Many perennial fruit crops are commercially grafted, an ancient practice in which a scion selected for desirable fruit-producing traits is joined to a rootstock chosen for its resistance to biotic and abiotic stresses or its ability to modify scion phenotypes (Warschesky et al., 2016). Grafting is essential in viticulture worldwide, particularly since the phylloxera crisis in the late 19th century. Most cultivated grapevines are Vitis vinifera scions grafted onto rootstocks derived from North American Vitis spp. and their hybrids. The production of a bench-grafted grapevine is a complex process, starting with winter harvesting of the canes that will be grafted and ending with the selection of viable plants in the nursery and their planting in the vineyard (Waite et al., 2015). Successful establishment and long-term growth of grafted plants depends primarily on the compatibility between the two genotypes and their ability to form robust vascular connections during the formation of the graft union (Tedesco et al., 2022; Loupit et al., 2023). It also depends on the ability of (i) the scion to develop leafy shoots from dormant buds and (ii) the rootstock to initiate roots and develop a functional, well-structured root system.

The geometrical, topological and morphological traits of roots, including the apical diameter, specific length, tropism, branching angle, branching density and lateral root expansion, determine the spatial configuration of the root system (i.e., the root system architecture, RSA) and influence plant performance and resilience during soil environmental stresses situations (Bardgett et al., 2014). The spatial configuration of the root system develops to ensure anchorage in the soil, to promote interactions with beneficial microbial communities in the rhizosphere, and to uptake and transport water and nutrients from the soil, especially under heterogeneous or limited resource availability (Lynch, 2007). The RSA is shaped by a complex set of individual root developmental processes, including initiation, emergence, growth (axial and radial), branching, tropism, senescence and decay. These processes are organised in space and time and are genetically determined, resulting in a great phenotypic diversity in form and structure between species, populations or clones (Hodge et al., 2009). The expression of these root processes is influenced by both endogenous signals and exogenous factors, such as the physico-chemical properties of the soil, temperature or humidity, which fluctuate greatly over the lifetime of the plant (Rellan-Alvarez et al., 2016). As a result, the number, location, and direction of growth of each root in the system is highly variable, even among genetically identical plants (i.e., root plasticity).

In contrast to seed propagation, where primary roots emerge from the seed and give rise to secondary lateral roots, vegetative propagation results in a root system composed entirely of adventitious roots (ARs) from which lateral roots branch off (Albrecht et al., 2017). In hardwood cuttings, AR rhizogenesis and root system development are strongly influenced not only by their genetic origin but also by the physiological and biochemical state of the parent plant from which the cuttings were taken (Smart et al., 2003). Endogenous hormones are key regulators of AR formation, with auxin acting as the primary signal to trigger rhizogenesis from non-root tissues (Druege et al., 2019). Recent studies, mainly in model plants, have demonstrated that other plant growth regulators including nitric oxide, ethylene, polyamines, jasmonic acid and strigolactones may also be involved in the process of root primordia induction (reviewed by Bannoud & Bellini, 2021; Altamura et al., 2023). In addition to the aforementioned growth regulators, the initial levels of endogenous carbohydrates, mineral nutrients and other biochemical components, such as phenolic compounds, that may act as rooting cofactors or modulators of auxin metabolism and transport, also impact AR formation and the subsequent root development in cuttings (Da Costa et al., 2013; Bannoud & Bellini, 2021). In grapevine, as early as 1927, Winkler observed a positive correlation between the degree of iodine staining of the starch present in the mother canes and the AR formation in cuttings (Winkler, 1927). More recently, the concentration of carbohydrates in the wood used for grafting (originating from different geographical areas) has been linked to grafted plant survival and development in grapevine (Bartolini et al., 1996). In poplar, carbohydrate availability in the base of cuttings was demonstrated to be higher in easy-to-root cultivars than in difficult-to-root cultivars, and soluble sugars were found to be more influential than starch in initiating root primordia (Denaxa et al., 2012). Rooting also tends to improve with increasing cutting length, as observed in woody species such as poplar (Desrocher & Thomas, 2003) and olive (Porfírio et al., 2016), possibly due to the presence of higher amounts of carbohydrate resources and/or higher enzymatic activity to remobilise and transport them to the base of the cutting. In grapevine, the results obtained are not always consistent, as they are highly dependent on the cultivars and cutting type (Boeno & Zuffellato-Ribas, 2022). Moreover, the role of carbohydrates in AR formation and early root growth remains unclear (Phillips et al., 2015). Finally, in the majority of studies investigating adventitious rooting in hardwood cuttings, root system characterisation is limited to a few traits such as the number of roots emitted after a few days or the length of the longest root. This is likely due to the difficulty and time-consuming nature of phenotyping roots in situ, even on young plants at an early stage of development (Vielba et al., 2020). To date, no comprehensive studies have been conducted to characterise the morphological and architectural features of the grapevine root system in relation to the initial state of the hardwoods used for grafting.

The objectives of this study were to investigate in young, grafted grapevines (i) the relationship between the quantity of reserves and architectural and morphological traits of the roots, and (ii) the genetic variation of these architectural and morphological traits. The quantity of reserves was manipulated by grafting scions with rootstock hardwoods of varying lengths. Genetic variation was examined using two commercial rootstocks, selected for their agronomic performance and contrasting root system patterns in vineyards. A deeper understanding of the mechanisms underlying AR formation and early RSA development is essential for improving propagation strategies in viticulture, particularly for hard-to-root rootstocks like V. berlandieri hybrids. This knowledge can help nurseries and grape growers to produce uniform, high-quality plants with greater consistency, thereby enhancing the success and long-term viability of vineyard plantings.

In both experiments, the plants were harvested to characterise several shoot and root traits (Table 1). The root systems were completely extracted and the substrate was removed by gentle washing with tap water. The length of the longest root (MLr1, cm) was measured with a ruler, and the number of ARs and leaves was counted. Subsequently, individual roots (three chosen at random in Exp. A and all roots were selected in Exp. B) were scanned with a flatbed scanner (Epson Perfection V850 Pro, America, Inc., USA) by placing them in a Plexiglas tray (22 cm × 30 cm, 1.5 cm) (Figure 1A). The Plexiglas tray was filled with a few millimetres of water, which was covered with a rigid transparent plastic sheet to disentangle the roots and minimise overlapping. Images were captured in TIFF format with a resolution of 1200 dots per inch (dpi) using the transparent mode. Finally, shoot and root samples were dried (in an oven at 60 °C until reaching a constant mass) and weighed. This allowed the calculation of the root biomass fraction (RMF, g g^-1) and the root-to-shoot mass fraction (RSR, g g^-1) (Table 1).

Vitis vinifera cv. Cabernet Sauvignon was grafted onto varying lengths of V. riparia cv. Riparia Gloire de Montpellier (RGM) and the V. berlandieri × V. rupestris hybrid cv. 1103 Paulsen (1103P). Results of a two-way ANOVA (with rootstock length and genotype as factors) are provided. ANOVA significance, ns: p-value < 0.05, * : 0.05 > p-value > 0.01, ** : 0.01 > p-value > 0.001, *** : p-value < 0.001

In Exp. B, the visible root systems of the rhizotrons were scanned periodically with an A3 flatbed scanner (Epson Expression 1640xl, America, Inc., USA) three times a week from 7 d after planting until harvest (day 30). Black and white images were captured in TIFF format at a resolution of 600 dpi (Figure 1C).

Rhizotron images obtained in Exp. B were semi-automatically vectorised in an RSML format using the Smartroot v4.21 plugin (https://smartroot.github.io/; Lobet et al., 2011) of the ImageJ v1.52 software (https://imagej.net/ij/). Each vectorised root was defined by a set of nodes characterised by 2D spatial coordinates, diameters, root orders and parental nodes (Figure 1B). For time-lapse images of root systems in rhizotrons, only first-order ARs emerging from the base of the rootstock cutting were analysed (Figure 1D). For each date, the vectorised root system from the previous date was superimposed on the image, and the new root tips were incorporated into the RSML files (Lobet et al., 2015). An in-house R script, adapted from the ArchiDART package (Delory et al., 2016), was then used to calculate, for each date, the number (Nr) and total length (TRLr1, cm) of ARs visible on the transparent side of the rhizotron, as well as the maximum depth (MRD, cm) and horizontal width (MRW, cm) reached by the root system and the surface area of the convex hull (CHA, cm²). To quantify the maximum spreading angle of a root system (i.e., the insertion angle between the two outermost roots) (MRA, °deg), a slightly modified version of the protocol described in Dayoub et al. (2021) was applied to the most recent image of the time series (Figure 1D).

Additional root morphological and topological information was also collected from 2D images of individual roots sampled after one month in rhizotrons. Four individual roots per plant in Exp. B, for a total of 128 roots, were fully vectorised using SmartRoot software as described above. Several traits were determined for each AR of order 1 to 3 (Table 1), including (i) the apical diameter (Dr1 and Dr2, mm), (ii) the length of the apical unramified zone (LAUZ, mm), which is the length between the root tip and the nearest lateral root, (iii) the inter-branching distance (IBD, mm), which is the average length between two consecutive lateral roots, (iv) the total root length by root order (RL, cm), (v) the average length of one daughter root per parent root (meanRLr2, mm) and (vi) the length of the first-order root branching zone (LBZ, cm). Branching density was quantified using IBD (its reciprocal) because this variable could be measured for each lateral root. Root length, mass and volume allow the calculation of specific root length (SRL, m g^-1) and root tissue density (RTD, g cm^-3).

Finally, all individual roots were also analysed using the Rhizovision Analyser software (https://www.rhizovision.com/; Seethepalli et al., 2021) to measure the total length of the roots on scans and root volume that allow us to calculate the integrated traits SRL and RTD on a large batch of samples. A detailed description of the traits is given in Table 1.

All graphs, data calculations and analyses were performed using R software (R Development Core Team 2023; R version 4.1.1; http://www.r-project.org/). A home-made R script using the ‘XML’ and ‘dplyr’ libraries was used to read RSML files and calculate root traits. Analyses of variance to compare rootstock length and genotype effects were performed using the ‘aov’ function in the ‘stat’ library and post-hoc Student–Newman–Keuls analyses using the ‘SNK.test’ function in the ‘agricolae’ library. Prior to analysis, the normality and homoscedasticity of the data were assessed using the Shapiro–Wilk test in the 'stat' library and the Levene test in the 'rstatix' library, respectively. Principal component analyses were performed using the PCA function in the ‘FactoMineR’ library. Plotting was done using the ‘ggplot2’ library. To compare the morphological root traits in Figure 6, the data were normalised between 0 and 1 using the normalisation method: (data-min)/(max-min).

Plants grew for 112 d in pots (Exp. A) or 30 d in rhizotrons (Exp. B) and showed no visual signs of stress or mineral deficiency. A 1:1 linear correlation was observed for each scion/rootstock combination between the number of ARs or root mass production in plants grown in pots and those grown in rhizotrons, highlighting comparable root system development across both containers (Figure S1). Rootstock length had a significant effect on 100 % of shoot traits, 75 % of global root traits and 40 % of root morphological traits (Table 1). In Exp. A, plant survival was not significantly affected by rootstock length, but there was a trend towards more viable grafts with a 50 cm rootstock than with an 8 cm rootstock (Figure 2A). Grafts with 50 cm rootstocks produced twice as much shoot and root biomass as those with 20 or 30 cm rootstocks and four times as much as those with 8 cm rootstocks throughout the time course; these growth differences occurred from the first time point measured, 28 d after planting (Figure 2B). Root biomass increased dramatically over time, multiplying by 5 to 10 times between the first and the fourth month after planting, while shoot biomass increased by only 1.5 to 2 times during the same period. Consequently, the root mass fraction (RMF) doubled over this time (Figure 2C). However, the increase in root biomass was not proportional to the increase in root length, resulting in a 50 % decrease in specific root length (SRL) from one to four months after planting (Figure 2D).

Figure 2. Effects of rootstock length on (A) plant survival rate, (B) shoot (solid line) and root (dotted line) dry mass accumulation, (C) root mass fraction (RMF) and (D) specific root length (SRL) in two scion/rootstocks combinations.

Vitis vinifera cv. Cabernet Sauvignon was grafted onto four lengths of V. riparia cv. Riparia Gloire de Montpellier (RGM) and the V. berlandieri × V. rupestris hybrid cv. 1103 Paulsen (1103P) and grown for 112 days in pots (Exp. A). (A) rootstock length does not affect grafted plant survival (Chi² test). (B-D) ANOVA results: not significant, ns; 0.05 > p-value > 0.01, *; 0.01 > p-value > 0.001, **; p-value < 0.001, ***.

Significant differences in root morphological traits were observed between CS/1103P and CS/RGM and between different rootstock lengths, particularly in the diameter and mean length of lateral roots (Table 1). Traits related to the first-order axis (Dr1 and LAUZ) did not vary significantly with rootstock length for CS/1103P, but Dr1 was affected by rootstock length for CS/RGM. For both combinations, lateral root traits were not affected by rootstock length (Figure 6). The SmartRoot data showed that the mean inter-branch distance (IBD) increased along the length of the branching zone (LBZ) of the first-order root axis for both rootstock genotypes (Figure 7A, 7B and 7C). In the basal, oldest parts of the first-order roots, the mean IBD was 2.5 mm, increasing to over 5 mm in the youngest parts (Figure 7A). This increase was consistent along the first-order axis for all CS/RGM rootstock lengths (Figure 7C), but less so for CS/1103P, which showed elevated IBD values in the mid-root zone (Figure 7B). The mean length of lateral roots decreased progressively along the LBZ, from 3 cm to 0.5 cm in CS/RGM and from 4 cm to 0.5 cm in CS/1103P (Figure 7D). This decrease was more pronounced in CS/1103P than in CS/RGM, irrespective of the length of the rootstock (Figure 7E and 7F). The apical diameter of lateral roots remained constant at 0.3 mm in CS/RGM, but decreased from 0.3 mm to 0.2 mm along the first-order root axis in CS/1103P, being significantly smaller in the distal LBZ (Figure 7G). In addition, the shorter hardwood resulted in a smaller Dr2 for CS/RGM (Figure 7H and 7I).

Figure 6. Effects of rootstock length on six morphological traits of individual roots of Vitis vinifera cv. Cabernet Sauvignon grafted onto four lengths of (A) the V. berlandieri × V. rupestris hybrid cv. 1103 Paulsen (1103P) and (B) V. riparia cv. Riparia Gloire de Montpellier (RGM) after 30 d of growth in rhizotrons.

Traits: First-order root length (RLr1), inter-branch distance (IBD), mean lateral root length (meanRLr2), lateral root apical diameter (Dr2), first-order root apical diameter (Dr1) and length of the apical unramified zone (LAUZ). Data are normalised between 0 and 1. Results of ANOVA: not significant, ns; 0.05 > p-value > 0.01, *; 0.01 > p-value > 0.001, **; p-value < 0.001, ***.

The length of rootstock material used for grafting significantly impacted the recovery and early-stage growth of young grafted grapevine during the first month after planting, encompassing both shoot and root development. This phenomenon is well-documented among grape growers (Waite et al., 2015) and has been observed in other woody species such as poplar (Desrocher & Thomas, 2003) and olive (Porfírio et al., 2016). Longer hardwood cuttings accelerated AR emergence and increased the maximum number of ARs, regardless of rootstock genotype. Moreover, longer grafts enhanced total root system length and mass, facilitating rapid anchoring in the soil and enhancing water and mineral uptake to support shoot development (Lynch, 2013).

The variation in rhizogenesis among genotypes in response to graft length can be attributed to the physiological and biochemical characteristics of the hardwood cuttings taken from the mother plants. Smart et al. (2003) showed that cuttings from the base of the shoot had a higher rooting capacity than those from the apex, due to their larger diameter and greater maturity, characterised by more advanced wood lignification. Therefore, to ensure consistent rooting potential, all cuttings in this study were standardised according to their position on the mother plant and their diameter.

Longer grafts are also more likely to root easily than shorter ones due to potentially higher levels of auxins, carbohydrates, mineral nutrients, or other biochemical components such as phenolic compounds (Bannoud & Bellini, 2021). Carbohydrates play a crucial role, not only in providing energy and carbon chains for biosynthetic processes in new root meristems but also in influencing gene expression in conjunction with auxin (Druege et al., 2019). The positive correlation observed between initial graft biomass and the number of roots produced suggests that greater carbohydrate allocation at the root formation site may have favoured AR formation in longer grafts. Additionally, the maximum number of ARs was achieved approximately 15 d after planting under each condition, possibly due to a progressive shift in phytohormonal balance from promoting to repressing adventitious rooting (Da Costa et al., 2013). The positive influence of graft length on rooting capacity exhibited significant variation among rootstock genotypes. Notably, CS/1103P had a higher maximum number of ARs than CS/RGM for equivalent graft lengths, contrary to existing literature suggesting lower rooting efficiency in 1103P compared to RGM (Gautier et al., 2018). Although CS/1103P produced a greater number of roots than CS/RGM, there was no significant difference in total root mass between the combinations. This indicates a decoupling in the mechanisms of AR emergence and subsequent growth.

A number of early root morphological traits were quantified at the individual root level, with only a few influenced by rootstock length or genotype (Figure 1B, Figure 6 and Table 1). For both CS/1103P and CS/RGM, the average length of the first-order roots (RLr1) was slightly greater in the 50 cm grafts compared to the 8 cm grafts. This outcome is consistent with the hypothesis proposed by Thaler and Pagès (1998), which suggests that individual root elongation rates are partially regulated by assimilate availability. However, traits such as the length of the unbranched apical zone (LAUZ) and the apical diameter of the first-order axis (Dr1), which are often considered predictive indicators of root elongation rate (Pagès et al., 2010), were not significantly affected by rootstock length in our study. Interestingly, only CS/RGM exhibited a smaller Dr1 with the shortest rootstocks. In general, the development of very fine roots in response to limited assimilates is considered a strategy to increase soil–root exchange surface area at minimal cost (Eissenstat et al., 2000). In this context, RGM appears to display greater plasticity compared to 1103P.

Similarly, root traits associated with branching, such as the number of lateral roots (Nr2), the total length of lateral roots (RLr2), and the inter-branch distance (IBD), were minimally impacted by rootstock length. Notably, only the lateral roots of the CS/RGM displayed a larger apical diameter (Dr2) in longer grafts compared to shorter ones, consistent with the apical diameter of their parent root (Dr1). This finding is consistent with previous non-grapevine studies, which have shown that root traits related to emission and elongation are more influenced by the physiological state of the plant, whereas traits related to root branching are more influenced by the surrounding soil environment (Bui et al., 2015).

Another consistent trend observed using Smartroot root image segmentation was the decrease in lateral root length, apical diameter, and to some extent, branching density along first-order roots, regardless of rootstock length (Figure 7). Typically, older roots near the base are longer than those recently formed at the apex (Pagès et al., 1993). However, for CS/1103P, the reduction in lateral root length and apical diameter along the primary axis may also result from reduced carbon assimilate availability during plant development or reduced sink capacity of apical root meristems to attract these assimilates. Furthermore, IBD increased progressively from the basal to the distal part of first-order roots for CS/RGM, while IBD was more variable for CS/1103P. Similar increases or longitudinal variations in branching density have been documented in several species (Pagès, 2014). While soil gradients, especially nutrient availability, often account for such fluctuations, however, the relatively homogeneous substrate and consistent nutrient solution watering in this study suggest otherwise. Instead, differences in root elongation rate between genotypes during the first 30 d might explain these variations.

The 2D RSA varied significantly between the two scion/rootstock combinations. For the same graft length, CS/1103P produced more ARs and had a greater total length of primary roots compared to CS/RGM, even though both genotypes had similar total root biomass. Additionally, while both rootstocks had similar branching densities, the lateral roots of CS/1103P were significantly shorter and thinner. The roots of CS/1103P grew at narrower angles from the cutting, indicating a stronger gravitropic response and a strategy giving priority to the exploration of deeper soil layers. In contrast, CS/RGM spread its larger ARs more horizontally. These distinct root system patterns reflect different strategies for soil exploration and resource use (Dayoub et al., 2021). The narrower angles and deeper roots of CS/1103P suggest a strategy optimised for water uptake, while the horizontal root spread of CS/RGM might be better suited for accessing immobile nutrients in the upper soil layers. These findings highlight the importance of considering RSA in rootstock selection to optimise grapevine performance under diverse environmental conditions.

Here, we present an original study characterising the architectural and morphological traits of young grafted grapevine root systems in response to variations in rootstock length and genotype. The results indicate a direct correlation between rootstock length and increased adventitious rhizogenesis and overall root system length, which could enhance soil exploration. Despite variations in total root number, individual root morphology remained unchanged in response to rootstock length, suggesting in particular that root branching was likely insensitive to carbon availability. These findings are consistent across two distinct rootstock genotypes with different RSA patterns and no common genetic background. Future research should focus on the long-term effects of rootstock length on grapevine development, using molecular and biochemical analyses to elucidate the underlying mechanisms. This knowledge will help to refine rootstock selection protocol and optimise planting practices, thereby contributing to more efficient and sustainable grapevine-growing strategies.

We thank C. Hevin, N. Hocquard, M. Lafargue and J.P. Petit for their technical support in preparing the plant material and monitoring the greenhouse, as well as the EGFV staff for their assistance with root phenotyping. The PhD studentship of M. Larrey was funded by INRAE and the Nouvelle-Aquitaine Region as part of the 'Archivit' project (22001535/6218). This study has been carried out in the framework of the University of Bordeaux's France 2030 programme/GPR Bordeaux Plant Sciences.