Evidence of Graft Incompatibility and Rootstock Scion Interactions in Cacao

Abstract

This study sought to quantify and characterize diverse rootstock scion interactions in cacao around graft compatibility, disease resistance, nutrient use efficiency, vigor traits, and translocation of nonstructural carbohydrates. In total, 106 grafts were performed with three scion cultivars (Matina 1/6, Criollo 22, Pound 7) and nine diverse open-pollinated seedling populations (BYNC, EQX 3348, GNV 360, IMC 14, PA 107, SCA 6, T 294, T 384, T 484). We found evidence for both local and translocated graft incompatibility. Cross sections and Micro-XCT imaging revealed anatomical anomalies, including necrosis and cavitation at the junction and accumulation of starch in the rootstock directly below the graft junction. Scion genetics were a significant factor in explaining differences in graft take, and graft take varied from 47% (Criollo 22) to 72% (Pound 7). Rootstock and scion identity both accounted for differences in survival over the course of the 30-month greenhouse study, with a low of 28.5% survival of Criollo 22 scions and a high of 72% for Pound 7 scions. Survival by rootstocks varied from 14.3% on GNV 360 to 100% survival on T 294 rootstock. A positive correlation of 0.34 (p = 0.098) was found between the graft success of different rootstock–scion combinations and their kinship coefficient, suggesting that relatedness of stock and scion could be a driver of incompatibility. Significant rootstock–scion effects were also observed for nutrient use efficiency, plant vigor, and resistance to Phytophthora palmivora. These findings, while preliminary in nature, highlight the potential of rootstock breeding to improve plant nutrition, resilience, and disease resistance in cacao.

1. Introduction

Theobroma cacao L., the cacao tree and the products made from it, has been cultivated and consumed by humans for over 5000 years [1]. It continues to be an economically important crop that provides cocoa powder, cocoa butter, and cacao pulp for chocolate, confections, and cosmetics worldwide. Globally, cacao is cultivated across an estimated 11.5 million ha, predominantly by smallholder farmers on less than 5 ha of land [2]. Despite high global demand for products of the cacao tree, yields have remained lower than the demonstrated potential of improved planting material in many regions. Today, the availability of planting materials from elite breeding selections offers more options to farmers in different geographies with improved yield, disease resistance, and stability of performance under adverse conditions [3,4,5,6]. Currently it is estimated that 75% of farmers are planting material of unknown provenance [7]. Although clonal cacao still comprises a small percentage of production worldwide, access to improved planting material has accounted for yield intensification in countries such as Ecuador, Peru, and Brazil [6], offering opportunities to improve yields at the farm level and reduce the deforestation associated with expansion of cacao production.

Improved varieties of cacao are replicated through grafting onto open-pollinated seedling rootstock. In contrast to the advancements made in scion breeding, development of improved rootstock varieties and characterization of rootstock effects in cacao are limited. Recent studies have demonstrated that cacao rootstocks could perform an important role in regulating cadmium uptake into the scion [8], explaining variance in total pods [9] and affecting growth and stomatal conductance under water deficit conditions [10]. A challenge impeding progress in the understanding of rootstock–scion interactions and the advancement of rootstock improvement is the currently prevailing utilization of segregating open-pollinated seedling populations for rootstock, which introduces variability into the traits being studied [11]. The dimorphic nature of cacao growth and recalcitrance to tissue culture methods have been additional obstacles to the widespread use of clonal rootstock [11,12].

Clonal rootstock is not yet utilized commercially for cacao, and farmers rely instead on open-pollinated populations of locally available varieties, although the use of segregating populations may lead to increased variance and heterogeneity in yield and performance [13,14].

Systematic improvement of rootstock and validation of widespread rootstock–scion interactions have been difficult in part because of the diversity of planting material used. In Brazil, open-pollinated seedlings from the clones TSH 1118, Comum, and CEPEC 2002 are utilized due to their resistance to ceratocystis wilt disease (Ceratocystis cacaofunesta Engelbrecht and Harrington) [15]. In Ecuador, EET 399, EET 400, and IMC 67 are recommended, and the latter two have demonstrated better performance under water deficit conditions [10,16]. In Puerto Rico, EET 400 is also recommended for improving performance and yield stability [17], although it has been shown through molecular data to be distinct from the Ecuadorian EET 400 [18]. IMC 67, the most commonly utilized rootstock in Colombia, Peru, and Ecuador, has shown improved vigor, improved resistance to ceratocystis wilt disease, and only moderate susceptibility to several diseases including witches’ broom disease, although it is susceptible to Phytophthora palmivora [19,20]. Rootstock improvement has lagged considerably behind scion breeding efforts for many reasons, including geographic variation, high levels of genotype-by-environment interactions that affect phenotyping, a high prevalence of off-types in collections, and the limited geographic distribution of selections used as rootstock [21].

There is thus a critical need for better understanding of rootstock-modulated traits and rootstock–scion interactions in cacao to inform future breeding programs for elite rootstock varieties. This greenhouse study explored the hypothesis that rootstock regulates variation in scion graft success, survival, vigor, foliar nutrition, and susceptibility to Phytophthora palmivora.

2. Materials and Methods

2.1. Planting Materials

Three scion cultivars grafted onto nine open-pollinated rootstock populations with four replicates were evaluated for 30 months in greenhouses in Miami, FL, USA. Open-pollinated seedlings from nine maternal clones in the germplasm collection at the USDA-SHRS in Miami, FL, USA were used for rootstock in this study. Rootstocks were selected from nine diverse open-pollinated seedling populations from the maternal clones BYNC, EQX 3348, GNV 360, IMC 14, PA 107, SCA 6, T 294, T 384, and T 484. Based on STRUCTURE [22] results run across a common set of 21,657 SNP markers, the materials used in this study capture sufficient diversity based on the 10 genetic ancestry groups [23] (Figure 1). To our knowledge, none of the populations tested are used as commercial rootstock [24]. Once germinated, seedlings were grown in 3.8 L plastic polybags in an autoclaved woody mix composed of equal parts redwood:peat:sand:perlite (1:1:1:1). The plants were fertilized biweekly with modified Hoagland Solution and supplemented with ~1 teaspoon slow-release Nutricote^TM (Profile Products, Buffalo Grove, IL, USA) 18-6-8 granular fertilizer every 6 months. The plants were hand-watered based on their irrigation needs, ranging from daily to twice a week. Trees, once established, were transplanted into 26 L pots and maintained on a bench with 30.5 cm spacing between their centers in the Mars Breeding Greenhouse of the United States Department of Agriculture-Subtropical Horticultural Research Station in Miami, FL, USA (Figure 2A). As greenhouse space was limited, the plants were pruned quarterly to maintain bench spacing and reduce canopy interference. The effect of canopy pruning, as well as the constraints of the container size, may limit the inference of some results, particularly morphological traits, and field validation is recommended. Total shading was 71% (15% shade provided by the polycarbonate greenhouse structure and 56% shade provided by ALUMINET^® shade screen (Ecologic Technologies Inc., Myrtle Beach, South Carolina, USA)), the plants experienced average daily PAR values ranging from 80 to 400 PPFD, the annual temperature range was 18–32 °C, and the annual humidity range was from 60% to 90%.

2.2. Grafting

Scions were selected from three cacao cultivars in the germplasm collection: Criollo 22, Matina 1/6, and Pound 7. The three scions were selected as representatives of three different genetic ancestry groups in cacao––Criollo (Criollo 22), Amelonado (Matina 1/6) and Nanay/admixed (Pound 7)––which together are broadly representative of cultivated material prevalent in different regions. Criollo cultivars are widespread in Mesoamerica, Amelonado backgrounds predominate in tropical West Africa, and Pound 7 represents upper Amazon material cultivated widely in Latin America and the Caribbean. Additionally, they represent complementary genomic diversity, as Matina 1/6 and Criollo 22 are both over 97% homozygous, with ~30% co-ancestry [25,26]. Pound 7, by contrast, represents a highly heterozygous, admixed accession [27]. Eighteen-month-old seedlings were cleft-grafted 46 cm from the base by a single grafter. The grafting height was selected to ensure that the diameter of scions matched the rootstock. Grafting was performed in three successive rounds, whereby plants were regrafted after 2 months if the graft failed until four replicates of the rootstock–scion combination were achieved. The lowest number of grafts made per treatment was four, while the highest was eleven (Table S4), and the mean number of grafts needed to reach four replicates was 5.6. In total, 150 grafts were performed to produce the 106 surviving plants used for this study.

2.3. Morphological Measurements

Survival and vigor measurements were assessed at multiple time points during the experiment to establish growth rates and assess temporal trends. The rootstock basal diameter was measured with digital calipers at the base of the tree at three time points: 16, 22, and 30 months after grafting. The rootstock diameter 8 cm below the graft junction was taken 4, 16, and 30 months after grafting and scion diameter was measured 5 cm above the graft junction 16 and 30 months after grafting. Flush length, defined here as the length of six internodes from the most recently developed dormant shoots, was taken from the average of three measurements per tree 30 months after grafting. At this time, the plants were harvested and separated into roots, trunk (up to the graft junction), branches (above graft junction), and leaves (Figure 2B). Samples were air-dried for 3 weeks and weighed for biomass measurements.

2.4. Graft Junction Imaging

Imaging of the graft junction of individual trees was performed at the Center for Molecular and Genomic Imaging (University of California, Davis) by X-ray CT using a MicroXCT-200 specimen CT scanner (Zeiss, Oberkochen, Germany). The samples were received freshly cut and capped with parafilm to slow desiccation and glued vertically to a sample holder to orient the apical-basal axis parallel with the imaging cylinder. Scans were acquired at 29 µm. The CT scanner has a variable X-ray source capable of a voltage range of 20–90 kV with 1–8 W of power. Scan parameters were adjusted based on the manufacturer’s recommended guidelines. First, the source and detector distances were adjusted based on sample size and the optimal field of view for the given region of interest. An Xradia (Zeiss, Oberkochen, Germany) LE2 filter was applied. The optimal voltage and power settings were determined for optimal contrast; 40 kV and 200 μA were used for all scans. We obtained 16,000 projections (camera binning 2) over 360 degrees with 3.5 s per projection. The tomographic images were reconstructed with appropriate center shifts and beam-hardening parameter values to minimize image artifacts. A smoothing filter of kernel size 0.5 was applied during reconstruction. Images were reconstructed into 16-bit values, visualized in Dragonfly 3D World Ver. 2024.1.

In order to visualize graft junction anatomy and starch accumulation in the rootstock and scion, cross sections were cut 8 cm above and below the graft junction and cut longitudinally on a band saw and sanded with 220 grit sandpaper. Cross sections were stained with Lugol’s solution consisting of 10 g potassium iodide (KI) and 5 g iodine dissolved in 100 mL distilled water. Stain was applied for 12 h, and samples were analyzed visually for the distribution of starch in the tissues.

2.5. Nutrient and Carbohydrate Quantification

Five mature leaves from around the canopy were submitted to Waypoint Analytical (Richmond, VA, USA) for quantification of 13 micro- and macronutrients. Leaf samples submitted included samples taken 4 months after grafting for 74 plants, 16 months after grafting for 71 plants, and 30 months after grafting for 55 plants. The number of plants sampled at each time point reflects the availability of healthy leaves of the proper developmental stage.

Quantification of chlorophyll and starch content was conducted 30 months after grafting, using the third leaf of the most recently hardened flush on three branches per tree. Chlorophyll measurements were taken using an Apogee MC-100 Chlorophyll Concentration Meter (Apogee Instruments, Inc., Logan, UT, USA). Three sampling points were taken per tree, in the center of the third leaf of the most recently matured flush. Starch samples were punched from the same point as the chlorophyll measurements. Leaf discs were collected using a 6.0 mm diameter biopsy hole punch and placed in 1 mL of ethanol. Samples were shipped to Davis, CA and stored at −20 °C until analysis. Starch was measured using the following protocol adapted from Warren, et al. (2015) and Teixeira et al. (2012) [28,29]. Leaf discs were removed from the collection tubes and placed in 2 mL SPEX SamplePrep 2303-MM1 microcentrifuge tubes loaded with garnet grinding beads (Cole-Parmer, Vernon Hills, IL, USA), to which 1 mL of fresh ethanol was added. After soaking overnight at 4 °C, the ethanol was discarded. Next, 1 mL of 8:2 methanol:water (v/v) was added, and the leaf discs were soaked overnight again at 4 °C. The supernatant was removed, and 1 mL of pure methanol was added. After soaking for 1 h in methanol at room temperature, the supernatant was discarded, and the leaf discs were dried in a SpeedVac (Eppendorf, Hamburg, Germany).

A solution of 5 U/mL amylase (Sigma A4551; Sigma-Aldrich, St. Louis, MO, USA) and 2.5 U/mL amyloglucosidase (Roche 11202367001; Roche, Basel, Switzerland) was prepared in 0.2 M acetate buffer (pH 6) with 200 mM CaCl₂ and 0.5 mM MgCl₂ [28]. Activities were calculated using the reported activities specific to the enzymes’ lot numbers. Next, 1 mL of the enzyme solution was added to each tube with cacao leaf discs. Leaf discs were pulverized in the buffer using a Precellys^® Evolution (Bertin Technologies, Montigny-le-Bretonneux, France) tissue homogenizer. The samples were incubated at 50 °C for 3 h and were mixed by inverting at the 1 and 2 h time points. Debris was then pelleted by centrifugation (5000× g, 10 min, room temperature). Then, 250 µL of the supernatant was mixed with 250 µL of dinitrosalicylic acid (DNSA, Thermo Scientific 156441000, Thermo Fisher Scientific, Waltham, MA, USA), prepared by dissolving, in 100 mL water, 1 g DNSA, 1.6 g sodium hydroxide, and 30 g sodium potassium tartrate [29]. The samples were then heated to 90 °C for 10 min on a heat block, cooled to room temperature, and then measured for absorbance at 540 nm.

To generate a standard curve for quantification, a 1 mg/mL starch (Thermo Scientific AAA1196136, Thermo Fisher Scientific, Waltham, MA, USA) solution was prepared in water (requires heating and sonication). The solution was aliquoted into 2 mL screw-cap tubes to deliver 300, 200, 100, 50, 25, 10, 5, 1, or 0 (blank) µg starch. These aliquots were dried to completion in a SpeedVac. Then, 1 mL of amylase/amyloglucosidase solution was added, and the standards were processed in parallel with the cacao leaf samples. The resulting standard curve was used for quantification:

y = 0.0024x + 0.1398 (R² = 0.9994; y = µg starch; x = Abs540 nm)

2.6. Statistical Analysis

Rootstock effects on morphological, nutrient, and carbohydrate traits were analyzed by ANOVA to determine differences between groups (Table S1), and a linear mixed model was used to determine the percent of variance explained by rootstock group membership. The linear mixed model was run using the package LME4 in R statistical software Ver. 4.4.1 [30,31].

Y = X ∗ β + Z ∗ µ + e

• X = Matrix of fixed effect coefficients
• Z = Matrix of random effect coefficients
• β = Scion cultivar
• µ = Rootstock family

A correlation matrix was also generated in R to explore relationships between traits and trends in the repeated measures (Table S2, Figure S1). Principal component analysis was performed on leaf nutrient and carbohydrate data from the 30-month time point using the prcomp R Core Package (Figure S2).

Kinship coefficients between scions and the maternal parent of each rootstock populations were generated using the package rrBLUP [32], with 21,657 SNP markers spaced throughout the genome. Maternal accessions were used rather than the individual segregating open-pollinated rootstock in the study, which did not have available sequence data.

2.7. Phytophthora Palmivora Inoculations

2.7.1. Rootstock Inoculations

This study aimed to determine whether susceptibility to Phytophthora is mediated by the rootstock—that is, whether the rootstock’s susceptibility to Phytophthora inoculation corresponds to the scion’s response when grafted onto rootstock populations with contrasting levels of resistance. Wounded stem inoculations were made on ungrafted 2- to 3-year-old seedlings from each rootstock group. Seedlings were derived from open-pollinated pods from eight of the nine cultivars and ranged from 84 to 124 cm in height and 1.3 to 2.8 cm in diameter. IMC 14 was excluded from inoculations on account of not having enough material for sufficient replication. Trees were inoculated as described in Puig et al. (2021) [33], by wounding with a 6 mm diameter cork borer (4 mm depth), and the excised tissue was replaced with 6 mm mycelial disks from 2- to 4-day old cultures of P. palmivora (isolate H33) from T. cacao in Hawaii [34]. Inoculation sites were wrapped with parafilm and kept in Percival growth chambers (25 °C; 12 h light/dark). Lesion diameters were measured 5 weeks after inoculation and used to calculate the lesion area using the following equation for calculating the area of an ellipse, where a is one-half the vertical diameter and b is one-half the horizontal diameter of the lesion:

area of ellipse (lesion) = π × a × b.

Each tree was inoculated in two different locations (Figure 3A) and averaged to produce a single lesion size per tree. The average lesion area among groups was analyzed using by non-parametric Kruskal–Wallis Test, using SAS Ver. 9.4 (SAS Institute, Cary, NC, USA). Each group had 3 to 16 replicates (seedlings), based on availability. To confirm P. palmivora as the cause of the symptoms, material was taken from a subset of the lesion, surface-disinfested, and plated on half-strength potato dextrose agar (PDA) (Sigma Chemical Co., St. Louis, MO, USA; 19.5 g PDA, 7.5 g agar, and 1 L distilled water). The re-isolated organisms were identified as P. palmivora based on morphology.

2.7.2. Scion Inoculations

To assess the influence of rootstock on the susceptibility of grafted trees to lesions caused by Phytophthora palmivora, inoculations were performed on Matina 1/6 scions grafted onto open-pollinated rootstocks of BYNC (n = 4) and SCA 6 (n = 3). (Figure 3A). These rootstocks were selected based on contrasting lesion size of the seedling populations from 2.5.1, and the availability of at least three replicates.. Inoculations were carried out as described above; however, due to the narrower branch diameters, a 2 mm diameter probe was used instead of a cork borer. Lesions were measured 5 weeks after inoculation, and mean lesion area was analyzed using the non-parametric Wilcoxon test in SAS 9.4.

3. Results

3.1. Graft Incompatibility

We used binary logistic regression to model the log-odds of graft take and survival as a function of rootstock, scion, and their interaction:

$$\log \left({\displaystyle \frac{p}{1-p}}\right)={\beta }_0+{\beta }_1 (\text{Rootstock})+{\beta }_2 (\text{Scion})+{\beta }_3 (\text{Rootstock}\times \text{Scion})+\mathsf{\epsilon }$$

where p is the probability of graft establishment, β₀ is the intercept, and β₁, β₂, and β₃ are the coefficients associated with the main and interaction effects of the categorical predictors. Based on 106 graft events, scion and specific rootstock–scion combinations were significant predictors of graft take (

Table 1. Binary logistic regression model for factors influencing graft take and survival.

Trait	Variables	ChiSq	DF	Pr (>ChiSq)
Graft Take	Rootstock	10.92	8	0.21
	Scion	7.79	2	0.02 *
	Rootstock × Scion	28.19	16	0.03 *
Survival	Rootstock	19.39	8	0.01 *
	Scion	11.8	2	0.003 **
	Rootstock × Scion	14.99	16	0.525

* Statistical significance at a 5% threshold. ** Statistical significance at a 1% threshold.

). Out of the three scions tested, Criollo 22 had the lowest graft success across rootstocks (47% successful from 53 attempted), while Pound 7 had the highest graft take (72% successful from 44 attempted). Graft success of rootstock populations ranged from a low of 39% success for T 384 (of 23 attempted grafts) to 84.6% success for PA 107 (of 13 attempted grafts) (Figure 4A).

Trends were evident in the graft success of specific rootstock–scion combinations. For instance, rootstock populations from SCA 6, PA 107, and EQX 3348 showed the least specificity and highest consistency in graft take across all scion cultivars, with coefficients of variation ranging from 0% to 26% (Table S2). On the other hand, rootstock populations T 384 and BYNC showed high specificity, with coefficients of variation of 53% and 84%, respectively. Notably, both had 25% graft success with Criollo 22 as a scion but 100% success with Pound 7 as a scion (Figure 4A).

The kinship coefficients between scion and rootstock maternal cultivar ranged from −0.51 (Matina 1/6 on SCA 6 OP rootstock) to 0.16 (Matina 1/6 on T 484), with an average of −0.09 (Table S3). A Pearson correlation coefficient of 0.324 (p = 0.098) was observed between the kinship coefficient and the graft success percentage across 27 rootstock–scion pairs. When analyzed using linear regression, the kinship coefficient explained approximately 11% of the variance in graft compatibility (Figure 4B).

Two scions grafted to the same rootstock were selected for comparison based on contrasting graft success rates: Criollo 22 grafted onto T 294 (100% success) and Pound 7 grafted onto T 294 (56% success) (Figure 4A). The kinship coefficients for these pairs were 0.023 and −0.16, respectively. All grafts exhibited a necrotic region marking the cut site of the stem on the rootstock where grafting occurred (Figure 5A–I). However, the incompatible combination (Figure 5D–F) developed a pronounced necrotic layer extending vertically along the graft union and outward to the cambium. In contrast, the compatible combination showed a more gradual transition between scion and rootstock tissues, with lighter discoloration and less distinct necrosis at the union interface (Figure 5A–C).

3.2. Survival

Out of 106 initial grafts made to generate the material for this study, 90 plants survived for 16 months after grafting, 70 survived 22 months after grafting, and after 30 months, 52 plants remained (Figure 6). Survival of Pound 7 grafts remained stable after the second year, whereas Criollo 22 and Matina 1/6 scions continued to decline in the third year after grafting. Only 28.5% of Criollo 22 grafts survived for 3 years, compared with 48.6% of Matina 1/6 and 72% of Pound 7 grafts. Survival by rootstock varied widely, with the lowest rates seen on GNV 360 across all tested scion cultivars (14.3%) and the highest on T 294 (100%).

Many of the plants that died in the second and third year of the study showed symptoms consistent with delayed incompatibility responses, including shoot tip dieback and leaf drop following mechanical and abiotic stresses from pruning and water deficit. Stained cross sections from a sample of plants in this study revealed a pattern of starch accumulation in the rootstock of some samples, but not the scion (Figure 5G–I). This could indicate translocated incompatibility, with the rootstock acting as a sink and accumulating starch due to impaired translocation to the scion. Notably, this is in contrast to the starch accumulation patterns described in incompatible Citrus grafts, where starch content accumulates on the scion side of the graft, while the rootstock displays ribboning and necrosis [35]. Potential phloem and xylem degradation at the graft junction could disrupt phloem loading, as well as result in xylem embolism. Both of these vascular disruptions could impact the recovery of plants following stress.

3.3. Morphological Traits

3.3.1. Rootstock Diameter

The rootstock diameter at the base of the plant differed significantly by rootstock population at all time points measured (p < 0.05) (

Table 2. ANOVA summary of biomass traits grouped according to rootstock and scion.

Rootstock										Scion
	DF	SS	MS	F	p	DF	SS	MS	F	p
Diameter below graft, 4 MAG	8	251.6	31.5	5.9	3.96 × 10−6 ***	2	23.1	11.6	1.6	0.208
Diameter below graft, 16 MAG	8	139.8	17.4	3.1	0.00459 **	2	15.7	7.8	1.2	0.318
Diameter below graft, 30 MAG	8	100.4	12.6	1.9	0.0925	2	32.8	16.4	2.2	0.117
Δ Diameter below graft, 4–30 MAG	8	72.56	9.1	1.5	0.189	2	77.5	38.7	7.5	0.00152 **
Basal diameter, 16 MAG	8	288.9	36.1	5.0	0.0036 **	2	14.4	7.2	0.7	0.49
Basal Diameter, 30 MAG	8	227.5	28.4	2.4	0.0322 *	2	16.3	8.2	0.6	0.579
Δ Basal diameter, 16–30 MAG	8	192.3	24.0	2.6	0.0214 *	2	36.5	18.23	1.6	0.213
Scion diameter, 16 MAG	8	79.8	10	1.3	0.25	2	20.3	10.2	1.3	0.276
Scion diameter, 30 MAG	8	36.5	4.6	0.5	0.823	2	21.4	10.7	1.4	0.264
Δ Scion diameter, 16–30 MAG	8	77.6	9.704	1.2	0.328	2	26.8	13.4	1.6	0.206
Flush length, 30 MAG	8	80	10	0.7	0.692	2	79.1	39.5	2.9	0.0593
Leaf dry weight, 30 MAG	8	5147.9	643.5	1.2	0.302	2	114	56.9	0.1	0.904
Branch dry weight, 30 MAG	8	7816.7	977.1	0.7	0.696	2	5648	2824	2.2	0.122
Trunk dry weight 1, 30 MAG	8	39,421.8	4927.7	2.3	0.0373 *	2	1395	697.7	0.3	0.769
Root dry weight, 30 MAG	8	61,920	7740	1.3	0.255	2	14,814	7407	1.2	0.304
Whole plant dry weight, 30 MAG	8	233,617.4	29,202.2	1.1	0.397	2	28,546	14,273	0.5	0.603

* p < 0.05; ** p < 0.01; *** p < 0.001; MAG—months after grafting; ¹—the trunk was part of the rootstock.

and Table S1). PA 107 was the most vigorous rootstock at all time points, while EQX 3348 consistently ranked last (Table S1). While rootstock basal diameter was the only morphological trait to show significant differences into the third year of the study, the average basal diameters of the five most vigorous rootstocks at 30 months were all within 1.05 mm of each other (Table S1), indicating that the pot size had likely become limiting with respect to diameter growth, even for this trait.

Rootstock diameter below the graft was measured at time points 4, 16, and 30 months after grafting. The measurement was taken 8 cm below the top of the graft to avoid capturing the graft junction region. Across these time points, the correlation with rootstock base diameter decreased from 0.72 (4 months to 16 months) to 0.39 (4 months to 30 months) (Table S2, Figure S1). Rootstock effects on this trait were significant at 4 months and 16 months, but by month 30 after grafting, this trait differed significantly by scion and not rootstock (Table S1). Cumulative rootstock diameter growth below the graft was calculated by taking the difference between the 4-month and 30-month measurements. While there was no significant difference in cumulative growth among rootstock groups, there was a strong significant difference in rootstock growth based on scion cultivar. Rootstocks grafted with Criollo 22 had the largest growth increment below the graft, and those grafted with Pound 7 had the lowest (Figure 7). This was in contrast with the measure of diameter at the base of the trunk, which was influenced only by rootstock population. Notably, scion diameters above the graft at the two time points, as well as the change in scion diameter from 16 to 30 months, did not differ among scion cultivars (Table 2 and Table S1).

3.3.2. Scion Diameter

There was no effect of either rootstock or scion on differences in the diameter of scion at any time point. A Pearson correlation analysis showed a strong positive correlation between the basal diameter and scion diameter (r = 0.56), but where there were multiple branches following grafting, the correlation dropped. Scions with a single branch at 12 months had an average diameter 10.6 mm; with the addition of a second branch the diameter of the first decreased to 9.77 mm, and with a third, to 8.9 mm (Table S1). There was no change in vigor for the second branch with or without a third shoot (7.78 vs. 7.78 mm), suggesting that competition is strongest between the first branch and subsequent shoots in the first year after grafting. A strong correlation (r = 0.93) was found between scion diameter at 22 months and rootstock basal diameter at 12 months (Table S2, Figure S1).

3.3.3. Flush Length

The rootstock population did not have any significant effect on flush length at 30 months, but the scion cultivar was nearly significant (p = 0.054) (Table S2, Figure S1). Flush length did not correlate strongly with any of the plant vigor/growth data, except for a negative correlation with the diameter of the second shoot diameter (−0.46) at 30 months after grafting (Table S2).

3.3.4. Plant Biomass

At the time of harvest, trunk weight was the only biomass trait that differed by rootstock or scion, while there were no significant differences between rootstocks or scions for branch dry weight, leaf dry weight, root dry weight, or total plant biomass. If there was in fact an impact of pot size limiting treatment differences after 30 months, it is notable that the trunk diameter at the base and the trunk dry weight still differed significantly among populations after 30 months of container grown conditions (Table 2 and Table S1).

At harvest, the basal diameter of the rootstock was the strongest predictor of the scion traits dry leaf weight (r = 0.57), scion biomass (r = 0.64), stem weight (r = 0.78), rootstock biomass (r = 0.66), and whole plant biomass (r = 0.68). The diameter of the primary scion branch had the strongest correlation with branch weight (r = 0.74), root weight (r = 0.64), and scion biomass (r = 0.64). Notably, the trunk weight at 30 months had a strong correlation with base diameter measurements from 12 months (r = 0.55), demonstrating the utility of that trait as an early marker for later biomass and vigor (Table S2, Figure S1).

Linear mixed models were used to determine position effects on biomass measures. Bench row, which ran parallel to the greenhouse wall, explained 22% of the variance in total leaf dry weight when included as a random effect. This could be capturing an edge effect experienced by the exterior trees on the bench. No other biomass component aside from leaf dry weight was shown to be influenced by position.

The influence of rootstock on all morphological traits appeared to decline over the course of the study, according to a decreasing trend in the percentage of variation explained by rootstock (Figure 8). This decrease may be due to constraints on plant growth imposed by container size, an observation that is highly relevant for future long-term studies exploring biomass traits under greenhouse conditions. Stem and diameter traits showed the most influence from the rootstock.

3.3.5. Graft Junction Scans

CT scan renderings of a healthy 3-year-old grafted plant were analyzed to explore the site of the graft junction. Transverse sections revealed considerable splitting and cavitation from the rootstock pith, as well as fissures at the site of the graft junction just below the cambium (Figure 9). In compatible grafts, a callous bridge forms at the interface of the graft and heals the necrotic barrier [36]. These fissures demonstrate that, even though the cambium has reconnected all the way around the graft, the underlying rootstock and scion tissues were not fully fused even in a grafted tree with no apparent symptoms of incompatibility. This provides insights into putative factors influencing the health, physiological function, and survival of grafted trees and makes a case for further study of graft junction anatomy as a marker of graft incompatibility for cacao.

3.4. Nutrient and Carbohydrate Analysis

3.4.1. Foliar Nutrient Analysis

Four months after grafting, rootstock had the strongest influence on foliar nutrient content, as evidenced by the percentage of variance explained ranging from 0% (sodium) to 65.4% (aluminum) (

Table 3. ANOVA summary of nutrient traits grouped according to rootstock and scion.

Rootstock										Scion
	DF	SS	MS	F	p	DF	SS	MS	F	p
Nitrogen %—4 MAG	8	4706	588.3	2.89	0.00628 **	2	1538	768.9	3.5	0.0354 *
Nitrogen %—24 MAG	8	0.835	0.10435	1.089	0.383	2	0.041	0.02	0.204	0.816
Nitrogen %—30 MAG	7	1.192	0.1703	1.11	0.372	2	0.461	0.23	1.514	0.23
Sulfur %—4 MAG	8	0.0307	0.0038	1.563	0.154	2	0.00812	0.004	1.582	0.213
Sulfur %—24 MAG	8	0.1309	0.0163	0.922	0.505	2	0.0214	0.01068	0.6	0.552
Sulfur %—30 MAG	7	0.02021	0.0028	0.958	0.474	2	0.0169	0.008469	3.058	0.0562.
Phosphorus %—4 MAG	8	0.02715	0.003394	3.064	0.00553 **	2	0.00852	0.00426	3.332	0.0413 *
Phosphorus %—24 MAG	8	0.01712	0.002139	0.757	0.641	2	0.00228	0.001138	0.407	0.667
Phosphorus %—30 MAG	7	0.01397	0.001996	0.955	0.475	2	0.01137	0.005687	2.953	0.0617.
Potassium %—4 MAG	8	1.93	0.2418	2.459	0.0218 *	2	0.043	0.0214	0.183	0.833
Potassium %—24 MAG	8	3.438	0.4298	1.531	0.165	2	0.195	0.0974	0.32	0.727
Potassium %—30 MAG	7	0.37	0.05290	0.609	0.746	2	0.229	0.1145	1.417	0.252
Magnesium %—4 MAG	8	0.6075	0.07593	2.87	0.00991 **	2	0.0855	0.04277	1.329	0.271
Magnesium %—24 MAG	8	0.315	0.03938	1.335	0.244	2	0.0384	0.01922	0.62	0.541
Magnesium %—30 MAG	7	0.2792	0.03989	1.591	0.164	2	0.0459	0.02293	0.839	0.438
Calcium %—4 MAG	8	2.310	0.28874	5.023	7.36 × 10−5 ***	2	0.086	0.04295	0.509	0.603
Calcium %—24 MAG	8	1.741	0.2177	1.753	0.104	2	0.268	0.134	0.992	0.376
Calcium %—30 MAG	7	0.635	0.09067	1.025	0.428	2	0.14	0.07003	0.782	0.463
Sodium %—4 MAG	8	0.000176	2.2 × 10−5	0.631	0.749	2	0.00033	0.0001682	5.685	0.00516 **
Sodium %—24 MAG	8	0.01689	0.002	2.34	0.0291	2	0.00123	0.0006147	0.582	0.561
Sodium %—30 MAG	7	0.001799	0.000257	1.49	0.196	2	0.001466	0.0007329	4.539	0.0156 *
Boron PPM—4 MAG	8	8049	1006.1	8.227	1.38 × 10−7 ***	2	1683	841.3	4.149	0.0198 *
Boron PPM—24 MAG	8	845	105.6	0.945	0.487	2	46	22.85	0.201	0.818
Boron PPM—30 MAG	7	1380	197.1	1.196	0.326	2	447	223.3	1.336	0.272
Zinc PPM—4 MAG	8	82,281	10,285	6.567	3.15 × 10−6 ***	2	5008	2504	0.987	0.378
Zinc PPM—24 MAG	8	9478	1184.7	1.81	0.0924	2	1164	582	0.808	0.45
Zinc PPM—30 MAG	7	3466	495.1	0.667	0.699	2	193	96.5	0.132	0.877
Manganese PPM—4 MAG	8	474,670	59,334	3.497	0.00208 **	2	18,489	9244	0.42	0.659
Manganese PPM—24 MAG	8	572,322	71,540	2.286	0.0327 *	2	38,372	19,186	0.526	0.593
Manganese PPM—30 MAG	7	38,477	5497	0.726	0.651	2	10,075	5038	0.638	0.51
Iron PPM—4 MAG	8	190,087	23,761	17.85	1.13 × 10−13 ***	2	3561	1781	0.459	0.634
Iron PPM—24 MAG	8	169,223	21,153	4.513	0.000241 ***	2	8802	4401	0.661	0.52
Iron PPM—30 MAG	7	26,433	3776	1.646	0.148	2	2351	1176	0.46	0.634
Copper PPM—4 MAG	8	30.07	3.759	1.906	0.0744	2	10.31	5.155	2.471	0.0918
Copper PPM—24 MAG	8	188.6	23.580	3.933	0.000841 ***	2	1.9	0.962	0.117	0.89
Copper PPM—30 MAG	7	39.71	5.672	0.869	0.539	2	13.58	6.79	1.062	0.354
Aluminum PPM—4 MAG	8	1727	215.87	4.439	0.000258 ***	2	49	24.65	0.36	0.699
Aluminum PPM—24 MAG	8	1621	202.6	0634	0.746	2	290	144.9	0.466	0.629
Aluminum PPM—30 MAG	7	2728	389.7	7.947	3.62 × 10−6 ***	2	31	15.3	0.153	0.859
Chlorophyll—30 MAG	8	232,528	29,066	1.7	0.103	2	70,207	35,104	2.008	0.138
Starch—30 MAG	8	299.3	37.41	0.68	0.707	2	34.1	17.07	0.318	0.729

* p < 0.05; ** p < 0.01; *** p < 0.001; MAG—months after grafting.

). Over time, the influence of rootstock appears to decrease, ranging from 0% (sulfur, phosphorus, boron, aluminum) to 27.2% (iron), and by month 30 after grafting, it is below 10% for all traits except for aluminum, which jumps up again to 60% (Figure 10). This result supports the importance of rootstock in the health and nutrition of grafted plants in the nursery and for transplanting and establishment in the field, which usually occurs 4 months after grafting. The dynamic nature of metals including aluminum, copper, and iron demonstrate the particular relevance of rootstock in moderating heavy metal uptake.

The first two eigenvectors of the principal component analysis explained 46% of the variation in nutrients at 30 months after grafting based on rootstock; this was the same proportion explained as at 4 months, and slightly less than the 55% variance explained at 16 months after grafting [24] (Figure S2). The groupings of nutrients along the two axes followed the trends observed at 16 months after grafting, with phosphorus, nitrogen, and potassium having the highest loadings for the first eigenvector and all other nutrients falling along the second (Figure S2). There was not pronounced clustering according to rootstock membership, possibly reflecting the heterogeneous nutrient use efficiency of the segregating seedling rootstock that was used in the study and is typical of the industry.

3.4.2. Starch and Chlorophyll

Starch concentrations ranged from 2.98 to 36.5 µg per disc, with an average measurement of 11.4 µg. Starch measurements did not show strong correlations with any of the nutrients or treatments. The strongest correlation was seen with percent phosphorus (r = 0.32), followed by a negative correlation with chlorophyll (r = −0.3), then calcium and potassium (r = 0.28; r = 0.26). Chlorophyll, on the other hand, had a correlation of r = 0.74 with manganese, followed by calcium percent (r = 0.55) and boron (r = 0.48). According to the principal component analysis, starch, along with nitrogen, phosphorus, and potassium, drove the first eigenvector, which explained 30.6% of the variance, while chlorophyll fell along the second eigenvector in close association with manganese (Figure S2).

Patterns of starch accumulation in the stained cross sections of graft junctions illustrated that the pith is a major store of nonstructural carbohydrates (NSC) in the plant, particularly in the rootstock. Patterns of stain appear to show NSCs being translocated radially through tissues in ribbons of ray cells outwards toward the phloem and cambium (Figure 5). In some cases, they show an accumulation of starches below the graft junction (Figure 5G–I). In the sample of cross sectioned rootstocks, there were no examples of higher starch accumulation above the graft junction.

3.5. Phytophthora palmivora Lesions

3.5.1. Rootstock Lesions

The mean lesion area ranged from 1.8 to 26.4 cm², with high variability within groups, as evidenced by the high standard errors. No significant differences were found among groups based on non-parametric Kruskal–Wallis test (p = 0.072) (

Table 4. Mean lesion area on open-pollinated ungrafted seedling populations.

Population	N	Lesion Area (cm2)	Std Error
BYNC	3	1.8	0.74
PA 107	5	2.18	0.9
GNV 360	6	4.76	2.42
T 294	9	6.04	2.06
SCA 6	16	7.85	6.17
T 484	7	12.1	3.29
EQX 3348	8	15.77	7.5
T 384	5	26.4	18.98
p = 0.072

).

3.5.2. Scion Canker Lesion

The Matina 1/6 on SCA-6 trees developed larger lesions than Matina 1/6 on BYNC following inoculation with P. palmivora, with mean lesion areas of 0.72 (±0.13) and 2.12 (±0.75) cm², respectively (p = 0.034) (

Table 5. Mean lesion area on Matina 1/6 scion grafted onto open-pollinated rootstock of BYNC or SCA 6. Lesions were measured 5 weeks after inoculation with Phytophthora palmivora and found to be significantly different based on the Wilcoxon test (p = 0.034).

Scion	Rootstock	N	Lesion Area (cm2)	Std Error
Matina 1/6	BYNC	4	0.72 a	0.13
Matina 1/6	SCA 6	3	2.12 b	0.75

a and b indicate differences in mean lesion area were statistically significant (p = 0.034).

). By the end of the 5-week period, one of the Matina 1/6 on SCA-6 trees was starting to wilt (Figure 3C); however, no wilting was seen on any of the Matina 1/6 on BYNC trees.

4. Discussion

4.1. Graft Incompatibilities Increase with Genetic Distance

This study provides the first evidence in cacao that genotype-specific graft incompatibilities exist, and that the scion was a stronger predictor of graft compatibility than rootstock. A positive correlation of 0.34 between genetic distance and graft take supports that compatibility is influenced by the relatedness of scion and rootstock. Cacao as a crop has gone through relatively few breeding cycles and could be considered still largely undomesticated [11,28], which is also reflected in the provenance of the material used in this study. The genetic diversity of planting materials, reflecting recent wild ancestry, could support the observed intraspecific incompatibilities. It is unlikely that genetic proximity alone determines graft compatibility; however, a recent study of ten elite cultivars grafted to rootstock populations from their open-pollinated seedlings found rates of graft establishment that ranged from 5% to 66.7% [37].

An analysis of cut and stained graft junctions revealed prominent necrotic lines and browning around the graft junction interface of incompatible combinations. These could indicate the role of phenolics or oxidative stress in the incompatibility response between unrelated individuals. In various Prunus species, phenolic compounds, including flavonols, can negatively affect cell division and tissue development at the graft junction, leading to potential physical weakening and incompatibility [38]. Patterns in phenolic accumulation between compatible and incompatible combinations are sufficiently robust in woody plants that they have been proposed as early biochemical markers for detecting graft incompatibility in plums, chestnut, and grapevines [39,40,41,42,43]. With the prevalence of phenolic compounds in the woody tissues of cacao [44], this could be a promising avenue for screening for more broadly compatible rootstock in cacao as well. A recent study of Citrus analyzed the transcriptome in vascular tissues above and below the graft junction of compatible and incompatible graft combinations and found differentially expressed genes associated with oxidative stress and plant defense, similar to the pathogen-induced immune response, localized to the rootstock [45]. The authors of that study hypothesized that exchanges of molecules between rootstock and scion or the absence of signaling or receptors, were leading to the perception of scion by rootstock as non-self and triggering an immune response [45].

While scion identity had the strongest influence on graft take for the combinations of plants tested in this study, rootstock population was the most significant factor explaining plant survival over 30 months, with scion remaining significant as well. Accumulation of nonstructural carbohydrates in the rootstock of certain graft combinations could indicate xylem and phloem degeneration around the graft junction. This study found that, when used as a scion, Criollo 22, which is representative of many of the fine flavor varieties cultivated in Mesoamerica, showed the lowest rate of graft success across rootstocks, as well as the lowest survival after grafting. Grafting of Criollo is important not only for the preservation of unique phenotypes of this background, which is a high-value heritage cacao, but also because Criollo-type cacao has a longer juvenility period from seed and takes, on average, 2 years longer to fruit and flower than other varieties, possibly due to a higher genetic load from the bottleneck event that resulted in its domestication [46]. An unusual scion-mediated effect on rootstock was revealed for grafts with Criollo 22 scions. These plants displayed enhanced secondary thickening of the stem below the graft junction over the course of the study that was 14.4% and 34% larger, respectively, than rootstocks with Matina 1/6 or Pound 7 scions. This enhanced secondary growth below the graft junction could potentially reflect the accumulation of starches in the rootstock. The results suggest that genotype-specific mechanisms of the scion may be responsible for driving graft incompatibilities, particularly in highly homozygous cultivars of cacao. Matina 1/6 is a cultivar that is representative of the genetic background of much of the West African Amelonado type, and Matina 1/6 and Criollo 22 often contrasted in their graft take on specific rootstock (e.g., BYNC, T 294, and T 384), as well as in their survival on specific rootstock populations (e.g., BYNC, SCA 6, and PA 107). The more highly heterozygous scion, Pound 7, showed both the highest graft take and the highest post-grafting survival, and perhaps better represents the response of admixed cultivars. Practical recommendations resulting from this work include the need to test multiple scions with any rootstock selection. The ancestry of scions should be considered in the selection of rootstock, particularly for cultivars with a high coefficient of membership in a particular genetic background. Of the rootstocks tested, PA 107 was distinguished by having high rates of both graft take and survival across the different scion cultivars, as well as high vigor.

A micro XCT scan of a healthy graft junction representative of the planting materials used in this study revealed hidden cavitation at the site of the graft junction below the cambium and remnants of a necrotic layer at the graft interface. Transverse sections of graft junctions also revealed a more pronounced necrotic layer between rootstock and scion tissues in incompatible genotypes than in compatible genotypes. A small positive correlation between genetic relatedness and graft success could provide more context regarding the differential responses amongst genotypes, although wider sampling will be needed. Finally, evidence of starch accumulation below the graft junction could indicate translocated incompatibilities and signal that rootstock may play an important role in storing, partitioning, and translocating carbohydrates in healthy unions. This study provides insights into genetic, physiological, and anatomical markers of graft incompatibility and may provide context for the observed differences in the survival of grafted plants.

4.2. Vigor Is a Complex Trait Reflecting Direct and Interactive Genetic Effects, as Well as the Environment

This study sought to determine which vigor characteristics of young grafted plants could be regulated by rootstock. Rootstock vigor was observed to drive scion vigor in the first year after grafting. Three years after grafting, however, rootstock no longer accounted for variation in scion traits. Vigor differences between plants stabilized, potentially suggesting a size capacity constraint of the pot over the course of the 30-month study. The results of this pilot suggest a bi-directional relationship driven in part by both rootstock and scion, as the scion influenced the vigor of rootstock below the graft in the third year of the study. Leaf dry weight was shown to be influenced more by row position than rootstock, highlighting the strong degree of micro-environmental interaction with certain traits, even within a controlled greenhouse study.

4.3. Rootstock and Scion Affect Foliar Nutrition in a Nutrient-Specific and Dynamic Manner

Rootstock was a significant driver of differences in 10 of 13 nutrients measured at 4 months, but only 2 after 24 months and 1 after 30 months. Fernández-Paz and colleagues [46] also found that rootstock effects on cacao scion nutrition were significant 2 months after grafting but no longer significant 4 months after grafting. This supports the importance of rootstock for the health of young grafted plants in the nursery and transplanting in the field. Aluminum was the nutrient that varied significantly by rootstock at two time points. This was the strongest association of any of the elements with either a rootstock or a scion effect and supports the body of literature reporting that rootstock will be a critical tool in regulating heavy metal uptake in contaminated soils [47,48,49]. In this study, scion was a significant driver of nutrient variation between treatments for sulfur and sodium at the 30-month time point. Elemental sulfur accumulation in cacao has previously been linked to fungal resistance against verticillium wilt [50]. Foliar sodium has been linked to osmotic adjustment in response to drought stress in cacao [51], and was the only element that appeared to be influenced by scion in two of the three time point measurements.

We found a weak, but significant, correlation between leaf starch content and phosphorous. This could be due to the role of phosphorous in both the formation of starch using glucose-1-phosphate and the phosphorolytic breakdown of starch, which are both catalyzed by the enzyme starch phosphorylase [52,53]. Storage of starch phosphate monoesters could also be a contributing factor [54], although the degree of phosphorylation of cacao transitory starch during accumulation and breakdown has not been evaluated, to our knowledge.

4.4. Phytophthora palmivora Resistance

We observed significant differences in lesion area among inoculated Matina 1/6 scions grafted onto different rootstocks. Notably, the rootstock rankings based on lesion size in these grafted scions mirrored the rankings observed when the same rootstocks were evaluated as ungrafted seedling populations. While the sample sizes were small, considering segregation of open-pollinated rootstock, the findings provide preliminary evidence that selecting for Phytophthora resistance in the rootstock confers some resistance to the scion.

5. Conclusions

In this study, we explored the effects of rootstock–scion interactions on a diversity of traits related to graft take and survival, morphology, nutrient and carbohydrate uptake, and resistance to Phytophthora palmivora. This work was exploratory in that it involved a small number of replicates and was limited to testing under greenhouse conditions, but nonetheless revealed several novel findings of practical importance for rootstock breeding and nursery management efforts.

• Evidence of both local and translocated graft incompatibilities in cacao is presented for the first time, and genetic distance and phenolic responses at the graft junction are proposed as further steps in understanding the incompatibility response. The accumulation of starches in the rootstock tissue below the graft junction and hypertrophy of rootstock with Criollo 22 scions, also presented for the first time, potentially indicate other complications of graft incompatibility that could impact the resilience or productivity of the scion over time. The role of cacao root systems in the storage and translocation of nonstructural carbohydrates should be further explored.
• This study demonstrated applications of rootstock in driving nutrient use efficiency and limiting heavy metal uptake, particularly in the first 4 months after grafting. The rootstock showed significant effects on potassium, magnesium, calcium, zinc, and copper 4 months after grafting and manganese, iron, and aluminum 24 and 30 months after grafting. Scion identity was significant in explaining foliar sodium at two of the three time points measured. Both rootstock and scion had significant effects on nitrogen, phosphorus, sodium, and boron.
• It was shown that seedling response to Phytophthora palmivora inoculations could be predictive of conferred resistance when used as a rootstock. This could help to expedite the screening of rootstock, but should be validated in field studies and with larger populations.
• Early-stage greenhouse studies of rootstock–scion interactions can be informative, but factors like microclimate and container effects may influence certain traits disproportionately when experiments last for multiple years. Leaf biomass was particularly sensitive to position-related microclimatic variation within the greenhouse. Although certain morphological traits—such as root, leaf, and branch biomass—appeared to converge toward limits potentially imposed by container size, trunk biomass and rootstock diameter continued to exhibit significant cultivar-specific variation throughout the study.

This work provides novel insights regarding rootstock–scion interactions in cacao. Further work should explore the performance of commercial varieties with the inclusion of more replicates and validate these discoveries in field-grown trees.