While discrete regulatory mechanisms have been identified, a unified model for the transcriptional activation of photoperiodic flowering in short-day (SD) plants has not been delineated.

Although the GIGANTEA-CONSTANS-FLOWERING LOCUS T (GI-CO-FT) pathway appears to be highly conserved, each species may differ in its blueprint for activation of FT and floral meristem

identity. In cannabis, an SD plant, PSEUDO-RESPONSE-REGULATOR 37 (PRR37) has a causal effect in determining photoperiod sensitivity. This study identifies a network of WITH NO LYSINE (K)

kinases (WNK) which may be regulating the PRR proteins and downstream genes, including florigen (FT). CO-like genes have been identified in cannabis, but their function in regulating FT has

inversely related with expression of FT. Building on these insights, we propose a model for the CDL-gated regulation of FT expression in cannabis.

Plants recognize sunlight via specialized molecules in the cytosol called photoreceptors1. There are several classes of photoreceptors including phytochrome, cryptochrome, phototropin,

ZEITLUPE (ZTL) and FLAVIN BINDING KELCH REPEAT F-BOX 1 (FKF1). These molecules control genes which determine the photoperiodic flowering response, circadian rhythms, photosynthesis,

photorespiration, cellular respiration at night, and more2,3.

When the phytochrome chromophore is irradiated by red light, it moves into its active conformer, which interacts with phytochrome interacting factors (PIFs) to modify gene expression3.

Far-red light, however, rapidly converts the active conformer Pfr to the inactive conformer Pr; darkness also converts Pfr to Pr, but more slowly4. During the day, most but not all the

phytochrome is activated, and the ratio of Pr to Pfr will dictate certain physiological responses, including the shade avoidance response, which is induced due to the nature of shady light

to contain a high ratio of far-red light5. This is because far-red light, having a longer wavelength, can penetrate more deeply through the canopy. The ratio of Pr to Pfr is known as the

PSS, or Phytochrome Stationary State6. It has been shown that the PSS can be predicted based on the ratio of red to far-red light that is being received by a plant tissue7. Thus, we can

model PSS in a leaf tissue given the following formula:

The entrainment of photoperiod pathway genes through uninterrupted skotoperiod length is the primary signal for the initiation of photoperiodic flowering, which takes effect through

increased expression of FT in the leaves8. Cannabis is a highly polymorphic single species; there are subspecies of cannabis that are photoperiod insensitive, and others which flower under

short days: however, no cannabis plants flower exclusively under long days9,10. Cannabis is a short-day photoperiodic flowering plant and although sexually mature flowers form prior, it

produces its inflorescences when the critical day length is under 12–14 hours10. Photoperiod sensitive cannabis tends to flower when the night length exceeds ten hours. Although we refer to

critical day length (CDL) instead of critical night length, It is important to understand that photoperiodic flowering depends not on the constancy of the photoperiod, but on the constancy

of the dark period11. The critical day length at which cannabis begins to flower is a key determinant in success when it comes to cultivation, because the longer the CDL, the sooner in the

season flowers emerge12, which guarantees that the flower can ripen before cold weather in the northern latitudes. Therefore, understanding the genetic mechanisms which trigger CDL-gated

flowering activation cascades is critical for cannabis flower production.

The signaling cascade which initiates the photoperiodic flowering response is well characterized in long day (LD) plants13. In Arabidopsis, a model LD plant, CO acts as a central hub for

integrating environmental signals to determine flowering time14. The FT gene is activated in leaves via the sequential activation of GI and CO15, which are regulated by both phytochromes and

cryptochromes16. Upon entering long days, the FT protein is positively regulated and translocated to the meristem, where it works with FD and SUPPRESSOR OF CONSTANS 1 (SOC1) to activate

floral meristem identity, which initiates the development of flowers13,17.

The CCT (CONSTANS, CONSTANS-like, TOC1) domain or CONSTANS domain is a highly conserved genetic feature common to CONSTANS genes, among others. The encoded domain enables protein–protein

interactions with TFs like Nuclear Factor Y (NFY) that directly bind the CCAAT box in the FT promoter18. Although it has been shown that in Arabidopsis the CCT domain of the CONSTANS protein

is responsible for FT activation, pseudo-response regulators like those in rice and cannabis possess CCT domains as well19. This group of pseudo-response regulators have been shown to be

substrates for phosphorylation by WITH NO LYSINE (K) (WNK) kinases, which are regulated by circadian rhythms20.

Like Arabidopsis, rice (an SD plant) possesses a CO-like gene, but this gene is known as HEADING DATE 1 (HD1)21,22. Unlike in Arabidopsis, HD1 in rice is not a central hub for signal

integration, but rather a bifunctional regulator which performs different functions based on the presence of multiple interactor proteins22. Soybean is also an SD plant, and has a system of

activating FT expression based on CDL gated genes and circadian rhythm entrainment23. A group of “E genes” and GmELF3 regulate the photoperiodic flowering response in soybean. The E2 gene is

an ortholog of GI. Not dissimilarly to rice, GI regulates the expression of FT in soybean, inhibiting its expression during long days24. Although several models have been proposed to

explain the protein–protein and protein-DNA interactions which control SD flower initiation, a conclusive model has not been established21. The known mechanisms of SD flowering initiation in

rice and soy provide clear insight into the potential mechanisms for initiation of flowering in cannabis.

Cannabis CONSTANS-like (CO, COL, CO-like) genes have been identified, and there are at least 13 expressed CONSTANS-like genes25. However, no evidence has demonstrated that CO is required for

FT activation. In cannabis, florigen-like, PRR37-like, and SOC1-like genes have also been identified19,26, and CsPRR37 has been shown to be causal to and required for the activation of FT,

whether or not it requires the presence of CO. It has been shown that a truncation of the CsPRR37 protein, removing the CCT domain, is correlated with the autoflowering trait, or lack of

response to photoperiod19.

COL genes have varying expression levels in different tissues. Seven (4, 8, 1, 6, 11, 9, 12) are related and highly expressed in leaves, three are related and expressed in leaves and flowers

(7, 5, 10), one is highly expressed in stems (13), and two are highly expressed in flowers(2, 3)25. COL3 and COL7 exhibit amino acid differences among early flowering and late flowering

cultivars25. It has been shown that all the cannabis CO genes have diurnal expression patterns in both LD and SD conditions25, however no data showing the expression of these genes during SD

transition has been published. There seems to be a dichotomy in CO genes regarding expression levels in early and late-flowering genotypes, where CsCOL4 and 11 expression levels are higher

in early flowering genotypes, and CsCOL6, 7, 9 and 12 are lower. “This indicates that there may be multiple CsCOL genes functioning as promoters or suppressors of flowering to regulate

flowering time in C. sativa”27.

Cannabis FT (florigen) has been identified, and sequence variations in the FT protein have been associated with the photoperiod sensitive/insensitive phenotype26. There are 4 known FT-like

genes in the cannabis genome, which are part of a larger family of PHOSPHATYIDYLETHANOLAMINE BINDING PROTEIN (PEBP) genes involved in flowering. These include three closely related to MOTHER

OF FT (MFT), two related to TERMINAL FLOWER (TFL), and three BROTHER OF FT (BFT)/CEN genes. In solitary flowers in vegetative plants, six PEBP genes were differentially expressed. In the

reproductive nodes, the three CsBFT/CEN, CsTFL and CsMFT3 genes exhibited reduced expression, while the CsFT4 gene expression increased27,28.

Although the GI-CO-FT photoperiod-dependent flowering pathway appears to be highly conserved across different crops including cannabis24,29, the mechanisms and functions of each component of

the signaling pathway can vary. Understanding the specific mechanisms of this pathway is an ideal path for elucidating the regulation of flowering time in cannabis27.

Although the inductive, short photoperiod “flowering” phase is when the inflorescence forms, and is also when the majority of flowers are produced, it is not the point of sexual maturity30.

The cannabis plant produces solitary flowers at the leaf and stem axis after reaching the mature vegetative phase (non-inductive), and these flowers are in fact sexually mature and capable

of producing seed. These solitary flowers can emerge at the 4th–6th internodes (Cervantes), and transcriptomic results have shown that several floral induction genes are upregulated,

including MFT, SOC1, LEAFY (LFY), and APETALA (AP1), while several repressors TEMPRANILLO (TEM), TFL1, and BFT were downregulated28. SPL genes were also activated.

The age-dependent flowering pathway in cannabis has been elucidated28 and mechanisms homologous to Arabidopsis have been identified, including miRNA31. The function of miRNA172 and 156 in

regulating flowering genes is one of the most highly conserved mechanisms in the plant kingdom. As such, it is not unreasonable to assume that these miRNAs are having a significant impact on

cannabis flowering.

One important mechanism for age-dependent flowering in Arabidopsis are “count-down” and “count-up” miRNA timers. miRNA156/157 act as “count-down” timers, meaning their expression decreases

as the plant ages, and miRNA172 expression increases as the plant ages. When the expression levels of these genes reach critical thresholds, they initiate flowering32,33. Target genes for

miRNA156/157 include SPL genes, which are critical in regulating the transition from the juvenile to the mature phase32,34. miRNA172 affects the AP2-family TF’s, which are also flowering

repressors. Therefore, the increased expression of miRNA172 results in the decreased expression of these flowering repressors, which initiates flowering35,36.

Other miRNA’s have been shown to function as flowering mechanisms in rice, where differentially expressed osa-miR171 and osa-miR1432 were associated with effects on yield37. A group of

microRNA’s including csa-miR156 and csa-miR172 have been identified in the cannabis genome, and csa-miR156 has been shown to target SPL genes while csa-miR172 targets transcription factors

that regulate flowering time31.

The differential expression analysis below displays the fold change in gene expression of several different photoperiodic-flowering related genes in cannabis leaves. The first sample set

represents LD conditions (last long day), while the second sample set represents SD conditions (seventh short day). Thus, these data demonstrate gene expression changes in cannabis leaves

across SD transition. Multiple known genes were identified, and their CDL-gated response characterized, while several other genes were uncovered, homologous to genes in soy and Arabidopsis.

The sequenced transcriptomes were mapped to nine different reference genome assemblies due to what transcripts are defined directly affects what alignment and quantification is possible with

short read- RNA-Seq. After aligning the sequenced reads using Salmon, the mapping rate ranged from 20 to 70% with cs10 achieving the lowest mapping rate at 25% (supplemental Fig. 1).

Supplemental Fig. 2 displays an alternative volcano plot for the analysis. Supplemental Figs. 3 and 4 display the effect of shrinkage. The highest mapping rate was achieved by Jamaican Lion,

Purple Kush, and CBRX18 at ~ 70%. Out of those three transcriptomes, Jamaican Lion38 had the highest number of annotated transcripts, which we continued to use for the study. To identify

significant differentially expressed genes (DEGs), we used DESeq to identify thousands of relevant targets with a wide range of annotation on function.

Differential Expression Analysis. Panel (A): Experimental design. Panel (B): Scatterplot: red dots indicate up regulated genes, while blue dots indicate down-regulated genes. The X-axis

indicates the log2 fold change, either negative or positive. The Y-axis indicates the log10p-values associated with each gene’s fold change. Greater values indicate higher confidence.

CsCOL10 was downregulated with a high level of confidence, while HD3A, FT and VRN are upregulated with a high level of confidence.

qPCR data describing the fold change in gene expression in cannabis leaves across SD transition. Four biological replicates per assay, from low RFR group. UB used as reference, delta-delta

Ct method used to calculate relative expression change. In each assay, the last long day is set as control (1). For all experiments, student’s t-test was utilized for calculating statistical

significance of means separation. One star (*) indicates a P-value below 0.05, while two stars (**) indicates a P-value below 0.005. Panel A, CO-like genes: All but one CO-like genes were

downregulated promptly by the third short day, which persisted into the seventh short day. COL9 expression remained relatively unchanged, which is characteristic of AtCO. The greatest fold

change was exhibited by CsCOL10. Trend line indicates average fold change across all COL genes. Individual expression values across all biological replicates and genes used to calculate fold

change, standard error and significance. Overall, the CO-like gene family in cannabis was down-regulated significantly upon SD transition. A one-way ANOVA comparing LLD to SD7 revealed

significant differences in gene expression between groups across CO-like genes (F(1, 44) = 14.81, p