Materials and Methods

Growing media and seeding.

Seedlings of kale (Brassica napus L. ‘Red Russian’), cabbage (Brassica oleracea L.), arugula (Eruca sativa L.), and mustard (Brassica juncea L. ‘Ruby Streaks’) were grown in fiber trays (23.5 × 48.5 × 3.5 cm) for 10 to 11 d after sowing. The growing substrate comprised (by volume) 30% peat, 30% compost, 30% coir, and 10% perlite. The substrate analysis indicated that the macronutrient concentrations (mg·kg⁻¹ DW) were 1410 K, 1390 Ca, 329 P, 295 Mg, 220 S, and 42 N–NO₃; the micronutrient concentrations (mg·kg⁻¹ DW) were 218 Na, 47 Fe, 10 Mn, 9.4 Zn, 0.7 B, and 0.6 Cu. Trays were machine-filled with substrate and preseeded at Greenbelt Microgreens (Lynden, Ontario, Canada) on the day each replication was started. Seeding rates were the same as used in production at Greenbelt Microgreens. Per-tray seeding rates, which varied according to seed size, were 8 g for kale (2.78 mg/seed), 10 g for cabbage (2.84 mg/seed), and 4 g for arugula (1.56 mg/seed), and mustard (1.63 mg/seed), resulting in seeding densities ranging from 2.15 to 3.09 seeds/cm². Trays of microgreens were top-irrigated with well water as needed.

Growth chamber environment.

The experiment was conducted at the University of Guelph (Ontario, Canada), with sequential replications starting on 13 Oct., 14 Nov., and 12 Dec. 2017, respectively. A 29 m² walk-in growth chamber was divided into six 1.5-m² treatment plots, surrounded by opaque white vertical blinds to prevent light contamination between treatments, while still allowing sufficient airflow to maintain uniform temperature and humidity throughout the growth chamber. Each plot was divided into four subplots that were the same dimensions as a single tray. The four species were each randomly assigned to one of the subplots within each plot. Tray positions rotated clockwise daily, while maintaining a constant aspect, to enhance uniformity of crops’ exposure to light. The growing area for each light treatment was 0.46 m² and was centered directly below the lights.

Lighting Setup.

LED arrays (LX601C, Heliospectra AB, Gothenburg, Sweden) were used to provide six PPFD treatments: 100, 200, 300, 400, 500, and 600 μmol·m⁻²·s⁻¹ (Table 1), measured at the level of the top of the substrate. These fixtures have four separately controllable spectrum channels: B (peak 445 nm), R (peak 660 nm), white (W, 5700K broad spectrum), and FR (peak 735 nm). Only the B and R channels were used in the present study, with the fixtures configured to provide a photon flux ratio of B15:R85. For each treatment, two LED arrays, spaced 36 cm apart (on center, with the long sides of the fixtures parallel), were mounted 56.5 cm above the top of the substrate (measured from the bottom of the LED array). For the 600 μmol·m⁻²·s⁻¹ treatment, a third fixture was centered between the other two at the same elevation and aspect, which resulted in 1-cm gaps between the near sides of adjacent fixtures. LI and spectral quality were measured using a radiometrically calibrated ultraviolet-VIS Flame spectrometer (Ocean Optics, Dunedin, FL) coupled to a 400 nm × 1.9 m patch cord with a CC3 cosine corrector and tethered to a laptop running the Spectrasuite software package. Spectral power distribution was converted to photon flux using the PARspec subroutine. Before commencing each replicate, LI and B:R were measured in 25 locations in each plot, on a 16.7 × 16.7 cm square grid, centered under the LED arrays (Table 1). Intensity and uniformity measurements were repeated after harvest to confirm that the average PPFD of B:R did not change more than ±3%, within each plot, over the course of each respective replicate.

Table 1.Mean photosynthetic photon flux density (PPFD) from red (R) and blue (B) light (and combined) delivered from sole-source LEDs to achieve a target light ratio (%) of red to blue 85:15 (R₈₅:B₁₅) with target PPFD levels of 100, 200, 300, 400, 500, and 600 μmol·m⁻²·s⁻¹. Mean values reported are the average of 25 spectral scans before each respective replicate.

View Fullscreen

The LED arrays and growth chamber environmental control were set to the same daily 16-h light (L)/8-h dark (D) photoperiod with L/D temperature setpoints of 21 °C/17 °C and a constant 80% relative humidity (RH) setpoint. Three data loggers (HOBO U12-013; Onset Computer Corporation, Bourne, MA) were used to record air temperature and RH every 120 s. For each replicate, the loggers were positioned on the bench adjacent to the microgreen trays in three randomly selected plots. Halfway through each trial, the logger locations were switched to the other three plots to provide plot-level environmental data for each plot for half of the duration of each replication. This resulted in 18 unique data-collection events that, when the means for each event were averaged, resulted in L/D temperatures and RH (mean ± SD), of 21 ± 0.42 °C/17 ± 0.31 °C and RH of 86 ± 1.9%. Supplemental CO₂ was not used in the experiment, and the mean chamber CO₂ concentration (mean ± SD) during the daylight period was 470 ± 22 ppm.

Growth and morphology measurements.

Growth and morphology measurements were taken 10 d after sowing for kale and cabbage and 11 d after sowing for arugula and mustard. Five representative plants from each subplot were selected to determine HL, which was measured from the base of the hypocotyl to the shoot apical meristem using a ruler. Three subsamples, each comprising five representative plants from the remaining plants in each subplot, were collected to measure LA using an LA meter (LI-3100C; LI-COR Biosciences, Lincoln, NE). If visible, true leaves were of insignificant size, and therefore only cotyledons were used for LA measurements. Three additional representative samples (i.e., full plants and substrate) were collected using a cylindrical core sampler (76.4 cm²) from each subplot to measure fresh and dry yield. All plants within each core were cut just above substrate level and combined to determine FW. Each sample was dried in an oven at 70 °C for 3 d to constant weight, and DWs were recorded.

Digital image analysis of cotyledons.

Five representative cotyledons (one per plant from five plants) per subplot were scanned (CanoScan LiDE 25; Canon Inc., Tokyo, Japan) in JPEG format at 297 pixels per inch. ImageJ 1.42 software (https://imagej.nih.gov/ij/download.html) was used to determine red, green, and blue (RGB) values for each image, which were then converted into hue and saturation values using the formulas outlined by Karcher and Richardson (2003). Hue angle (HA) refers to a position on a continuous circular scale (0° to 360°), and saturation refers to the purity of a color with 0% and 100% representing gray and full saturation, respectively (Karcher and Richardson, 2003). A calibration curve was obtained by fitting a linear regression equation to HA values measured by scanning 12 color chips from the Munsell Color Charts for Plant Tissues (GretagMacbeth LLC, New Windsor, NY) with varying HAs, ranging from green to red (5Y 6/6, 2.5GY 6/6, 5GY 6/6, 7.5GY 6/6, 2.5G 6/6, 5G 6/6, 5Y 6/6, 2.5GY 6/6, 5GY 6/6, 7.5GY 6/6, 2.5G 6/6, and 5G 6/6). The present study’s HA values were corrected based on the resulting linear regression equation: y = 0.993x + 10.7 (r² = 0.996) where y = actual HA of Munsell Color Chart for Plant Tissues color chips and x = HA quantified by digital image analysis.

Statistical analysis.

The experiment was a randomized complete block design with six LI treatments, four microgreen genotypes, and three consecutive replications. Data were analyzed using R statistical software (RStudio 1.1.453; Auckland, New Zealand). HL, LA, HA, and saturation were individually analyzed using linear regressions, and FW and DW were individually analyzed using asymptotic light response curves (described subsequently), with both the independent (i.e., PPFD and TLI) and dependent (i.e., production and harvest indices) values as continuous variables. The relationships between LI and yield (FW or DW) were determined using the asymptotic model y = a + be^(cx) (Delgado et al., 1993), where y, x, a, and e represent yield, LI (i.e., PPFD or TLI), estimated maximum yield, and Euler’s number, respectively. The parameters for a, b, and c were derived through nonlinear regressions. All regression analyses were evaluated at P ≤ 0.05 level of statistical significance. The best-fit model regressions and equations are only presented for production and harvest indices with significant regressions. Normality of residuals and homoscedascity of variances were confirmed by using the Shapiro-Wilk and Levene test, respectively.

Results

Results are discussed primarily in terms of PPFD to be most relatable to similar studies. However, because PPFD and TLI are proportional in this study (because LI and photoperiod are fixed levels), the models are also presented in terms of TLI (Fig. 1) and the respective models share a common r². Also note that the growing period for arugula and mustard was 1 d (≈9%) longer than cabbage and kale. Therefore, while the PPFD levels (on the x-axes) are the same for all genotypes, the TLI for cabbage and kale are smaller than for arugula and mustard. Therefore, when considered on a PPFD-basis, the time-weighted effects of LI on harvest metrics may be more pronounced between some genotypes than the harvest data indicates.

View Figure

• Download as Powerpoint
• Download Figure

As LI increased from 100 to 600 μmol·m⁻²·s⁻¹, HL decreased linearly by 1.5 cm (24%), 0.80 cm (23%), 1.1 cm (37%), and 2.3 cm (62%) for kale, cabbage, arugula, and mustard, respectively (Fig. 1A–D). There were no treatment effects on per-plant LA (Fig. 1E–H). Both FW and DW increased asymptotically as LI increased. As LI increased from 100 to 600 μmol·m⁻²·s⁻¹, FW increased by 0.59 kg·m⁻² (36%), 0.70 kg·m⁻² (56%), 0.71 kg·m⁻² (76%), and 0.67 kg·m⁻² (82%) for kale, cabbage, arugula, and mustard, respectively (Fig. 1I–L). As LI increased from 100 to 600 μmol·m⁻²·s⁻¹, DW increased by 47 g·m⁻² (65%), 45 g·m⁻² (69%), 64 g·m⁻² (122%), and 65 g·m⁻² (145%) for kale, cabbage, arugula, and mustard, respectively (Fig. 1M–P). As LI increased from 100 to 600 μmol·m⁻²·s⁻¹, HA decreased by 35, 19, 19, and 25° for kale, cabbage, arugula, and mustard, respectively (Fig. 1Q–T). Saturation of kale and arugula was not influenced by LI (Fig. 1U and W); however, saturation of cabbage and mustard increased by 18% and 36%, respectively, as LI increased from 100 to 600 μmol·m⁻²·s⁻¹ (Fig. 1V and X).

Discussion

From a production standpoint, microgreens represent a unique commodity in that the input biomass (i.e., the seeds) and the harvested crop (i.e., aboveground biomass of young plant at or before the first true leaf stage) may not be that different. Unlike more mature plant life forms, embryos of germinating seeds rely solely on stored energy resources before the development and activation of the photosynthetic machinery. The transition from heterotrophic to autotrophic growth is a dynamic process that is influenced by myriad factors, such as seed size, planting depth, temperature, and availability of PAR. Although there were no direct measurements made on photosynthesis or the balance of autotrophic vs. heterotrophic growth in this study, the modeled relationships between LI and harvest metrics (Fig. 1) clearly show the autotrophic proclivity of the crops by the time of harvest.

Increasing LI is commonly associated with more compact growth (i.e., shorter internodes) in mature vegetative plant tissues (Burkholder, 1936; Butler, 1963; Yeh and Hsu, 2004; Zervoudakis et al., 2012), but the influence of LI on HL of young seedlings has not been as well documented. Some studies on seedlings of floriculture and vegetable crops have presented mixed results on the effects of LI on HL (Craver et al., 2018, 2019; Hernández and Kubota, 2014; Poel and Runkle, 2017; Pramuk and Runkle, 2005; Randall and Lopez, 2015), usually with either reductions or no change in HL with increasing LI. Seedlings of Brassica genotypes all tend to grow shorter under higher LI (Gerovac et al., 2016; Potter et al., 1999; Samuolienė; et al., 2013), although the magnitude of the effects may be influenced by genotype, seed size, seeding density, and time between sowing and harvest. When harvested between 10 and 15 d after sowing (DAS), Gerovac et al. (2016) reported reductions in HL of up to 30% in kohlrabi, mizuna, and mustard microgreens (seeding rate between 12 and 20 mg·cm⁻²) as LI increased from 105 to 315 μmol·m⁻²·s⁻¹, under most genotype-by-PFR combinations. Samuolienė; et al. (2013) reported reductions in HL ranging between 9% and 28% in 10-d-old mustard, red pak choi, tatsoi, and kohlrabi microgreens (sown at between 5 and 10 mg·cm⁻²) when grown at 545 vs. 110 μmol·m⁻²·s⁻¹. In contrast to these studies, Potter et al. (1999) showed ≈3-fold shorter HL, both 7 and 14 DAS, in canola (Brassica napus L. ‘Westar’) seedlings grown (one seedling per cell) at 500 vs. 150 μmol·m⁻²·s⁻¹. Given these varying responses of similar genotypes to LI under different planting densities, plant responses to vegetated shade may be antagonistic to the full phenotypic expression of LI-induced reductions in HL, such as in the production of endogenous gibberellins (Potter et al., 1999). The differences between genotypes in their LI effects on HL may have also been influenced by the availability of stored energy resources (i.e., size of seed). To illustrate this, kale and cabbage, the two tallest genotypes (for a given LI) in the present study, had about twice the seed size and also had higher seeding densities than arugula and mustard. High LI-mediated reductions in HL may have important ramifications for microgreen production as shorter HL may increase overall crop robustness and lengthen postharvest shelf life; however, shorter HL can also increase difficulty of harvesting, particularly if harvesting is done by machine. Depending on the importance of HL at harvest, relative to other production metrics (e.g., yield), it may be possible to counteract the LI-induced reductions in HL by using targeted light spectrum treatments such as a reduced photon flux ratio of R to FR light (Blom et al., 1995; Fletcher et al., 2005; Hisamatsu et al., 2008; Mah et al., 2018) or monochromatic blue light treatments (Hata et al., 2013; Hernández and Kubota, 2016; Kim et al., 2014; Kong et al., 2018) to stretch the plants.

Probably the most important harvest metric for commercial microgreen production is fresh yield (i.e., kg_(FW)·m⁻²) because microgreens are typically sold on a “per-FW” basis. All genotypes in the present study showed asymptotic trends of increasing FW with increasing LI. Kale and cabbage had higher yield at a given LI than arugula and mustard, which may have been partly due their higher seed sizes and seeding densities. The FW data (on a per-plant basis) reported by Gerovac et al. (2016) only showed treatment effects on some genotype × PFR combinations, with no LI treatment effects in kohlrabi or mizuna regardless of PFR and 19% and 34% increase in FW in mustard for the PFRs with 7% FR and 18% green (G, 500 to 600 nm), respectively, as LI increased from 105 to 315 μmol·m⁻²·s⁻¹. Their FW results for mizuna are conflicting between their text (15% increase in FW regardless of PFR) and graphical results (no difference in FW), as LI increased from 105 to 315 μmol·m⁻²·s⁻¹. On the basis of the asymptotic models of FW response to LI in the present study, the crops in Gerovac et al. (2016) should have been well below the saturation point on their respective light response curves, and thus readily responded to increasing LI with increasing biomass production. However, it is still not readily apparent how increasing LI by a factor of 3 (i.e., increasing DLI from 6 to 18 mol·m⁻²·d⁻¹) did not have substantial LI treatment effects on fresh yield for many genotype × PFR combinations in the Gerovac et al. (2016) study. This may indicate that other (unknown) factors related to growing environment or crop husbandry may have been limiting plant growth and yield in their experiments. Notably, kohlrabi, which was the genotype that showed the lowest FW response to increasing LI, also had the largest seeds (i.e., ≈2-fold higher mass per seed than mizuna and mustard) based on current information from their seed supplier (Johnny’s Selected Seeds, Winslow, ME). Regrettably, the Samuolienė; et al. (2013) paper did not present any FW data for their crops. Both Gerovac et al. (2016) and Samuolienė; et al. (2013) presented “percent DW” data, but this metric is essentially a measure of water content at the time of harvest and does not relate directly to biomass production. The trend of increasing percent DW (i.e., decreasing water content) with increasing LI in the present study (data not directly shown) are similar to the trends reported by both Gerovac et al. (2016) and Samuolienė; et al. (2013). These observations may be indicative of inherently reduced water content in plants grown under higher LI or may be an artifact of inadequate irrigation strategies in some studies, perhaps related to disparate water demand of plants grown under large LI ranges.

DW yield data, although less directly applicable to commercial production goals than FW, is the most widely used metric in academia for assessing treatment effects on biomass accumulation, particularly in foliar tissues. In the present study, DW followed similar trends as FW. Overall, a 6-fold increase in LI resulted in a 1.6- to 2.5-fold increase in (aboveground) DW production. When an asymptotic model was applied to the aboveground DW data presented in Potter et al. (1999), there was an ≈3-fold increase in canola yield over the same range of LI (model extrapolated to 600 μmol·m⁻²·s⁻¹). However, because their plants were harvested 17 DAS, the TLI was more than 35% higher than the TLI in the present study (for a given PPFD); therefore, over a comparable TLI the increase in aboveground DW would be closer to 2-fold. This comparison exemplifies why extreme caution is required when relating results of different LI trials on the basis of PPFD or (to a lesser extent) DLI. Further, although it has been said that 1% more light should result in concomitant 0.5% to 1% increases in crop yields (Marcelis et al., 2006), this generalization appears to be exaggerated for microgreen production, where growing cycles are short and seed storage likely provides a disproportionate contribution to the total harvest biomass. The net gains in DW were 47.0 g·m⁻², 45.1 g·m⁻², 63.7 g·m⁻², and 64.7 g·m⁻² for kale, cabbage, arugula, and mustard, respectively, as LI increased from 100 to 600 μmol·m⁻²·s⁻¹. When considering these gains in DW along with the initial seed (input) mass of the different genotypes, it would appear that seedlings of smaller-seeded genotypes, such as arugula and mustard, may have considerably higher phenotypic plasticity for biomass accumulation responses to increasing LI, even when accounting for their higher DAS (i.e., higher TLI). This may be an important consideration for growers to balance when considering the various input costs associated with the two main drivers of harvestable biomass accumulation in microgreen production: initial seed resources and photosynthetic light inputs.

While there were increasing trends for aboveground fresh and dry biomass metrics with increasing LI, there were no LI treatment effects on leaf area for any of the genotypes. Although the relative biomass allocation to hypocotyl and leaf tissues were not assessed in this study, it is likely that higher LI treatments produced plants with greater leaf thickness because this trend has been observed in other species and production scenarios (Givnish et al., 2004; Matos et al., 2009; Morais et al., 2004; Sims and Pearcy, 1994). Further, because tissue water potential and morphological factors that can influence the rate of postharvest water loss may also have a substantial influence on shelf life, more research is needed to elucidate how varying LI affects leaf morphology (particularly leaf thickness and stomatal densities) and relate these metrics to shelf life.

The general decreases in HA observed in the cotyledons of all genotypes were indicative of increases in the proportions of yellow and red pigments in the cotyledons of kale, cabbage, and arugula (all green colored plants), and purple (likely associated with anthocyanins) in the cotyledons of mustard (reddish-purple colored). Only mustard cotyledons had LI treatment effects on saturation (i.e., purity of color relative to gray), with increasing saturation at higher LI. These observations may be indicative of increased concentrations of carotenoids and anthocyanins at higher LI. Carotenoids in the chloroplast play an essential photoprotective role in reducing photooxidative damage to the photosynthetic apparatus caused by excess LI (Krinsky, 1979). Chloroplast-specific carotenoids are synthesized in response to light to protect against photoinhibition (Bou-Torrent et al., 2015; Llorente et al., 2017; Toledo-Ortiz et al., 2010). Anthocyanins have also been shown to screen photosynthetic tissues from high light intensities (Smillie and Hetherington, 1999), and increased anthocyanin synthesis has been shown to occur under high LI (Krol et al., 1995; Mancinelli, 1983). It is also likely that higher LI reduced the quality of the visual appearance of Brassicaceae microgreens, although how the magnitude of these differences relates to normal human perception of plant color was beyond the scope of this study. Because plant color is the most important visual indicator of plant quality and health (Barrett et al., 2010), it is possible that deleterious effects of higher LI on the visual appearance of Brassicaceae microgreens may somewhat offset any increases in yield.

Overall, the magnitude of the morphological and yield responses expressed by the four genotypes indicated that arugula and mustard exhibited greater levels of phenotypic plasticity to LI than kale and cabbage. This may have important production implications as growers may wish to assign different levels of lighting infrastructure and energy budget to growing commodities that exhibit greater yield responses to increasing LI. Photoperiod and spectrum were kept at fixed levels, with varying PPFD levels in the present study. Lighting at lower intensities for longer photoperiods (up to and including continuous lighting) and gradually increasing the LI as the crop matures (either using dimmers or by moving the plants between zones with different LI) are other common methods for reducing costs associated with crop lighting. In some cases, crops are even kept in the dark until germination is complete and the cotyledons have unfolded. These additional strategies also deserve attention in the scientific community. As more data become available, expression of lighting levels in terms of TLI will permit more relevant comparisons between studies of similar crops grown in different production scenarios.

Conclusion

Overall, the results of this study are demonstrative of how SS LI can affect the interplay between the most relevant factors of growth, yield, and quality for several economically important genotypes of Brassicaceae microgreens. Indoor growers of Brassicaceae microgreens can use these results, especially the light response curves for yield, to help determine the economic optimum LI to target for their production systems, genotypes, and production goals.