2.1 Experimental design

The companion plant treatment was employed as the main plot factor applied at the tunnel level (n = 6 total high tunnels, three each for with vs. without companion plants). The six tunnels were divided into three blocks across the north–south gradient. Within each of the blocks, one tunnel was randomly designated for tomato alone and the other for tomato cultivated with a companion planting. The control tunnels were planted with only tomato across the 4-year experiment to replicate a standard high-intensity production system. Tunnels assigned to companion planting were cultivated with tomato combined with white clover (Trifolium repens). Aside from the addition of clover, all other aspects of production were identical (e.g., irrigation, pruning, fertilization). For the initial 2 years of the experiment, kale was also implemented as an overwintering crop in the clover-added tunnels due to its cold tolerance and use by growers in winter production. However, due to poor and inconsistent establishment this practice was abandoned.

The grafting treatment was employed as a subplot factor within all high tunnels. All tunnels contained both heirloom tomato cultivars (CP, BK) as scions, which were grafted onto two alternative rootstocks—the commercially available Maxifort (MF) and Solanum pimpinellifolium (SP), a wild relative of tomato. MF was selected due to its high vigour and resistance to multiple soil-borne pathogens (Grieneisen et al., 2018; Rivard & Louws, 2008), as described by the seed vendor (Johnny Selected Seeds, Winslow, Maine). The wild tomato S. pimpinellifolium was selected because wild relatives of tomato are thought to have greater resistance and/or tolerance to pests and pathogens (Ferrero et al., 2020; Zhang et al., 2017). Also, wild tomatoes showed weaker negative PSFs than domesticated tomatoes (Carrillo et al., 2019), suggesting that wild species like S. pimpinellifolium have roots that are less prone to yield reductions when grown consecutively in the same soil. In addition to MF and SP rootstocks, ungrafted and self-grafted treatments were included to control for the effect of grafting alone, independent of rootstock. These combinations resulted in eight plant types: two scions × four graft types (BK, BK/BK, BK/MF, BK/SP and CP, CP/CP, CP/MF, CP/SP). Within each tunnel, three of the six rows were cultivated with BK (rows 1, 3, 5), while the other three were cultivated with CP (rows 2, 4, 6). Within each row, all possible grafting combinations were present in a randomized complete block design. Each row consisted of four replicated blocks with each block having one plant of the four grafting treatments (i.e., 16 plants per row). At the tunnel level, this resulted in 12 plant replicates of each of the eight grafting treatments, and 72 plant replicates per treatment combination across all six tunnels included in the experiment (for a visual depiction of the layout of these treatments, see Supporting Information: Figure S1).

2.2 Transplant production

Tomato seeds were sown in 72 cell trays filled with soilless potting mix (Metro mix; Sun Gro Horticulture) and propagated in a greenhouse on the Purdue University campus under supplemental high-pressure sodium lamps (1000 W; 06:00-22:00). Tomato seedlings designated for grafting were grown for 3–4 weeks or until the diameter of the stem was ~3 mm. Grafting was performed using the cleft technique where a ‘V’ shape was cut into the bottom end of the scion and inserted into a vertical cut made into the rootstock. A grafting clip was placed over the graft union and the grafted tomatoes were placed into plant trays covered with humidity domes. Grafted seedlings were kept in darkness at ~100% relative humidity for 1 week for the graft union to heal and show signs of callous formation. Thereafter, grafted plants were moved into low light conditions for another week before moving to standard light levels for plant growth.

2.3 High tunnel preparation and management

For initial preparation of the high tunnels in 2018, all weeds were mowed, and the ground was prepared using a rotary tiller. For subsequent years, the rotary tiller was only used for rows that were designated for tomato plants. Six beds were prepared in each tunnel with a hand-operated bed maker using black plastic mulch with 1.1 m spacing between adjacent rows and 0.75 m within row spacing between consecutive plants. From 2018 to 2021, tomatoes were transplanted on 28 May, 27 May, 1 June and 1 June, respectively. Each year, transplants that did not survive were replaced within 2 weeks, after which there were no further replacements throughout the growing season. In each tunnel, we installed an automated irrigation system (Rainbird, Model 1ZEHTMR) that connected to drip irrigation (12 mm) lines, which extended along the full length of each of the six beds beneath black plastic mulch used to control weeds.

In the companion plant treatment, the open soil between rows was sown with white clover at a rate of 10 lbs./acre. Clover was broadcast seeded during the first week of May and reseeded later in the growing season as needed to ensure establishment. However, in tunnels dedicated to the monoculture system, the interrow space was covered with black ground cloth to smother emerging weeds and ensure that tomato was the only plant growing in the soil. We also hand-removed weeds as needed throughout all tunnels in both treatments to maintain a consistent plant presence. Herbicides and other pesticides (e.g., insecticides, fungicides) were never applied. In control tunnels, tomatoes were always planted in the same location across years, whereas in the companion tunnels, tomato rows were shifted into the space cultivated with clover in the prior year.

For fertility, Jack's professional 20-20-20 fertilizer was applied via fertigation system at three times over the course of the growing season based on agronomic recommendations: one application at transplant, another application during the vegetative growth phase and one final application at fruit set to ensure the tomatoes were not deficient of macronutrients. Weekly throughout the growing season, all tomato plants were pruned and trellised continuously as they grew. For pruning, suckers and stems that were too close to the ground and/or diseased leaves were removed and discarded from the high tunnel to avoid contamination.

2.4 Crop performance

Once the fruits reached maturity (i.e., pink and ripe stage, ~75 days) they were picked and weighed from individual plants weekly. At the end of the growing season, total yield was calculated per plant. Yield data from the initial year of the experiment (2018) were collected based on a per tunnel average for each graft treatment so that we have sums for all plants within a grafting treatment. For this reason, we only present yield data from 2019 to 2021. Above-ground plant vegetative biomass was measured at the end of the growing season after the final harvest. To measure biomass, each tomato plant was detached from the root system at the soil surface, then individual plant fresh weights were recorded.

We used a Minolta SPAD 502 Plus Chlorophyll Metre to measure the chlorophyll content of the leaves. Three mature leaves were selected per plant, and leaf position was standardized across all treatments. Three measurements were taken per leaf for a total of nine measurements per plant and averaged to give one SPAD reading per plant. SPAD measurements were taken on three separate occasions (28 July, 27 August and 25 September) from all experimental plants during the final (2021) growing season.

To determine potential effects of treatments of fruit quality, tomatoes were collected from the different graft treatments and analyzed using a Brix refractometer (Atago, Japan). From each high tunnel, five representative tomatoes of each graft treatment were collected across all rows for a total of 180 tomatoes sampled and measured in 2021. Using a mortar and pestle, individual fruits were ground to a pulp and approximately 3 mL of liquid was transferred into the refractometer and Brix measurement was recorded. The instrument was cleaned thoroughly before the next measurement was taken.

2.5 Statistical analysis

All statistical analyses were performed using JMP Pro 16 using linear mixed models. Fixed effects included year, companion plant, tomato cultivar and grafting type with all possible two-way interactions tested. Tunnel and row nested within tunnel were included as random effects. Biomass data were square-root transformed to improve normality, whereas raw data were used for all other response variables (yield, SPAD, Brix). Year was not included as a factor for SPAD or Brix since these were only measured in 2021. For SPAD data, measurements over time were averaged across dates to create a single measurement per plant, which was used for analysis. Post hoc pairwise comparisons were performed using Tukey's HSD.

3.1 Biomass

There was no overall statistical difference in tomato biomass between management systems (monoculture = 1.44 ± 1.22, companion = 1.09 ± 0.84; kg/plant, mean ± SE); however, there was a significant interaction effect of Year*Companion (Table 1). Plant biomass increased across both tunnel treatments from 2018 to 2021, but the rate of increase was proportionally greater in monoculture (219%) than with clover added (163%). The only year where there was a significant difference between tunnel treatments was in 2020 when monoculture plants had 38% higher biomass than their counterparts with clover (Figure 1). There was no statistical interaction between companion planting and the other two variables in the model (Cultivar & Grafting).

Similarly, there was no overall statistical difference in tomato biomass between cultivars (BK = 1.29 ± 1.1, CP = 1.21 ± 0.98; kg/plant, mean ± SE); however, there was a significant interaction effect of Cultivar*Grafting (Table 1). This interaction was seemingly caused by cultivar-based differences in compatibility with the wild rootstock, where the median of biomass accumulation was 78% higher in CP than BK (Figure 2a). The grafting treatment also significantly interacted with year (Table 1). The Graft*Year interaction was primarily driven by changes in the performance of MF-grafted plants over time, which were not different from the ungrafted or self-grafted plants in years one and two but were significantly higher in years three (+71%) and four (+77%) (Figure 3a). The relative performance of the other three treatments remained comparatively stable over time with the ungrafted and self-grafted controls showing similar biomass accumulation and wild-grafted tomatoes growing nearly half as large (49% lower biomass).

3.2 Yield

Grafting had a significant main effect on fruit yield, but graft type also significantly interacted with year (Table 2). Similar to biomass, this interaction effect was primarily caused by yield increases in MF-grafted plants over time compared to the other three treatments; namely, a 38% higher yield than the ungrafted and self-grafted controls in the final year of the experiment (Figure 3b). Over all years combined, tomatoes grafted to wild rootstock had 81% lower yields.

There was also a significant interaction effect of Cultivar*Grafting (Table 2). MF rootstock performed better when grafted to BK compared to CP scions (Figure 2b). While a nonsignificant effect, there was a trend for higher yields when CP was grafted to wild rootstock compared to BK; a pattern that was also apparent in the biomass data.

Companion planting had no effect on tomato yields, either alone or interacting with other variables in the model. While time was a significant factor, tomato yields were far less variable from year to year compared with biomass.

3.3 Crop quality

None of the experimental factors significantly affected SPAD values, which mostly ranged from 46 to 48 (see means according to treatment in Table 3). Brix values were affected by the interaction between Companion*Cultivar (F = 7.56, p = 0.0066) and Cultivar*Grafting (F = 3.25, p = 0.0410). However, the biological magnitude of treatment differences was minimal, with the majority of fruits scoring between 4.8 and 5.3 (see Table 3 for values).

4.1 Changes to tomato productivity in high tunnel soil over time

We tracked tomato biomass and fruit yield as proxies for plant productivity, expecting that both variables would decrease over time in the monoculture system, while maintaining production in the companion plant treatment. Our data did not support either of these predictions. Neither vegetative aboveground biomass nor yield declined over 4 years of growing in the same high tunnel soil. In fact, biomass increased over time in both treatments, even though this pattern was not reflected in fruit yield. The differential response in vegetative biomass and fruit production is not surprising given that different factors affect these two variables. For example, heat accumulation may constrain tomato fruit production due to limitations in pollination or pollen viability (Pressman et al., 2002; Sato et al., 2000). Prolonged daytime temperatures >32°C negatively affect fruit set; a threshold that is routinely exceeded during summer production months in our region, even in well-ventilated tunnels (Ingwell & Kaplan, 2019).

The lack of observed yield penalties in serial tomato production could have been caused by a few factors. Our field trial was conducted on land that had never been used to cultivate tomato. Although we did not systematically sample soil or roots for pathogen presence, plants exhibiting foliar symptoms of disease were sent for diagnosis. This process only revealed a single case of confirmed soil-borne disease out of the >1000 tomato plants. This outcome suggests that 4 years was insufficient to develop disease pressure in a tomato-naïve soil. Notably, Fu et al. (2017) found that continuous tomato monoculture on naïve soil improved soil quality and plant yields in the short term while only observing negative effects after >10 crop cycles. Thus, the timeframe of our experiment, while twice the duration of many published studies, still might be insufficient for observing negative effects. Future studies would benefit from implementing on-farm trials using commercial high tunnels that have soil with a much longer history (i.e., a decade or more) of tomato cultivation.

Because the monoculture system did not develop disease pressure and/or reduce yields, it is not surprising that the clover treatment had little to no effect. Field trials of cover crops on tomatoes, in general, show variable effects on crop performance with positive, negative or neutral effects changing across years and sites (Chahal & Van Eerd, 2021; DiGiacomo et al., 2023; Summers et al., 2014). Other authors have demonstrated that positive effects of companion plants depend on the presence of soil pathogens such as Pythium or plant–parasitic nematodes (Hannula et al., 2020; Ma et al., 2018). Even in the absence of disease pressure, leguminous covers like clover could have beneficial effects on soil fertility, but this mechanism is more likely to operate in low-input or organic cropping systems (Abdul-Baki et al., 1996; Lenzi et al., 2009). Since we applied synthetic fertilizer, these applications could have masked any potential nutritional benefits of clover-derived nitrogen.

4.2 Impact of grafting on tomato productivity

Our study clearly shows that grafting affects biomass and yield of tomato plants, though, selecting an adequate grafting combination is key to its success. Our data suggest that grafting scions of both cultivars of heirloom tomatoes on the commercially available MF rootstock is beneficial. MF-grafted tomatoes showed increased biomass and yield compared to ungrafted and self-grafted plants. Interestingly, this effect was only apparent in years 3 and 4 of the experiment, demonstrating that grafting benefits may accrue over time rather than appearing immediately. Given that disease pressure did not appear to increase over time, the mechanism responsible for this pattern remains unclear. MF-induced benefits to tomato can be highly variable for unknown reasons (Buller et al., 2013; Rivard & Louws, 2008) and thus 2020 and 2021 may have simply presented environmental conditions conducive to positive effects, unrelated to time since establishment. Earlier work by Masterson et al. (2016) on MF-grafted tomatoes in high tunnels showed fairly dramatic (+53%) and immediate yield increases. However, a meta-analysis by Grieneisen et al. (2018) on tomato grafting emphasizes the variable outcomes, which are more typical in these studies. For example, grafted tomatoes produced significantly higher yields in only about one-third of trials. Despite this variation, tomato grafting, overall, caused a 37% yield increase for all studies combined; an effect size that is nearly identical to what we observed (+38%) in the final year of the study. Recent studies also emphasize that grafted plants can strengthen beneficial microbial associations in the rhizosphere (Ge et al., 2022; Ruan et al., 2020), including in MF-grafted tomato (Poudel et al., 2019). This opens the possibility that grafting benefited tomato growth and yield in our experiment via the microbiome.

In contrast to MF, our wild grafted tomatoes performed poorly in terms of establishment and production, resulting in >50% reductions in growth and yield. Tomato has long been known to show compatibility with interspecific grafts, including closely related Solanum sp. and other genera in the Solanaceae family (Dawson, 1942). Moreover, recent research demonstrates successful implementation of tomato grafting with its closest wild ancestor, S. pimpinellifolium (Mauro et al., 2020; Naik et al., 2021; Zhou et al., 2022). These experiments, however, were performed using hybrid varieties as scions. Heirloom tomatoes may be less compatible with wild rootstock due to morphological or developmental differences.