergence order (n ≥ 5 plants).

(F) Statistical analysis of the length and width of the longest leaf and rosette diameter.

(G) Statistical analysis of POV at various growth stages.

(H) Statistical analysis of leaf-level photosynthetic rate (A).

(I) Dynamic changes in Fv/Fm, YII, and NPQ across different growth stages (n ≥ 4 plants).

(J–O) Experimental data for Col and the clv1 mutant.

(J) Plant morphology images. Scale bar: 7.5 cm. Statistical analysis of leaf number, length, and width of the longest leaf, and rosette diameter.

(K) Statistical analysis of POV at various growth stages.

(L) Statistical analysis of leaf-level A.

(M) Dynamic changes in Fv/Fm, YII, and NPQ across different growth stages (n ≥ 4 plants).

(N) Statistical analysis of Ap at various growth stages.

(O) Statistical analysis of biomass. n ≥ 12.

Asterisks indicate significant differences determined by two-tailed t-tests (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001). Data are presented as means ± SD, with dashed lines representing error bars.

By contrast, the clv1 mutant had fewer green leaves (Figure 3J), with significant reductions in the length and width of the longest rosette leaf and in rosette diameter (Figure 3J), resulting in severely reduced POV compared with Col (Figure 3K and Supplemental Figure 2C). Although the clv1 mutant showed an elevated leaf-level photosynthetic rate (Figure 3L) with unchanged fluorescence parameters (Figure 3M and Supplemental Tables 1–3), both Ap and biomass were significantly lower than those of Col (Figure 3N and 3O). Thus, architectural limitations imposed by insufficient leaf numbers and restricted POV outweighed any compensatory increase in single-leaf efficiency.

Collectively, these results demonstrate that, whether leaf numbers increase (clv3) or decrease (clv1), architectural defects restrict POV and thus limit whole-plant photosynthesis and biomass accumulation. POV and Ap therefore serve as robust, integrative readouts of SAM homeostasis across various leaf-initiation and organ-size perturbations.

Plant photosynthesis and POV serve as sensitive indicators of SAM homeostasis in mutants with defective leaf expansion

To determine whether plant-level traits can serve as sensitive indicators of SAM homeostasis when leaf initiation is unaffected, we examined mutants that exhibited impaired leaf expansion but normal phyllotaxy. The single mutants rpk2 and clv2 exhibited normal leaf-number dynamics (Figures 4A–4C and Supplemental Figures 3A and 3B) but significantly reduced POV, Ap, and biomass compared with Col (Figure 4D and 4E and Supplemental Figure 3C). Leaf traits, including length, width, and area for all leaves according to emergence order, were strongly reduced in the rpk2 mutant (Figure 4F). Moreover, the length and width of the longest rosette leaf and rosette diameter were uniformly reduced in both rpk2 and clv2 mutants compared with Col (Figures 4G and 4H), whereas leaf-level photosynthetic rates and chlorophyll fluorescence parameters remained indistinguishable from those of Col (Figures 4I and 4J; Supplemental Tables S1–S3).

Figure 4.

Disrupted SAM homeostasis impairs leaf expansion but preserves normal phyllotaxy in rpk2 and clv2 mutants.

(A) Plant morphology images of Col, rpk2, and clv2. Scale bar: 7.5 cm.

(B) Dynamics of leaf emergence across growth stages and statistical comparison of green leaf number between Col and rpk2.

(C) Statistical analysis of leaf number in Col and clv2.

(D and E) Statistical analysis of POV, Ap, and biomass in Col compared with rpk2(D) and clv2(E) across various growth stages.

(F) Analysis of leaf length, leaf width, and area of all leaves from Col and rpk2 plants categorized by leaf emergence order (n ≥ 5 plants).

(G and H) Statistical analysis of rosette diameter and leaf length and width in Col and rpk2(G) and in Col and clv2(H).

(I and J) Dynamic changes in Fv/Fm, YII, and NPQ in Col and rpk2(I) and in Col and clv2(J) across different growth stages (n ≥ 4 plants) and statistical analysis of A in Col and rpk2(I) and in Col and clv2(J). n ≥ 12.

Asterisks indicate significant differences determined by two-tailed t-tests (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001). Data are presented as means ± SD, with dashed lines representing error bars.

Similarly, the triple mutants cik1/2/3 and cik2/3/4 exhibited normal green leaf number (Figures 5A and 5B and Supplemental Figures 4A and 4B), but their leaf expansion was strongly impaired throughout vegetative growth (Figure 5C and 5D). For cik1/2/3, this impairment resulted in a 10% reduction in rosette diameter and longest rosette leaf length, an 18% decrease in longest rosette leaf width, 2%–25% reductions in POV, 24% and 38% decreases in Ap at different developmental stages, and a 23% reduction in biomass relative to Col (Figures 5E–5H and Supplemental Figure 4C). By contrast, cik2/3/4 exhibited more pronounced defects, with 37% and 44% reductions in longest rosette leaf length and width, a 40% decrease in rosette diameter, 61%–75% reductions in POV, 64% and 52% decreases in Ap at different developmental stages, and a 60% reduction in biomass relative to Col (Figures 5E–5H and Supplemental Figure 4C). Phenotype severity was greater in cik2/3/4 than in cik1/2/3, consistent with the partial functional redundancy among CIK receptors (Hu et al., 2018). Leaf-level photosynthetic capacity was again unaffected in these mutants compared with Col (Figures 5I and 5J and Supplemental Tables S1–S3).

Figure 5.

Disrupted SAM homeostasis impairs leaf expansion but preserves normal phyllotaxy in cik123 and cik234 triple mutants.

(A) Plant morphology images of Col, cik123, and cik234. Scale bar: 7.5 cm.

(B) Dynamic changes in green leaf number in Col and cik123 and Col and in cik234 across different growth stages.

(C and D) Analysis of leaf length, leaf width, and area of all leaves in Col and cik123(C) and in Col and cik234(D) according to leaf emergence order (n ≥ 5 plants).

(E and F) Statistical analysis of rosette diameter and leaf length and width in Col and cik123(E) and in Col and cik234(F).

(G and H) Statistical analysis of POV, Ap, and biomass in Col and cik123(G) and in Col and cik234(H) across growth stages.

(I and J) Dynamic changes in Fv/Fm, YII, and NPQ in Col and cik123(I) and in Col and cik234(J) across different growth stages (n ≥ 4 plants) and statistical analysis of A in Col and cik123(I) and in Col and cik234(J). n ≥ 12.

Asterisks indicate significant differences determined by two-tailed t-tests (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001). Data are presented as means ± SD, with dashed lines representing error bars.

Thus, in mutants where SAM homeostasis disrupts organ size but not initiation rates, whole-plant photosynthesis and POV reliably capture growth impairment, outperforming traditional 2D metrics such as individual leaf size or rosette diameter. These integrative, non-destructively measured traits therefore serve as highly sensitive proxies for SAM-driven architectural defects across diverse genetic backgrounds.

Plant photosynthesis reveals subtle 3D architectural defects missed by conventional traits in bam1 and bam3 mutants

Building on the above analyses of mutants with altered leaf initiation or expansion, we next tested whether plant-level measurements could reveal growth defects arising from 3D architectural changes. We examined the bam1 and bam3 mutants (Supplemental Figures 5A and 5B). In the bam1 mutant, Ap and aboveground biomass were significantly lower than those of Col throughout the vegetative growth phase (Figures 6A–6C). These reductions coincided with markedly reduced POV in the bam1 mutant compared with Col (Figure 6D and Supplemental Figure 5C). Detailed measurements of individual leaf morphological and quantitative parameters, including the length and width of the longest rosette leaf and rosette diameter (Figure 6E), leaf emergence number and green leaf number (Figure 6F), and leaf length, leaf width, and area of all leaves according to leaf emergence order (Figure 6G), revealed no statistically significant differences between bam1 and Col. Similarly, leaf-level photosynthetic rate and chlorophyll fluorescence parameters of the bam1 mutant were indistinguishable from those of Col (Figure 6H; Supplemental Tables S1–S3). Thus, the reduced Ap in the bam1 mutant was attributed primarily to 3D architectural constraints captured by POV.

Figure 6.

Structural perturbations in 3D plant architecture underlie growth defects in bam1 and bam3 mutants.

Shown are phenotypic and physiological traits for Col and the bam1 mutant (A–H) and for Col and the bam3 mutant (I–P).

(A and I) Plant morphology images of Col and bam1(A) and Col and bam3(I). Scale bars: 7.5 cm.

(B and J) Statistical analysis of biomass in Col and bam1(B) and in Col and bam3(J).

(C and K) Statistical analysis of Ap in Col and bam1(C) and in Col and bam3(K) at various growth stages.

(D and L) Statistical analysis of POV in Col and bam1(D) and in Col and bam3(L) at various growth stages.

(E and M) Statistical analysis of rosette diameter, leaf length, and leaf width in Col and bam1(E) and in Col and bam3(M).

(F and N) Dynamics of leaf emergence number and statistical analysis of green leaf number in Col and bam1(F) and in Col and bam3(N) across different growth stages.

(G and O) Analysis of rosette leaf length, leaf width, and area of all leaves from Col, bam1, and bam3 plants according to leaf emergence order (n ≥ 5 plants).

(H and P) Dynamic changes in Fv/Fm, YII, and NPQ in Col and bam1(H) and in Col and bam3(P) across different growth stages (n ≥ 4 plants) and statistical analysis of A in Col and bam1(H) and in Col and bam3(P). n ≥ 12.

Asterisks indicate significant differences determined by two-tailed t-tests (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001). Data are presented as means ± SD, with dashed lines representing error bars.

The bam3 mutant displayed a phenotype similar to that of bam1, with lower Ap and biomass (Figures 6I–6K), reduced POV (Figure 6L and Supplemental Figure 5C), but no differences in leaf number, leaf morphology, rosette diameter (Figures 6M–6O), or leaf-level physiology compared with Col (Figure 6P).

These results establish that whole-plant photosynthesis is highly sensitive to subtle 3D architectural perturbations caused by SAM dysregulation, even when traditional 1D and 2D traits remain unaffected. POV emerges as a critical structural mediator that translates these otherwise hidden spatial defects into measurable reductions in whole-plant carbon assimilation and biomass accumulation.

To further quantify our platform’s sensitivity, we performed correlation analyses between aboveground biomass and Ap, as well as between biomass and POV, using datasets from Col, bam1, and bam3 plants. A significance level of p < 0.05 was used to determine the minimum detectable phenotypic difference threshold. In scenarios with subtle biomass reductions of 12%–27.4% in bam1 and bam3 mutants relative to Col, both Ap (A: r = 0.6880, p = 0.0279; B: r = 0.6473, p = 0.0229) and POV (C: r = 0.5370, p = 0.0320; D: r = 0.5099, p = 0.0365) were significantly positively correlated with aboveground biomass (p < 0.05), and the wild type exhibited a significantly higher Ap and POV values than the mutants (Supplemental Figure 6). These findings confirm that the platform is sufficiently sensitive to detect fine-grained phenotypic variations and their physiological/morphological correlations.

Plant photosynthesis and POV serve as universal quantitative proxies for SAM homeostasis

To further assess the reliability of Ap as a non-destructive measure of biomass changes, we performed a correlation analysis incorporating Ap and aboveground biomass. Across all genotypes and three mutant categories, Ap exhibited a robust positive correlation with biomass (Figures 7A–7D), establishing whole-plant photosynthesis as a robust, genotype-independent proxy for SAM-regulated growth and productivity and thereby addressing a significant gap in the quantitative assessment of meristem phenotyping.

Figure 7.

Correlations of Ap with biomass, leaf traits, rosette diameter, and POV across genotypes.

This figure presents correlations between various traits across all genotypes and three mutant categories (clv3 and clv1; rpk2, clv2, and cik mutants; and bam1 and bam3).

(A–D) Correlations between Ap and biomass for all genotypes (n = 99, A); clv3 and clv1 (n = 34, B); rpk2, clv2, and cik mutants (n = 54, C); and bam1 and bam3 (n = 32, D).

(E–H) Correlations between POV and Ap for all genotypes (n = 111, E); clv3 and clv1 (n = 34, F); rpk2, clv2, and cik mutants (n = 58, G); and bam1 and bam3 (n = 34, H).

(I–L) Correlations between Ap and green leaf number, leaf length, leaf width, rosette diameter, leaf inclination angles, and leaf curvature for all genotypes (n = 98–99, I); clv3 and clv1 (n = 34, J); rpk2, clv2, and cik mutants (n = 54, K), and bam1 and bam3 (n = 32, L). r denotes the Pearson correlation coefficient. p values are indicated for each plot (red: p < 0.05; blue: p ≥ 0.05).

To further evaluate the robustness of POV as a reliable structural proxy for plant performance, we performed a comprehensive correlation analysis of leaf number, leaf size, rosette diameter, refined architectural traits (including leaf inclination angles and curvature), and POV against Ap across all genotypes and three mutant categories. POV emerged as the leading structural determinant, with the highest correlation coefficients among all other traits across genotypes and within each mutant category (Figures 7E–7H). Although some traits showed significant correlations with Ap in the “all genotypes” group, their correlation coefficients were consistently lower than those of POV (Figures 7E–7L). Furthermore, unlike individual traits, which failed to maintain consistent significance across specific mutant categories, POV remained a robust and consistent predictor in every group (Figures 7E–7L). Notably, in the bam1, bam3 category, leaf curvature showed a significant positive correlation with Ap (Figure 7L; r = 0.5305, p = 0.0018). This positions POV as a superior integrative 3D trait that reliably captures SAM-regulated architectural variation, significantly outperforming traditional 2D metrics.

By contrast, correlations of individual morphological traits (green leaf number, size, rosette diameter, leaf inclination angles, and leaf curvature) with POV were highly genotype-specific, revealing four distinct SAM regulatory strategies (Figures 2J and Supplemental Figures 7A–7D). (1) wus disrupts both leaf initiation and expansion, causing a synergistic decline in POV. (2) clv1 and clv3 primarily restrict POV through impaired leaf initiation and the subsequent suppression of individual leaf expansion. (3) clv2, rpk2, and cik mutants restrict POV predominantly through reduced leaf expansion and have minimal impact on leaf number. (4) bam1 and bam3 impair POV primarily through changes in 3D spatial arrangement without affecting leaf number or 2D morphology, revealing a distinct correlation structure in which only leaf curvature shows a significant positive association with POV in this group (Supplemental Figure 7D: r = 0.3977, p = 0.0180).

Thus, whereas conventional traits provide only context-dependent information, Ap and POV together provide universal, non-destructive proxies that quantitatively integrate diverse SAM perturbations into consistent predictions of growth and biomass accumulation.

Discussion

Plant biomass and yield ultimately reflect cumulative net CO₂ assimilation throughout the life cycle (Evans, 2013; Wu et al., 2023). Although leaf-level traits define intrinsic photosynthetic capacity, our results demonstrate that Ap is the most reliable indicator of growth dynamics and biomass. Whole-plant photosynthesis is governed primarily by 3D architecture and its influence on light distribution, rather than by cellular biochemistry alone (Song et al., 2013, 2017). Consequently, biomass reductions in architectural mutants arise not simply from smaller organs but from impaired light interception caused by structural disorder (Foo et al., 2020; Zhu et al., 2020). Thus, Ap integrates developmental integrity with physiological performance, serving as the principal proxy for optimal plant growth.

Using a comprehensive panel of CLV–WUS pathway mutants, we demonstrated that diverse genetic perturbations—ranging from simultaneous disruption of leaf number and organ expansion (wus; Figures 2D and 2E), altered leaf initiation and size (clv1 and clv3; Figures 3B, 3E, 3F, and 3J), and restricted leaf expansion (clv2, rpk2, and cik mutants; Figures 4F–4H, and Figures 5–5F) to subtle perturbations in 3D spatial arrangement that do not affect 2D morphology (bam1 and bam3; Figures 6E–6G and 6M–6O)—uniformly converge on reduced POV. In all cases, reduced POV limits Ap and biomass accumulation. The superior sensitivity of our 3D framework is thus confirmed by its ability to detect subtle spatial rearrangements in mutants such as bam1 and bam3 that were undetectable by conventional 2D analysis. Although discrete parameters such as leaf number, leaf size, rosette diameter, leaf inclination angle, and leaf curvature capture specific structural components, POV reflects the cumulative biophysical effects of these variations (Figures 7E–7L). Accordingly, POV serves as the central structural hub that translates SAM dysregulation into whole-plant physiological outputs, particularly for complex 3D architecture features (e.g., leaf overlap, petiole elevation, and petiole curvature) at later stages. These results underscore a universal biophysical principle: the 3D spatial footprint, rather than individual 2D geometric traits, governs plant capacity for light interception (Monsi and Saeki, 2005).

Given the high evolutionary conservation of the CLV–WUS pathway in angiosperms (Kitagawa and Jackson, 2019; Demesa-Arevalo et al., 2024), functional orthologs of several pathway members have been identified in major crops (e.g., rice and maize). These orthologs may modulate SAM activity to optimize POV and architecture via mechanisms similar to those in Arabidopsis. Thus, the framework for non-destructive quantification of plant architecture established here provides a potential translational bridge between developmental biology and elite crop breeding for ideal architectures, laying a theoretical foundation for data-driven enhancement of crop photosynthetic efficiency and biomass.

Optimizing plant-level photosynthetic efficiency is a key objective in modern crop breeding (Tian et al., 2024). Current strategies for ideal plant architecture breeding typically target macroscopic traits such as leaf angle and plant height, often overlooking the dynamic coupling between plant structure and photosynthetic function. The non-destructive, high-resolution phenotyping platform developed here addresses this limitation by enabling (1) time-resolved monitoring of growth dynamics, (2) quantitative attribution of Ap to specific 3D architectural components, and (3) high-throughput genotype–phenotype mapping. To further optimize system throughput, the duration of gas exchange measurements can be reduced by adjusting flow rates for faster stabilization of CO₂ and H₂O concentrations. Notably, the traditional “structure from motion” algorithm adopted in our current workflow can be improved with 3D Gaussian Splatting (Jiang et al., 2025; Wang et al., 2025), which enables faster 3D reconstruction with fewer images per plant scan. Furthermore, our 3D phenotyping system captures high-quality imaging data for the generation of dense point clouds, enabling precise quantification of subtle architectural variations and complementing the core capabilities above, thereby reinforcing its utility for genetics-guided crop breeding research (Song et al., 2023). Our findings demonstrate that the integration of Ap and POV offers a robust framework for detecting subtle phenotypic differences, as shown by the system’s sensitivity to biomass reductions of 12%–27.4% in bam1 and bam3 mutants relative to Col (Supplemental Figure 6). This level of resolution highlights the platform’s potential for assessing subtle growth and physiological variation across diverse crop species. Overall, this system provides a robust, genetics-guided tool for optimizing plant architecture and accelerating the development of high-yielding crops with enhanced light-use efficiency.

Methods

Plant materials

All experiments used Arabidopsis thaliana accession Col as the wild-type background. CRISPR-Cas9-generated knockout mutants included clv3 (At2G27250), clv1 (At2G27250), clv2 (At1G65380), rpk2 (At3G02130), and wus (AT2G17950) produced in this study. The T-DNA insertion lines bam1 (SALK_059053C) and bam3 (SALK_107016C) were obtained from the Arabidopsis Biological Resource Center and have been characterized previously (Alonso et al., 2003; Qu and Qin, 2014). The triple mutants cik123 (cik1/cik2/cik3) and cik234 (cik2/cik3/cik4) were kindly provided by Prof. Xiaoping Gou at Lanzhou University (Cui et al., 2018) and were derived from the corresponding single T-DNA insertion lines cik1 (SALK_034037), cik2 (SALK_066568), cik3 (SALK_111290), and cik4 (SALK_049669).

Plant growth conditions

Plants were grown in a controlled-environment chamber under a 16-h light/8-h dark photoperiod with a photosynthetic photon flux density of 120 μmol m⁻² s⁻¹ (white LEDs), a constant temperature of 22°C ± 1°C, and a relative humidity of 60% ± 5% at the Core Facility and Service Center for the School of Life Sciences and Biotechnology, Shanghai Jiao Tong University.

Seeds were surface sterilized in 75% (v/v) ethanol for 1 min, followed by 2% (v/v) sodium hypochlorite for 10 min, rinsed 5 times with sterile water, and stratified at 4°C in darkness for 3 days. Sterilized seeds were germinated on Murashige and Skoog (MS) medium containing 1% (w/v) sucrose and 0.8% (w/v) agar (pH 5.8). Upon emergence of the first true leaf, seedlings were transplanted into pots containing a 10:10:1 (v/v/v) mixture of nutrient soil, vermiculite, and perlite.

We simultaneously cultivated and measured Arabidopsis wild-type Col and nine core CLV–WUS pathway mutants in 2022 and 2025. In 2022, we analyzed the length, width, and area of individual rosette leaves according to leaf emergence order. In 2025, measurements were expanded to include biomass, Ap, chlorophyll fluorescence, CO₂ assimilation rate per unit leaf area, POV, leaf number, rosette diameter, and the length and width of the longest rosette leaf.

CRISPR-Cas9 vector construction and plant transformation

Target-specific single-guide RNAs (sgRNAs) for CLV1, CLV2, CLV3, RPK2, and WUS were designed using CRISPR-P 3.0 (http://crispr.hzau.edu.cn/CRISPR2/) to minimize off-target activity (sgRNA sequences and primers are listed in Supplemental Table 4). Each sgRNA was cloned into the BsaI-digested binary vector pHEE401E (Addgene 71287) by T4 DNA ligase-mediated ligation (Feng et al., 2013). Constructs were verified by Sanger sequencing (primer sequences are listed in Supplemental Table 4) and introduced into Agrobacterium tumefaciens strain GV3101 by freeze–thaw transformation.

Col plants were transformed using the floral dip method with minor modifications (Clough and Bent, 1998). Agrobacterium cultures were grown to an OD₆₀₀ of 0.6–0.8 and resuspended in infiltration medium (1/2 MS salts, 5% sucrose, 0.04% Silwet L-77, 200 μg l⁻¹ 6-BA); inflorescences of 4-week-old plants were then dipped for 30–60 s. Infiltrated plants were incubated in darkness for 24 h; a second dip was optionally performed to increase transformation efficiency. T₀ seeds were surface sterilized and selected on MS medium containing 50 mg l⁻¹ hygromycin B. Resistant T₀ seedlings were transferred to soil (nutrient soil:vermiculite:perlite, 10:10:1 [v/v/v]) and self-pollinated, and homozygous T₁ or T₂ lines were identified by segregation analysis and genotyping.

Leaf emergence number

Visible leaf emergence number was recorded for each genotype 2, 4, 7, 10, 13, 16, 19, and 21 days after transplantation, starting from the day of transplantation until the appearance of floral buds. Only leaves with a lamina length ≥ 3 mm were counted.

Green leaf number

The number of photosynthetically active leaves (excluding senescent leaves) was quantified 19 days after transplantation.

Individual rosette leaf morphology

During vegetative growth, rosette leaves were sequentially harvested according to emergence order and photographed on a flat surface. Leaf length (apex to petiole base), maximum width (perpendicular to the midrib), and blade area were measured using ImageJ (v.1.53). Total rosette leaf area was calculated using the Python cv2 package.

Rosette diameter and longest-leaf measurements

During vegetative growth, top-view photographs of Col and mutant plants were acquired 19 days after transplantation. Using ImageJ software, we measured the (1) longest-leaf length (from apex to petiole base), (2) longest-leaf maximum width (perpendicular to the midrib), and (3) rosette diameter (maximum distance between opposing leaf tips).

Whole-plant photosynthetic rate

Ap was measured using a custom pot chamber (20 × 20 × 15 cm) designed specifically for Arabidopsis (Figure 1A and Supplemental Figure 1C). The transparent chamber fully enclosed a single potted plant and was sealed with a water-filled trough. It was coupled to an LI-6400XT portable photosynthesis system (LI-COR Biosciences, Lincoln, NE, USA) to form a closed gas exchange loop. Illumination was provided by an integrated LED panel delivering 240 μmol m⁻² s⁻¹ at the plant level. Measurements were conducted under controlled conditions: 400 μmol mol⁻¹ CO₂, 22°C, and 240 μmol m⁻² s⁻¹ photosynthetic photon flux density (PPFD) (saturating for Arabidopsis rosettes). After stabilization (∼3–5 min), Ap was recorded as nmol CO₂ plant⁻¹ s⁻¹. Total projected leaf area was determined immediately after each measurement from top-view images using the Python cv2 package. The CO₂ assimilation rate per unit leaf area (A) was then calculated as A = Ap/total leaf area (μmol CO₂ m⁻² s⁻¹). The Python source code for whole-plant leaf-area segmentation and calculation is publicly available on GitHub (https://github.com/songqingfeng/AtPOVcalculator).

3D point cloud analysis pipeline for A. thaliana

We developed a computational pipeline for analyzing 3D point clouds of A. thaliana, which comprised six sequential processing modules. Each module was implemented as a standalone Python script using standard libraries (Open3D, NumPy, and plyfile). Initially, the raw Polygon File Format containing point coordinates, normals, and RGB color data, was cropped to a defined bounding box to isolate the plant (Box_process.py). Vegetative points were then segmented using the Excess Green index (ExG = 2G – R – B) with a fixed threshold of 40 (Color_exg.py). The resulting point cloud was denoised by removing statistical outliers based on neighborhood distances (Denoise_process.py) and uniformly downsampled via voxel grid simplification (Voxel_downsample.py). Individual plant instances were identified and isolated using Density-Based Spatial Clustering of Applications with Noise algorithm, followed by post-filtering to retain only substantial clusters (DBS_process.py). Finally, POV was calculated from a voxel-downsampled representation (compute_POV.py). The point cloud data and Python source code are publicly available on GitHub (https://github.com/songqingfeng/AtPOVcalculator).

Determination of plant occupation volume (POV)

Inspired by the established relationship between POV and plant photosynthesis in maize (Liu et al., 2021), we developed a robot-assisted 3D imaging system tailored for Arabidopsis (Figure 1B and Supplemental Video 1). The system consists of a six-axis robotic arm, a high-resolution RGB camera, and a motorized turntable. Each plant was imaged from 25 viewpoints (15° angular intervals), with 6 images acquired per viewpoint, yielding 150 images per scan. The automated 3D imaging process requires 2 min per plant. Dense 3D point clouds were reconstructed using structure-from-motion pipelines previously validated in plants (Song et al., 2023). Background noise and pot/soil points were removed using a custom Python pipeline based on the Python Open 3D package. Cleaned point clouds were voxelized at 1-mm resolution, and POV was calculated as the total volume of occupied voxels following a previously reported method (Liu et al., 2021). This approach provides robust, non-destructive quantification of 3D plant space occupation in Arabidopsis. The Python source code for POV calculation is publicly available on GitHub (https://github.com/songqingfeng/AtPOVcalculator).

Automated Mask3D-based 3D segmentation and refined trait extraction

An automated pipeline leveraging the Mask Transformer for 3D Instance Segmentation framework (Schult et al., 2022) was developed for joint semantic-instance segmentation of standardized Arabidopsis 3D point clouds. Subsequently, refined architectural traits, including leaf inclination angles and curvature, were quantified via graph-based spatial analysis using SciPy (Dijkstra and Delaunay) and scikit-learn (NearestNeighbors) within a high-throughput batch-processing workflow. The Python source code is publicly available on GitHub (https://github.com/songqingfeng/AtPOVcalculator).

Chlorophyll fluorescence measurement

Chlorophyll fluorescence was measured using a fluorescence dynamic monitor (FDM-S1, imaging area 35 × 40 cm). Plants were dark-adapted for ≥ 20 min, after which 10 potted plants of different genotypes were measured within the imaging chamber. The measurement protocol consisted of two phases. First, minimal fluorescence (Fo) was recorded under weak measuring light (0.27 μmol m⁻² s⁻¹ PPFD), followed by a saturating pulse (4800 μmol m⁻² s⁻¹ PPFD, 800 ms duration) to obtain maximum fluorescence yield (Fm). The maximum quantum efficiency of photosystem II was calculated as Fv/Fm = (Fm − Fo)/Fm. Following the dark-adapted measurements, a fast light curve protocol was initiated to assess light-adapted fluorescence parameters. Plants were sequentially exposed to actinic light intensities of 40.9, 89.2, 113.4, 185.9, and 234.2 μmol m⁻² s⁻¹. After 30 s at each step to reach a stable fluorescence signal, steady-state fluorescence yield (Ft) was recorded, immediately followed by a saturating pulse to determine light-adapted maximal fluorescence (Fm′). Effective photosystem II quantum yield (Y(II) = [Fm′ − Ft]/Fm′) and non-photochemical quenching (NPQ = (Fm − Fm′)/Fm′) were calculated automatically by the instrument software. Y(II) and NPQ values obtained at 113.4 μmol m⁻² s⁻¹ (closest to growth light intensity) were used for statistical comparison across genotypes.

Biomass determination

Plants were harvested 32 days after transplantation. Whole rosettes were oven dried at 110°C for 1 h to deactivate enzymes and then dried to constant weight at 80°C, after which aboveground dry biomass was recorded.

Data processing and statistical analysis

Data were processed and analyzed using R software (v.4.3.1) and visualized with GraphPad Prism v.8.0 (https://www.graphpad.com/). Quantitative traits, including whole-plant photosynthetic rate, biomass, POV, leaf number, leaf dimensions, and rosette diameter, are reported as means ± SD (n ≥ 3 biological replicates). Statistical comparisons were performed using two-tailed Student’s t-tests, with each mutant individually compared with Col. Asterisks indicate significant differences (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001). Pearson correlation coefficients and associated p values were calculated with the R package corrplot and visualized in GraphPad Prism. All figures were prepared in GraphPad Prism 8.0 to ensure publication-quality formatting.

Acknowledgments

We thank Dr. Jie Xu (Core Facility and Technical Service Center, School of Life Sciences and Biotechnology, Shanghai Jiao Tong University) for assistance with plant cultivation, Prof. Xiaoping Gou (Lanzhou University) for providing the cik123 and cik234 mutants, and Millet Hill Biotech for technical support. This work was supported by grants from the National Key Research and Development Program of China (2022YFD1200102 to Z.Y.), the Innovation Project of Shanghai Agricultural Science and Technology (V2024004 to Z.Y.), the National Natural Science Foundation of China (32470351 and 32170322 to Z.Y. and 32270428 to Q.S.), the Sino German Mobility Program (M-0141 to Z.Y.), the SMC Morningstar Young Scholarship of Shanghai Jiao Tong University (to Z.Y.), and the Strategic Priority Research Program of the Chinese Academy of Sciences (XDB0630000 to Q.S.). No conflict of interest is declared.

Author contributions

Q.S. and Z.Y. supervised the project. Y.S., P.W., and Q.W. performed most of the experiments. Y.S., Q.S., and Z.Y. analyzed the data and wrote the original draft. Z.Y. edited the final manuscript.

Published: May 20, 2026