Optimal nitrogen rates and clonal effects on cannabinoid yields of medicinal cannabis

Abstract

Nitrogen (N) nutrition and germplasm of clones can influence biomass and cannabinoid concentration in medicinal cannabis. However, there are discrepancies on the optimal nitrogen (N) application rate at the flowering stage to achieve maximum yield and if, or how, this interacts with clones from different seed lines of the same genotype. This research examined the relationship between N application rate, concentration of cannabinoids and biomass yield of a CBD-type medicinal cannabis cultivar in clones propagated from five different seed lines (hereafter referred to as clones). Clonal rooted cuttings were propagated from five mother plants germinated from seeds of cultivar ‘Tas1’. Five N levels (30, 90, 160, 240 and 400 mg/L N) were imposed at the start of the inflorescence period and continued until harvest eight weeks later. Some pollen contamination occurred during the trial so that seed biomass was assessed for each plant and included in statistical analysis. Weight of total biomass, leaves and inflorescence (from upper and lower canopy positions), N%, and cannabinoid concentrations were measured after the harvest. Results indicated that increasing N supply generated a clear upward trend in inflorescence biomass that peaked at 160 mg/L N after which it did not significantly change, while leaf biomass steadily increased with N. Delta9-tetrahydrocannabinol (THC) and cannabidiol (CBD) concentrations decreased significantly with increasing N concentration in leaves with a similar, but non-significant, trend for inflorescences. The CBD to THC ratio increased with increased N. Clone source was strongly correlated with cannabinoid concentration, but not leaf, inflorescence or total biomass, across all N treatments. Clones 13 and 27 developed greater cannabinoid concentrations relative to clones 18 and 26 irrespective of N treatment. Pollen contamination induced seed development that comprised up to 5% of inflorescence biomass dry weight but this did not significantly affect whole-plant biomass, N accumulation (N%), or cannabinoid concentration. These findings provide valuable insights for improving cannabinoid yield in this widely cultivated plant species.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-96761-6.

Introduction

Cannabis sativa L. is a multi-purpose crop, traditionally cultivated as a source of food, medicine, and its bast fibres used in textile, paper industry, and material for building^1–3.

After many years of prohibition, the commercial cultivation of cannabis for medicinal and scientific use was legalized in Australia by the Federal Government in 2016 with amendments to the Narcotic Drug Act 1967. Since then, many Australian companies have obtained the licence to grow cannabis and a steadily increasing number of patients have had access to cannabis products^4,5. The manufacturing of cannabis products is regulated by the Therapeutic Good Act 1989, while the Office of Drug Control oversees the licensing and regulation of cultivation to ensure compliance with the international obligations under the United Nations Single Convention on Narcotic Drugs 1961.

The past scientific literature on stem and seed yield of cannabis demonstrated that N supply stimulates protein synthesis, fibre and seed production^6–9. However, nitrogen supply also is found to influence inflorescence cannabinoid concentration and yield^10–12. Some findings suggest^12,13 an optimal range for maximum inflorescence cannabis yield is found around 160 mg/L N above which cannabis plants develop toxicity response such as stunt growth and diminished cannabinoid concentration. Other findings report the optimal N supply to be at 50 mg/L¹⁴. Moreover, the N form, ammonium (NH₄⁺) or nitrate (NO₃⁻), is a critical factor determining the plant metabolic responses¹⁵.

Of other important factors that contribute to the overall inflorescence yield, such as growing medium^12,16, plant architecture¹⁷, plant density^8,18, temperature¹⁹, water supply^20,21, light spectrum and intensity^22–26, mineral or organic fertigation²⁷, arguably plant genetics is the most influential^14,28,29. Genetic heritability largely determines plant morphology and its chemical profile in cannabis³⁰. It is also reported that the content of cannabinoid not only is dependent on the genotype but also on the growth stage of the cannabis plant³¹. Both for legal and commercial reasons, it is important for growers to know the concentration of cannabinoids (THC in particular) in each cannabis chemovar before cultivation is initiated.

Clonal propagation by cuttings ensures plants are ‘true to type’. Similarly, other crops are clonally propagated (i.e. from cuttings): cassava (Colocasia esculenta), sweet potato (Ipomoea batatas), yam (Dioscorea spp.), and hop (Humulus lupulus). This ensures that quality-specific plant chemical traits are conserved in crops and are passed to next generations^32–34. Clonal propagation however comes with the risk of the quality of plants becoming less vigorous with lower cannabinoid content over time³⁵. To the best of our knowledge no studies have compared the influence of N supply on biomass and cannabinoid concentration (THC and CBD) in clones grown from different mother plants but belonging to the same cultivar.

The present research builds on the findings of a previous study³⁶ and was undertaken to evaluate the following hypotheses: (1) Clone source has an impact on biomass of leaves and inflorescences and cannabinoid concentration; (2) intermediate rates of N are optimal for cannabis yield; (3) differences in the spatial distribution of cannabinoid concentration (i.e., upper vs. lower inflorescences) will be less extreme using higher N application rates and; (4) CBD/THC ratio can change at harvest due to N supply.

Materials and methods

Plant material, experimental design and growing conditions

Cannabis clones were propagated from mother plants arising from seeds of the cultivar ‘Tas1’ (seeds provided by Martha Jane Medical, Australia). The plants used in this trial were grown from cuttings derived from 5 different mother plants, each coming from 5 different seeds of the ‘TAS1’ cultivar. This cultivar grows to less than two metres in height (under typical glasshouse production at 42 degrees south) and is chemotype III³⁷, which means it is rich in CBD (10–14% of dry weight for top inflorescences) and has a low concentration of THC (< 0.4% of dry weight). The TAS1 cultivar, and its clones used in this study, typically has CBD concentrations within the middle of the range seen globally [CBD in USA and European cultivars is typically between 5 and 20% of dry weight (Fig. 3)³⁸] After rooting, healthy cuttings 25 in total, rooted using Clonex rooting hormone (Yates) were potted in 35 L grow bags containing coarse sand on 6 April 2022, Hobart, Tasmania. During the first eight weeks all plants received uniform fertilization comprised of a standard Hoagland solution³⁹. During this period (8 weeks vegetative growth) the clones were cultivated under 18-hour light/6-hour dark cycle (18/6 h) in a controlled environment using LED grow lights (Spider farmer LED SP4000 PPFD MAP). LED lights were kept about 0.5 m above the top of plant canopies and adjusted thereafter to maintain a consistent height from the top of the plant canopy, with PPFDs ranging from 700 to 1000 µmol·m⁻² s⁻¹.

Fig. 3.

N percentage (% DW) of biomass (x-axis) versus the ratio of CBD to THC. Colours identify clones. Treatment labels are given as adjacent text (N levels in terms of mg/L). The lines are estimated using models based on Eq. (2), but in each case lines are drawn using the maximally complex model (in terms of the statistical significance of the fixed-effect x-variables from Eq. (2), according to p < 0.05), justified via a series of nested-model F-tests.

After eight weeks of vegetative growth, on 7 June 2022, cannabis plants were randomly allocated into five treatment groups of: 30, 90, 160, 240 and 400 mg/L N (treatments 1, 2, 3, 4 and 5, respectively). The experimental trial was performed in a split-plot arrangement by placing clone cuttings in main plots and applying nitrogen rates in sub-plots; the main plot contained five clones (2, 13, 18, 26, 27, these numbers represent the seed ID from which they derived) and the subplot had five nitrogen treatments (30, 90, 160, 240, and 400 mg/L N). N was supplied in the form of CaNO₃. Each clone ID (2, 13, 18, 26, 27) was part of each five nitrogen sub-plot treatment. For the sake of clarity, each N treatment was applied once to each clone; statistical inferences can be made from this design by using ANOVA models which include dummy variables for clone IDs and treatment labels. Clone 18 was assigned to 2 pots for treatment 2 and did not have a pot for treatment 5; and clone 27 was assigned to 2 pots for treatment 5 and did not have a pot for treatment 2. Thus, the N treatments were balanced (five pots for each treatment) but the clones were not perfectly spread across treatments (two clones did not have every treatment).

Each plot contained five pots, positioned on a 2.5 m x 1 m bench with a 1 m spacing between benches along the long side, with the short side facing the main windows (plants were grown using both natural light and additional LED light, as described above, to extend the daylight to 18 h). In total, the 25 plants were distributed within 27.6 m². Plant density was 0.9 plant/m².

For the duration of the trial, all plants were routinely moved (and randomized) so that their position in the glasshouse was rearranged every three weeks to neutralize the sun-edge effect that was observed in a previous trial³⁶.

All plants were irrigated with two drippers per pot using automated pumps that delivered a modified Hoagland solution (with adjusted N) six times a day with equal volume. This irrigation system ensured that the growing medium in each pot was never below 20% water content by volume. Pots were weekly flushed with tap water to prevent accumulation of nutrients in the media. The amount of nutrient solution provided (in mL, constant concentration) increased along with plant growth and the optimal range was adjusted to the plant need until there was about 10% leaching after every irrigation. At the beginning of the trial, each plant received 800 mL nutrient solution that was increased to 2200 mL after a month because of the rapid plant development and increased transpiration rate. This cultural management was maintained until harvest, 16 weeks after potting (8 weeks 18 h light, vegetative, and 8 weeks 12 h light, flowering). The temperature inside the glasshouse ranged between 18 / 28 °C.

Plant height, base stem diameter, and estimation of foliar chlorophyll levels (Konika Minolta SPAD-502, Japan) were measured during plant growth, starting from the 56th day after potting and approximately every three weeks after that until harvest. Plant height was measured as the distance from the base of the plant to the top of the main stem (inflorescence tip). Stem diameter was measured at 25 mm from the potting media surface with a digital calliper.

Biomass of the plant organs was assessed at the termination of the experiment, 16 weeks after potting. The plant was divided at mid-height and separated into upper and lower canopy to evaluate dry weight of inflorescences and leaves. All above-ground plant parts (inflorescences, stem, and leaves) were weighed after drying for 5 days at 50 °C to a constant weight in a ventilated oven.

This study analysed the effect of minor pollination on CBD and THC variations using statistical tools. Due to pollen contamination inside the glasshouse, some inflorescences were partially pollinated (this study analysed concentrations and biomass devoid of seeds). The extent to which plants were pollinated varied and both inflorescence biomass and seed biomass were calculated. For each sample we defined a “purity” measure, which is the proportion of ‘pure’ inflorescence biomass devoid of seeds: Purity = (inflorescence biomass devoid of seeds)/(inflorescence biomass including seeds)x100, with ‘100’ meaning inflorescences with no seeds.

Nitrogen elemental analysis

Total N and carbon were analysed at the Central Science Laboratory, University of Tasmania, using a Thermo FlashSmart Elemental Analyser (USA). Between 0.7 and 1.7 mg of samples were weighed into tin capsules using a Cubis II ultra-microbalance (Sartorius, Germany) with an accuracy of ± 0.1 µg. Combustion of the pressed tin cups was achieved in ultra-high purity oxygen at 1000 ºC using tungstic oxide on alumina as an oxidising agent followed by reduced copper wires as a reducing agent. The results were calibrated using a certified sulphanilamide standard.

Cannabinoid analysis

The dried inflorescences and leaves were ground manually to a homogenous mixture and about 5 g were pulverized using a ball mill. For each sample, about 100 mg of the pulverized plant material was placed in a 20 mL tube, 5 mL methanol (Sigma-Aldrich, USA) were added, and the tube was vortexed (Chiltern Scientific, UK) for 10 min at room temperature and then filtered with a 0.45 μm PVD filter. Cannabinoid concentrations in the filtered plant extracts were analysed using HPLC (Agilent 1260 Infinity II, USA), which consisted of a quaternary pump, an autosampler, a heated column compartment, and a Diode Array Detector. The detection was carried out in a spectrum mode, at the wavelength range 230 nm to 285 nm. Chromatographic separations were carried out with Agilent 2.7 μm InfinityLab Poroshell 3.0 × 100 mm 12 EC-C18 in methanol: water (0.1% formic acid in water, 0.05% formic acid in methanol), kept at a temperature of 50 °C. The sample injected was 3 µl and analysed at a gradient of MeOH: H2O (63:37) to 100% methanol over 20 min at a flow rate of 1.0 mL min⁻¹. Calculation of cannabinoid concentrations were based on pure analytical standards: cannabichromene (CBC), cannabichromenic acid (CBCA), cannabigerol (CBG), cannabigerolic acid (CBGA), cannabidiol (CBD), cannabidiolic acid (CBDA), cannabidivarin (CBDV), Δ9-tetrahydrocannabinolic acid (THCA), Δ9-tetrahydrocannabinol (THC), (Cerilliant, Texas, USA). No other cannabinoids were detected using HPLC analysis from the cannabis cultivar grown in this trial. R² values for linear regressions of the calibrations curves of all cannabinoid standards were > 0.995.

Cannabinoid concentration was calculated in percentage of dry weight (grams of cannabinoids/ grams of dry material x 100). Total CBD% (CBD% DW + CBDA% DW) was calculated with the following formula which takes into account the difference in molecular weight between CBD and CBDA⁴⁰:

All references to ‘CBD total’, ‘THC total’ and ‘N percentage’ mean percentage concentration of dry weight (all samples were freeze-dried prior to analysis). ‘Biomass’ is the dry weight in grams. And yields, which multiply concentrations by biomass, are referred to as ‘CBD yield’ and ‘THC yield’.

Statistical analysis and models

Models and correlations for data corresponding to inflorescences, leaves and total (inflorescences plus leaves) measurements were estimated. Outcomes analysed, including cannabinoid concentrations, plant N and biomass were presented as ‘y-variables’ with linear regression models that were analogous to ANCOVA set-ups. The ‘x-variables’ used in each model were baseline height measurements of the plants before treatment application, a dummy variable created to represent the clone genetics and a dummy variable for the treatment. Formally, each model structure can be summarised with the following equation. 

1

In which was the outcome (cannabinoid concentration, plant N or biomass) of plant (for ) which was in treatment group (for ) and clone group k(2, 13, 18, 26 or 27), was the baseline height of plant before any treatment was applied, was the effect of plant genetics for clone group k, was the effect of treatment , and was a random error term which we assumed to be distributed independently and identically Normal (with a mean of zero and constant variance for all ). The estimated parameters of the model were (the intercept), (the baseline covariate effect), (the ‘clone effect’), (the ‘treatment effect’ for treatment ) and .

Key model results (Fig. 1) used the average level of outcome variables predicted from the models based on Eq. (1), with the clone effect removed (in practice this entailed estimating the model and issuing predictions for each plant with set to zero for all k). The error bars in the chart represented the average absolute difference (above or below) between any two treatments that would be notionally “statistically significant” according to a Tukey multiple comparison contrast analysis within the model (the Tukey set-up focuses on unique pairwise assessments and uses a Bonferroni adjustment for the total number of contrasts, assuming a Type I error rate of 0.05). This enabled coherent comparisons between treatment effects on an easy to interpret scale.

Fig. 1.

N concentration, biomass, total CBD and total THC (%DW) in top and bottom inflorescences, leaves and total (average of inflorescences plus leaves). Each bar displays the average level of the outcome variable predicted from the corresponding model based on Eq. (1), with the clone effect removed. The error bars represent the average absolute difference (above or below) between any two treatments that would be notionally “statistically significant” according to a Tukey multiple comparison contrast analysis within the model (assuming a Bonferroni adjustment for the six treatments and a Type I error rate of 0.05). This enables coherent comparisons between treatment effects on an easy to interpret scale. For example, on the top left graph, the lower limit of the error bar for treatment 3 does not overlap with the top of coloured bar for treatment 2 in the same graph: thus the p-value is < 0.05 for the test of the null hypothesis that the difference in the effect of treatment 3 and treatment 2 on top inflorescence N concentration is truly zero. Stars indicate which treatments have statistically significant (p < 0.05) differences to treatment 1 (30 mg/L) in pairwise tests, after adjustment for multiple testing.

In addition to the models implied by Eq. (1), simple linear regressions of outcomes versus plant N concentrations were estimated. These models used the plant N concentrations that were measured in place of treatment dummy variables, and they took the following form:

2

in which is the outcome (cannabinoid concentration, yield or plant biomass) for plant (for ) which was in treatment group (for ) and clone group k(2, 13, 18, 26 or 27), was the baseline height of plant before any treatment was applied, was the effect of plant genetics for clone group k, was the recorded N percentage concentration (of dry weight) for plant , is a polynomial term to enable a non-linear relationship between N and the outcome, the and terms allow for interaction terms between clone k and N, and was a random error term which was assumed to be distributed independently and identically Normal (with a mean of zero and constant variance for all ). Equation (2) is also used to model y-variables corresponding to the CBD to THC ratio for the top inflorescence, bottom inflorescence, leaf and total levels of cannabinoids.

The results analysis includes categorical contrasts between the estimated clone effects from Eq. (1). Overall model fits in terms of adjusted r-squared values and F-test p-values were assessed and residuals checked for IID Normality.

Equation (2) was used to define the maximal complexity between N and the outcomes. According to a series of simplifications based on nested F-tests, only results for the minimal models are presented (for example in some cases there is evidence of polynomial relationship or interactions between N and clone categories).

We define “heritability” in this manuscript as the proportion of variance in a measurement that is attributable to clone effects, after accounting for the effect of nitrogen levels. Estimates for this heritability were made for the yield measurements of biomass (grams), CBD total (grams) and THC total (grams). The heritability estimates used linear mixed models with the same regression structure as Eq. (2), except that the clone x-variable was only included as a random effect and no interactions were used (from these linear mixed models the estimate of heritability was the variance of the random effect divided by the sum of the variance of the random effect and the variance of the residual).

All analysis was done with R (version 4.1.1), using the base installation packages. P-values were reported to three decimal places (or ‘<0.001’, as appropriate).

Results

Tables 1 and 2; Fig. 2 (alongside its details in Table S3) are the basis for formal statistical inferences about treatment and clone effects. The purpose of Fig. 1 is to identify trends and to enable ad-hoc pairwise comparisons between treatments and measurements (so the charts are uncluttered with preset “statistical significance” letters).

Table 1.

ANOVA model results from Eq. (1), in which the “treatment” results are for and the “clone” results are for .

	Upper inflorescence	Lower inflorescence	Leaves	Total	Total
	p-value	p-value	p-value	p-value	change in R^2
Clone ( )
Biomass (grams)	0.046	0.176	0.034	0.044	-2.5%
CBD total (%DW)	0.021	0.455	0.001	0.007	21.1%
THC total (%DW)	0.002	0.182	0.102	< 0.001	14.6%
Treatment ( )
Biomass (grams)	0.006	0.011	<0.001	< 0.001	78.3%
CBD total (%DW)	0.132	0.631	< 0.001	0.004	39.7%
THC total (%DW)	0.008	0.502	< 0.001	< 0.001	66.6%

Each model has a y-variable corresponding to the label in the row (Biomass, CBD total or THC total). Each column represents sub-sets of data (top and bottom inflorescence, leaf and the total of inflorescences and leaves) for which models are built. The x-variables in each model are baseline height (numeric), clone (categorical) and treatment (categorical). In each model n = 25, with 5 clones and 5 treatments. The p-values presented are from standard F-tests from a conventional ANOVA decomposition of the models (in which the null hypothesis is that the corresponding coefficients of the variable indicated are jointly zero); in the ANOVA test decomposition the clone variable is second last and the treatment variable is last. P-values are shown to three decimal places. The ‘change in R^2’ column, presented only for the models built for the total data set, is the marginal increase in adjusted R^2 made by the introduction of the variable (clone or treatment); for example, the introduction of the treatment variable to the ‘CBD total’ model (for the total data set) that already contains x-variables representing baseline height and clone, produces an increase in adjusted R^2 of 72.3%.

Table 2.

Clone contrasts (13 & 27 versus 18 & 26) for ANOVA models based on Eq. (1), using the total data set, using biomass, CBD total and THC total values as y-variables.

	Average clone difference (13 & 27) vs. (18 & 26)	95% CI Lower	95% CI Upper	p-value
Biomass (grams)	53.95	-21.04	128.94	0.145
CBD total (%DW)	2.65	1.36	3.95	0.001
THC total (%DW)	0.10	0.07	0.13	< 0.001

The x-variables in each model are baseline height (numeric), clone (categorical) and treatment (categorical). The contrasts being estimated and tested are based on a combination of clones. In each case, results for clone 13 and clone 27 are compared to results for clone 18 and clone 26. Formally, for each model the applicable null hypothesis is that true mean of ‘clone 13 and 27, combined’ is the same as the true mean of ‘clone 18 and 26, combined’ (i.e. that the difference in true means between these two groups is zero). The estimate column is the sample estimate of the average difference between these groups (e.g. the average CBD total concentration was 2.65 higher in clones 13 & 27, compared to clones 18 & 26), with the 95% CI for the difference in means being presented in the next two columns. The final column above is the p-value corresponding to the null hypothesis. A full set of individual clone-to-clone contrasts, including Bonferroni adjusted p-values, is provided in the supplementary material (Table S1).

Fig. 2.

N percentage (% DW) of biomass (x-axis) versus cannabis biomass, CBD and THC yield (grams), respectively. Colours identify clones. Treatment labels are given as adjacent text (N levels in terms of mg/L). The lines are estimated using models based on Eq. (2), but in each case lines are drawn using the maximally complex model (in terms of statistically significant x-variables from Eq. (2), according to p < 0.05), justified via a series of nested-model F-tests (see supplementary material, Table S3).

The effects of varying levels of N

Table 1 summarises the cannabinoid concentration and biomass of the models defined by Eq. (1). The N supply treatment variable explained 78.3%, 39.7% and 66.6% of the variance (in terms of increased r-squared, after accounting for clone effects) in total biomass, CBD total (CBD + CBDA, whole plant) concentration, and THC total concentration (THC + THCA, whole plant), respectively.

Figure 1 summarises the N% (dry weight), cannabinoid concentration and biomass results with respect to Eq. (1), but with the clone effect removed. The nitrogen concentration in plant tissues (observed N%), biomass, and cannabinoid concentrations were significantly influenced by the treatments as N supply increased. The N% in inflorescences was generally higher for all N treatments than the N% of leaves. Leaf N% increased with each treatment step, except at the highest application rates (treatments 4 and 5). The N% for treatments 1 and 2 was significantly lower than N% in treatments 3, 4, and 5 (p < 0.05) for both inflorescences and leaves. The difference in N% values induced by different N treatments were more pronounced in leaves than in inflorescences. Specifically (from the numbers underlying Fig. 1), N% of leaves for treatment 1 and 5 was 2.6% and 4.9% (an increase of 88%), respectively; N% of upper inflorescences for treatment 1 and 5 was 3.9% and 5.1% (an increase of 31%) respectively.

Figure 1 also clearly shows that leaf biomass increased with N supply. Upper inflorescence biomass was lowest at treatment 1 (30 mg/L of N) and peaked at treatment 3 (160 mg/L of N). The biomass of the bottom inflorescences, though relatively small compared to the top inflorescence biomass, increased as N supply increased.

Total biomass (inflorescences and leaves combined) increased by around 100% as N% increased from 3 to 5% (Fig. 2) while the concentration of CBD and THC in all plant organs tended to decrease as N% increased.

The variability of the data for inflorescence CBD and THC concentrations was higher than the variability in leaf CBD and THC concentrations, as shown by the relative widths of the error bars in Fig. 1.

Figure 2 summarises the cannabinoid yield and biomass analysis of the models defined by Eq. (2). The results summarised in Fig. 2 quantify the clear quadratic relationship that observed N% (dry weight) has with biomass and CBD total yield. There is also a notable absence of relationship between N% and THC yield (as N% increased, the drop in THC concentrations was offset by increased biomass). From the numbers underlying Fig. 2, averaged across clones, the highest CBD yield of 27 g per plant was reached with 160 mg/L N (treatment 3) which corresponded to 24 g of CBD per m², with a mean N% of 4.25.

Figure 3 shows the results of modelling with Eq. (2) when the y-variable is the CBD to THC ratio for the top inflorescence, bottom inflorescence, leaf and total levels of cannabinoids. Note that the ratio of CBD to THC is the same for both the concentration and yields (since biomass cancels in the case of yield figures). The lines in Fig. 3 show the predictions of models that were simplified on the basis of statistically insignificant x-variables (p > 0.05). In the case of the top inflorescence model, there was a p-value just slightly lower than 0.05 for the interaction between the square of N and clone but this was ignored, on the principle of parsimony. Overall, the total CBD to THC ratio increases as N percentage (% DW) increases. The leaf results demonstrate a similar linear increase to the overall data. In contrast, the top inflorescence results indicate a non-linear change, and the bottom inflorescence results indicate no change, in the CBD to THC ratio as N percentage increases.

Clone effects

Table 1 shows ANOVA results for biomass, CBD concentration, and THC concentration when clone and N treatment effects were modelled jointly. The clone effect on the performance of each plant was significant in terms of cannabinoid concentrations (p = 0.007 and p < 0.001; and an increase in adjusted r-squared of 21.1% and 14.6%; for CBD and THC, respectively). However, the biomass was not significantly influenced by clone (p = 0.044 but the increase in adjusted r-squared is slightly negative, indicating slight statistical significance but little or no practical effect). Estimates of heritability attributable to clonal effects for biomass and cannabinoid yields were as follows: 8.8% for Biomass (grams); 45.9% for CBD total (grams); and 52.9% for THC total (grams).

Figure 2 summarises the clone effect on N% (dry wight) with biomass and cannabinoid yields (concentration multiplied by biomass). For biomass there is a non-linear relationS1ship with N% and no clone effect; for CBD yield there is a non-linear relationship with N% and a constant difference in means between clones along different levels of N% (but no interactions between the slope and clone were detected – see Table S3); and for THC yield there is no detectable relationship with N%, but there is a difference in means between clones.

The practical significance of the individual clone effects can be seen in Fig. 2: CBD and THC mean values for clones 13 and 27 (red and orange lines, Fig. 2) were significantly different, in a practical sense, than mean values for clones 18 and 26 (purple and green lines, Fig. 2).

Table 2 describes pairwise differences between clones. In Table 2, the average clone differences for biomass and cannabinoid concentrations, between the set containing clones 13 and 27, and a set containing clones 18 and 26, indicate that CBD concentration and THC concentration were statistically different between these groups of clones, while biomass was not. For completeness all individual pairwise clone comparisons in terms of concentrations are provided in the supplementary material, Table S1.

Figure 3 indicates that, in top inflorescences the CBD to THC ratio had a non-linear relationship with N% and that the results differed among clones. In bottom inflorescences there was no relationship between the CBD to THC ratio and N%, nor any clonal effects. And for leaves and overall, whilst there was evidence of a positive linear relationship between the CBD to THC ratio and N%, there were no clonal differences.

Other results

The salient visual phenotypic effects of the nitrogen treatments were that the plants with the lowest levels of N (30 mg/L and 90 mg/L) displayed yellowing of foliage, reduced leaf area and reduced canopy volume (see Supplementary Figure S1 for photographs).

The shortest plants were ones that received 30 mg/L N while the tallest received 160 mg/L N. Above the 160 mg/L N application rate, plant heights did not increase; a similar pattern was found for base stem diameter. Chlorophyll concentration estimates (SPAD) increased with each increase in N application treatment. See supplementary Figure S2 for all these measurements.

Discussion

This study found that N application significantly affected CBD and THC concentrations and biomass in Cannabis sativa grown in a controlled environment and that clone ID significantly influenced CBD and THC concentrations, but not biomass. Overall plant biomass increased by 100% as N% DW increased from 3 to 5%, but at the same time cannabinoid concentrations decreased by approximately 25%. Treatments of 160 mg/L N or above had no significant impact on N% DW suggesting that not all N provided was taken up by the cannabis plants. Analysis ruled out significant confounding effects of baseline height, plant location in the glasshouse, and pollen contamination.

Leaves and inflorescence biomass doubled from treatment 1 to 5 but the increase was not to the extent of the 5-fold increase observed in a previous study with similar application rates³⁶. This can be related to the differences in the treatment initiation period. In the current study, different N treatments were initiated at the beginning of the flowering stage, which was preceded by eight weeks of full vegetative growth with a standard solution provided, during which each plant could attain full vegetative maturity. Conversely, in the previous trial³⁶, the period with standard nutrition was shorter (only four weeks) prior to treatment application rates being applied. The result was that plants that received lower N rates were smaller at the start of the flowering stage. This observation is consistent with a report of higher final biomass when higher levels of growth (bigger plants) were achieved prior to the initiation of the flowering phase⁴¹. In this trial, the height of plants that received more than 160 mg/L N showed a decreasing trend (Figure S2, supp info). This finding indicates reduced tolerance to higher N rates probably due to oxidative stress⁴² and increasing osmotic potential of leaf tissue sap, generating a salinity response which stunted plant growth⁴³.

Overall biomass increased steadily with increasing N application rates up to 160 mg/L N after which it did not change significantly. This N rate, optimal for biomass production, is similar to what other studies reported^12,13: in particular, inflorescence biomass peaked at 160 mg/L N with a decreasing trend above this value. Leaves and inflorescence biomass almost doubled, from low to high N treatments. This observation is consistent with other reports in the literature^13,14. While a lack of nitrogen is linked to reduced plant growth, reduced photosynthesis and leaf area^13,18,44, availability of nitrogen allows the synthesis of more N-containing molecules (e.g. chlorophyll, proteins) with consequent biomass production⁴⁵. Conversely, a different study reported a peak inflorescence biomass at 50 mg/L N after which it decreased¹⁴. The discrepancy in N supply for optimal inflorescence biomass may be due to differences in growing conditions and genetic material between the respective studies. Anderson¹⁴ used a peat growing media and because of the presence of mycorrhizae, a more efficient nutrient acquisition was possible^46,47 .

Accumulation of N in cannabis leaves when N supply increased was reported by Yang, Zha⁴⁸. In the current experimental trial, a similar trend was observed but for the two higher N treatments (4 and 5), more N supply did not translate into significantly higher N% in either leaves or inflorescences. This phenomenon is linked to reduced nitrogen use efficiency when N supply is high⁴⁹. With increasing N rates, cannabinoid concentration of the upper inflorescences showed a decreasing trend up to 160 mg/L N that was not statistically significant, although concentration in leaves was^48,49. Inverse correlation between concentration of secondary compounds and N supply is a common phenomenon observed not only in cannabis^12,36,50,51, but also in other crops grown for phytochemicals such as Stevia rebaudiana⁵² and Carlina acaulis^52–54 due to a dilution effect caused by biomass increases along with increased N supply.

In this experiment, the variation in both biomass and N% of inflorescences showed a narrower range compared to that observed in leaves across all treatments. This suggests that when N supply is low translocation of N from leaves to inflorescences occurs, so that nutrients are re-directed to more important reproductive organs. Plants have developed recycling mechanisms to guarantee plant fitness and survival when resources are scarce⁵⁵. This mechanism reflects the Optimal Defence hypothesis, which predicts that plant defences (and nutrients necessary to produce them) are concentrated in tissues and in organs that are the most valuable for survival and fitness⁵⁶.

During the course of this trial, it was observed that the upper inflorescence biomass and CBD% exhibited a response that followed a quadratic trend in relation to the N%. Conversely, the leaves and THC content displayed a linear response to changes in nitrogen levels. Similar trends, especially for biomass, were observed in other reported trials^12,13,57.

The mean value of CBD% in leaves and inflorescences was higher at low N rates than high N rates. When a higher rate of nitrogen is provided, a reduced concentration of secondary metabolites (per unit mass) is a common response^13,18,58. Other species have been reported to show a similar secondary metabolite concentration in response to N supply. For example, a study on oregano (Origanum vulgare subsp. Hirtum) by Ninou, Cook⁵⁹ reported that higher levels of applied N reduced essential oil content (mL 100 g⁻¹ DW) in all of the five strains investigated. Nonetheless, nitrogen fertilization increased essential oil yield through increased dry biomass production (+ 59%) compared to other treatments with lower N supply.

Different cannabis chemovars have specific CBD/THC ratios which are genetically determined⁶⁰. And the CBD/THC ratio has been reported to dynamically change over time when inflorescences mature^51,61; but to the best of our knowledge, CBD/THC ratio changes at harvest time in response to N supply have not been reported. Furthermore, in Fig. 2 CBD concentration is quadratic in response to N% while THC has a linear response. This trend is not clone specific, but it affects every clonal source.

The CBD/THC ratio in leaves and top inflorescences was smaller when plants receive a low dose of N than when they receive a high (though not excessively high in the case of top inflorescences) N supply. The reason for this phenomenon is not clear. However, nutrient availability causing changes in the relative ratios among different secondary compounds (i.e. linalool, β-myrcene, (Z)-3-hexenyl acetate, geranyl acetate) has been reported for other species, e.g. corn (Zea mays)⁵⁷. Other stressors (Ethephon, flooding, herbicide, powdery mildew, wounding) have been specifically tested to investigate CBD/THC ratio alteration, but the CBD/THC ratio was not found to be strongly influenced by environmental stress⁶². In contrast to Toth et al. (2021), our results indicate a strong influence of N deficit on CBD/THC ratio. From an ecological defence perspective this phenomenon could be interpreted as maximising investment into the psychologically active THC in order to protect from herbivory outcrossing potential and photosynthetic leaf area, respectively.

A major factor that significantly influenced cannabinoid concentration in this study was clone source. In particular, clones 13 and 27 produced CBD% and THC% that were consistently higher than other clones for every N treatment (in particular 18 and 26) (Fig. 2).

Our trial found that CBD% is positively correlated with THC%. This observation is in accordance with recent publications which attribute the presence of THC in high-CBD chemovar to the activity of CBDA synthase which is not 100% selective towards CBDA^62–64. Zirpel, Stehle⁶⁵ tested CBDA synthase selectivity in converting CBGA: these authors reported that the majority of CBGA precursor was converted to CBDA but they noted that about 5% of by-products were produced on average (including THCA and CBCA). This could explain why in our trial, for each N treatment, higher concentration of CBD was always linked to higher concentration of THC.

The better performance of some clones, in terms of higher cannabinoid concentrations, can be related to genetic heterozygosity of the cultivar employed in the trial. New cannabis cultivars are usually produced by bulk selection, hybrids (using two parent lines) or clones (cuttings). In bulk selection cannabis varieties are created from segregating genetically heterogeneous populations that meet selection criteria in order to enhance desired characteristics⁶⁶.

Cannabis sativa L. has a complex genetic structure across its global population⁶⁷, so ‘for comparative analyses [to improve upon academic research to date], a greater number of genetically well-defined hemp and drug-type Cannabis grown in parallel under standardized cultivation conditions is needed to identify traits of interest and obtain a functional understanding’⁶⁸. This study is a small contribution towards this goal: using a reproducible controlled environment, a wide range of nitrogen treatments and a targeted set of clones, the study produced results that will help plan future breeding programs and experiments; whilst it is acknowledged that many different cultivars and different growing conditions should also be examined.

Process-based modelling has been proposed as a formal tool to help with broader consideration of GxExM (Genotype-by-Environment-by-Management) interactions in plant breeding⁶⁹. To some extent, this approach has already begun to be deployed for Cannabis sativa^29,70. But in the future an even wider range of scenarios should be examined for at least the following three reasons. First, there is a significant impact of environment (for example annual temperature, altitude, and latitude) on CBD and THC, relative to cannabis germplasm collections⁷¹. Second, there is significant sensitivity in CBD and THC yields relative to the management options, such as the nitrogen treatments shown in this study (and this sensitivity may interact with the environment). Thirdly, this study suggests that even amongst apparently highly related clones, subtle but commercially relevant and management-dependant traits exist. In summary, plant breeding and related research disciplines should form multidisciplinary teams to improve the understanding of GxExM interactions⁶⁹, which in the case of Cannabis sativa means academics collaborating more with commercial partners from a rapidly expanding industry.

The current study found a total of about 300 g per plant of dry (upper + lower) inflorescence biomass (this was generally one order of magnitude higher than other studies) when 160 mg/L N were applied^11–13. This is a significant increase (about 300% more biomass) relative to a previous trial³⁶ when 15 L pots were used at a higher density per unit area (relative to the 35 L pots and lower plant density used in the present study) that developed less than 100 g per plant of dry inflorescences. Other trials on medicinal cannabis have reported much lower values ranging from 10 to 80 g per plant^11–14. All these studies have used a variety of different nutritional solutions and genetics, though pot sizes were notably very different: in this study 35 L pots were used while other authors used pot sizes from 3 to 6 L. Even higher inflorescence yields (> 700 g/plant) were also achieved by Yang, Berthold⁵¹ in an open field trial. The inflorescence biomass of the highest performer cultivar (Cherry Blossom) reached a CBD yield of about 82.7 g per plant while in this trial it only reached 27 g per plant. If the yield per unit of area is considered, Yang, Berthold⁵¹ reported 29.7 g per m² (0.36 plants/m²) which is comparable to the CBD yield of this trial (24 g/m² in 0.9 plants/m²). Poorter⁷² concluded that on average, a doubling of the pot size increased biomass production by 43% and root restriction in pots reduces net photosynthesis^73,74. This seems to confirm that non-root-constricted plants can produce higher flower yields.

A research study highlighted the difference in organ and location-specific cannabinoid accumulation in a high THC cultivar⁷⁵: it also reported higher CBC levels in the leaves at the top of the plants than flowers. The current study confirms that there is spatial variation in cannabinoid concentration: upper inflorescence biomass and cannabinoid concentration were considerably higher than the lower inflorescence across all N treatments. N rates did not alter cannabinoid concentration differences linked to spatial variation, in contrast to our hypothesis.

This study found that minor pollination of some inflorescences had no significant practical effect on cannabinoid concentration but there is a significant influence on inflorescence biomass and overall yields (this study analysed concentrations and biomass devoid of seeds). The extent to which plants were pollinated varied and both inflorescence biomass and seed biomass were calculated. For each sample we defined a “purity” measure, which is the proportion of ‘pure’ inflorescence biomass devoid of seeds: Purity = (inflorescence biomass devoid of seeds)/(inflorescence biomass including seeds), with ‘100’ meaning inflorescences with no seeds. The ‘purity’ sample average was 89 and standard deviation was 9. It could be assumed that when a cannabis plant is pollinated, internal resources are allocated to seed development at the expense of biosynthesis of cannabinoids. Indeed, a study reported a significant reduction in inflorescence concentrations of most cannabinoids (with some exceptions) after pollination⁷⁶. Another study⁷⁷ specifically investigated the difference in cannabinoid level between pollinated and unpollinated Cannabis sativa L. cv. Finola plants. The inflorescence biomass of pollinated plants was approximately double the unpollinated, but the difference of the total dry floral seed-free biomass was not significant. CBDA level in pollinated plants was negatively affected. The authors concluded that the loss of cannabinoids is likely proportionate to the extent of flower pollination. Conversely, in this paper no practically significant reduction in cannabinoid concentration was observed for lower purity levels (Table S2 in the supplementary material), probably due to low levels of pollination. The only statistically significant effect of pollination (p = 0.033) is a slightly higher biomass of upper inflorescences for higher purity levels (Table S2 in the supplementary material) and in accordance with Todd, Song⁷⁷ which reported higher biomass of unpollinated vs. pollinated inflorescence but the values were not significant because of high standard deviation).

This study found that, whilst biomass was not affected by clone ID, CBD and THC concentrations (% DW) and yields (grams) were affected by clone ID (Table 1; Fig. 2). The heritability estimates for total yields (grams) concurred: the heritability estimate for biomass (8.8%) was statistically insignificant (p > 0.05, according to a likelihood ratio test), whereas the heritability estimates for CBD yield (45.9%) and THC yield (52.9%) were significant in both statistical (p < 0.05) and practical terms. Furthermore, the effects of clone ID on CBD yield and THC yield were very similar: the clones with higher CBD total (grams) values were the same clones with higher THC total (grams) values (Fig. 2); this feature meant that there was no overall clone ID effect evident in the CBD to THC ratios (Fig. 3, chart for total). Finally, there was no significant interaction between clone ID and N (% DW), which meant that the differences between clone ID groups for CBD yield and THC yield appeared to be very similar across treatments (Fig. 2). Because of the low sample size, the experiment here should be regarded as a pilot study; but alongside a larger field trial, a more detailed genetic exploration^78–80 of these sorts of clones could yield illuminating information for a breeding program focused on CBD, THC and the ratio between these compounds.

Conclusions

This study found a clear effect of N treatment and clone source on biomass and cannabinoid production of medicinal cannabis. This research highlights the importance of precise N supply if the aim is to maximize cannabinoid yield. The significant effect of clone ID on CBD and THC concentrations found in this study underlines the importance of selective breeding for more abundant production.

Key findings of this study are that (i) upper inflorescence biomass was strongly and positively related to N supply in a quadratic relationship, (ii) cannabinoid concentrations were strongly and negatively related to N% in plant organs, (iii) clones from different seeds had higher cannabinoid concentrations, but not biomass, independent of N treatments, (iv) plant biomass was mainly determined by the N supply, (v) low-level pollination did not affect biomass of inflorescences nor cannabinoid concentration in seed-cleared inflorescences following unintended pollen contamination, and (vi) N supply, linked to N%, had the effect of changing CBD/THC ratio at harvest time.

This study confirms the data reported in a previous publication that baseline heights have little influence on final biomass and/or cannabinoid production in medicinal cannabis. It also confirms the biomass and cannabinoid trend with increasing N rates, but additionally identifies 160 mg/L N as the optimal N level for maximum yield. An in-depth study on the influence of different pot size, vegetative and flowering time on cannabis would be important to understand the best settings in a commercial growing environment for maximum yield.

In addition, our study found that the CBD/THC ratio in cannabis did not remain constant at harvest time but was increased with increasing N supply. Therefore, it can be suggested that cannabis producers can use fertilization as a tool to stay compliant with the legal THC thresholds, exploiting the fact that THC decreases more than CBD with increasing N doses, while simultaneously increasing biomass.

It is notable that in this study, some clones (same chemovar) had a significant higher cannabinoid concentration, irrespective of the N treatment, compared to others. N supply was the only factor found that could influence biomass production. This phenomenon can be used by plant growers for breeding program and genetic improvement of superior clones which deliver higher performances.

In conclusion, this study highlights the importance of precise agronomic practices to boost cannabinoid levels. Further studies on the interaction between N supply, the choice of clone and adequate pot size is warranted because it can potentially increase cannabinoid yield and growers’ profit margin.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Enrico Dilena: Conceptualization, Methodology, Validation, Investigation, Writing – Original Draft. Dugald C. Close: Conceptualization, Methodology, Writing - Review & Editing, Supervision, Funding acquisition. Ian Hunt: Writing - Review & Editing, Formal analysis, Supervision.

Data availability

Data is provided within the manuscript or supplementary information files.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval

The research was conducted under a medicinal cannabis licence and permit issued to Martha Jane Medical by the Australian Office of Drug Control and an industrial hemp licence issued to the University of Tasmania under the Tasmanian hemp legislative framework. The plant collection and use were in accordance with all the relevant institutional, national, and international guidelines and legislation. Permissions or licenses were obtained for the collection of plant material, and the voucher specimen was stored at the University of Tasmania under licence number IHR081. All the original plant material (seeds) used in this research was provided by its commercial owner, Martha Jane Medical, Australia.