TOC1 in Nicotiana attenuata regulates efﬁcient allocation of nitrogen to defense metabolites under herbivory stress

(cid:1) The circadian clock contextualizes plant responses to environmental signals. Plants use temporal information to respond to herbivory, but many of the functional roles of circadian clock components in these responses, and their contribution to ﬁtness, remain unknown. We investigate the role of the central clock regulator TIMING OF CAB EXPRESSION 1 (TOC1) in Nicotiana attenuata ’s defense responses to the specialist herbivore Manduca sexta under both ﬁeld and glasshouse conditions. We utilize 15 N pulse-labeling to quantify nitrogen incorporation into pools of three defense compounds: caffeoylputrescine (CP), dicaffeoyl spermidine (DCS) and nicotine. (cid:1) Nitrogen incorporation was decreased in CP and DCS and increased in nicotine pools in irTOC1 plants compared to empty vector (EV) under control conditions, but these differences were abolished after simulated herbivory. Differences between EV and irTOC1 plants in nicotine, but not phenolamide production, were abolished by treatment with the ethylene agonist 1-methylcyclopropene. Using micrografting, TOC1 ’s effect on nicotine was isolated to the root and did not affect the ﬁtness of heterografts under ﬁeld conditions. (cid:1) These results suggest that the circadian clock contributes to plant ﬁtness by balancing production of metabolically expensive nitrogen-rich defense compounds and mediating the allocation of resources between vegetative biomass and reproduction.


Introduction
Because they are both sessile organisms and at the base of most food chains, plants are under strong pressure to produce effective defenses in the face of mobile herbivores. Several theories have been developed in an attempt to quantify plant defense responses as a function of different factors, in order to explain and predict patterns observed under natural conditions, with varying degrees of success (see . Many of these theories attempt to determine trade-offs between growth and defense (Herms & Mattson, 1992), both of which depend on the same limited resources (Bryant et al., 1983;Chapin et al., 1990) and can be alternatively prioritized among plant tissues with different fitness values (McKey, 1974(McKey, , 1979Rhoades, 1979). Generating testable predictions from these theories has proven challenging largely because they address different levels of analysis, from mechanism to function (Tinbergen, 1963;Sherman, 1988), and the proposed mechanisms are not grounded in plant physiology. Furthermore, most plant defense theories do not consider the definition of defense as the contribution of a trait to the Darwinian fitness of plants under attack (Karban & Baldwin, 1997;Stamp, 2003;Massad et al., 2012;. One component is the importance of timing to the function of plant defense. Herbivory occurs unpredictably over a plant's lifetime, and plant defense responses are highly plastic and often elicited by herbivory Wu & Baldwin, 2010;Meldau et al., 2012;Wagner & Mitchell-Olds, 2018). Herbivore feeding behavior can be diurnally rhythmic, while plant defenses are likely constrained by the circadian clock in order to maximize resource use (Goodspeed et al., 2012(Goodspeed et al., , 2013b. The plant circadian clock is an endogenous mechanism that touches upon nearly every aspect of plant metabolism (Greenham & McClung, 2015;Nohales & Kay, 2016;Sanchez & Kay, 2016). As such, individual components of the clock have been shown to modulate responses to a variety of external stimuli and stresses, including herbivory (Takase et al., 2013;Seo & Mas, 2015;Sharma & Bhatt, 2015;Joo et al., 2018;Zhang et al., 2019).
Despite evidence for the clock's role in modulating responses, few studies have adressed its significance for maintaining plant fitness in natural, diurnal environments. The circadian clock component GIGANTEA (GI) maximizes the fitness of Oryza sativa sown at variable times under field conditions (Izawa et al., 2011;Izawa, 2012), while the central circadian clock component TIMING OF CAB EXPRESSION 1 (TOC1), which modulates a variety of growth and stress responses (Somers et al., 1998;Legnaioli et al., 2009;Soy et al., 2016;Zhu et al., 2016), enhances fitness during drought stress in the field (Valim et al., 2019). Although there have been few studies tying clock components to fitness outcomes under biotic stress, the clock is known to regulate responses to herbivory. Basal levels of the phytohormone jasmonic acid (JA), which is involved in regulating downstream responses to biotic stresses, are under circadian regulation in Arabidopsis (Goodspeed et al., 2013a). Jasmonic-acid-induced responses to herbivory are known to be under diurnal regulation in other plant species, with darkness affecting JA accumulation. However, whether JA accumulation is under circadian regulation remains unknown (see e.g. Phaseolus lutatus; Arimura et al., 2008). In the desert annual Nicotiana attenuata, performance of the herbivore Manduca sexta is similarly dominated by diurnal rather than circadian patterns (Herden et al., 2016), and several JA-induced secondary metabolites are under diurnal regulation (Halitschke et al., 2000;Kim et al., 2011;Gulati et al., 2014;Joo et al., 2018). However, the clock component Zeitlupe (ZTL) does regulate the JA-mediated induction of nicotine in N. attenuata (Li et al., 2018).
Jasmonic-acid-induced herbivory responses can be specific and are mediated by interactions with other phytohormone pathways. In the wild tobacco species Nicotiana sylvestris and N. attenuata, oral secretions of M. sexta dampen the JA-elicited increase in nicotine observed after elicitation with mechanical wounding or the application of exogenous JA Kahl et al., 2000). Fatty acid-amino acid conjugates (FACs) in the oral secretions of M. sexta are necessary and sufficient to dampen this nicotine response, and function by eliciting an ethylene burst Halitschke et al., 2001;Von Dahl et al., 2007). Ethylene signaling suppresses JA-induced nicotine production in Nicotiana spp. by suppressing expression of ethylene response factor (ERF) transcription factors that bind to the promoter regions of nicotine biosynthesis genes such as putrescine N-methyltransferase (PMT), and the ethylene burst leads to large-scale reconfiguration of transcriptomic responses to herbivory (Shoji et al., 2000(Shoji et al., , 2010Voelckel & Baldwin, 2004). Ethylene responses are under circadian regulation in Arabidopsis, and TOC1 silencing alters the rhythmicity of ethylene emissions (Thain et al., 2004). Nonetheless, it is unknown whether any functional benefits of rhythmic ethylene emissions are directly mediated by the circadian clock.
Although nicotine provides Nicotiana spp. with a potent defense in nature (Steppuhn et al., 2004), plants elicited by the nicotine-tolerant specialist M. sexta do not increase nicotine production, and subsequently experience lower fitness costs of elicitation than plants elicited by exogenous JA treatments or when ethylene signaling is blocked, leading to high nicotine accumulation after elicitation (Voelckel et al., 2001). This may be because plants mobilize their limited resources to more effective defenses, given nicotine's high metabolic cost, and/or because nicotine may be co-opted by M. sexta larvae as a defense against parasitism and predation (Thorpe & Barbosa, 1986;Kumar et al., 2014). Nicotiana attenuata contains many secondary metabolites of defensive value, with the phenolamides caffeoylputrescine (CP) and dicaffeoyl spermidine (DCS) effectively limiting M. sexta performance (Kaur et al., 2010). CP and DCS are highly abundant in M. sexta-elicited plants and are correlated with elicited jasmonyl isoleucine (JA-Ile) contents in leaves of natural accessions .
Transcriptional activation of MYB8, the key regulator of phenolamide biosynthesis, requires functioning JA signaling (Onkokesung et al., 2012). Silencing MYC2, the primary transcription factor at the interface of JA biosynthesis and downstream signaling, does not lead to decreases in CP and DCS after elicitation, suggesting an alternate JA-induced pathway independent of the COI1-mediated JAZ/MYC2 module that regulates nicotine biosynthesis (Chini et al., 2007;Thines et al., 2007;Sheard et al., 2010;Woldemariam et al., 2013). Nicotine biosynthesis is also directly regulated by an interaction between the circadian clock component ZTL and a JAZ/MYC2 module independently of COI1, while phenolamide levels are not strongly affected by silencing ZTL (Li et al., 2018). Phenolamides and nicotine are nitrogen (N)-rich and thus presumably metabolically expensive secondary metabolites, and efficient maintenance of N. attenuata's N composition is critical to its survival in environments characterized by high intraspecific competition Ullmann-Zeunert et al., 2012. Thus, we sought to examine the role of the circadian clock in the regulation of these defense metabolites at the intersection of JA and ethylene signaling, in addition to examining the implications of this role on plant fitness under field conditions by abrogating TOC1 expression.
Abrogating TOC1 expression via RNAi-mediated silencing leads to broad changes in plant responses to simulated herbivory. We observed that TOC1 silencing leads to delays in jasmonatedependent signaling after elicitation with M. sexta oral secretions, consistent with previous findings linking the circadian clock component ZTL to the jasmonate-mediated regulation of nicotine biosynthesis in N. attenuata. Interestingly, we also observed significant decreases in production of CP and DCS, and increased rather than decreased nicotine production. We performed 15 N pulse labeling experiments to ask whether N incorporation into phenolamide and nicotine pools is changed by elicitation, and observed that N introduced as 15 NO 3 À is preferentially incorporated into nicotine pools under control conditions in irTOC1 plants, while allocation becomes equally distributed between phenolamides and nicotine after elicitation with M. sexta oral secretions. These findings are consistent with previous evidence that herbivory disrupts the rhythmicity of the clock , and that nicotine pools in other Nicotiana species are homeostatic and do not function as sources of N .
Because irTOC1 plants overproduce nicotine, contrary to the phenotype which would be predicted by comparison with irZTL plants, we asked whether ethylene signaling may be compromised in irTOC1 plants. Although the ethylene burst is not different in irTOC1 plants, differences between empty vector (EV) and irTOC1 plants in nicotine production are abolished by treatment with 1-methylcyclopropene, an ethylene signaling receptor New Phytologist (2020) 228: 1227-1242 Ó 2020 The Authors New Phytologist Ó 2020 New Phytologist Foundation www.newphytologist.com

Research
New Phytologist blocker, pointing to an attenuation of the ethylene signaling feedback loop (Von Dahl et al., 2007). This effect was specific to nicotine, as differences in phenolamide accumulation remained largely unchanged. Finally, we find that restoring TOC1 function in shoots, but not in roots, via a seedling micrografting approach is sufficient to decouple differential phenolamide and nicotine accumulation and the fitness disadvantages of whole-plant irTOC1 silencing under field conditions, leading to plants with high levels of N-rich defense compounds and uncompromised fitness relative to EV plants. These results indicate tissue-specific roles for TOC1 in the production of metabolically expensive Nrich defense compounds via an interplay between JA and ethylene signaling.

Plant materials and constructs
Lines were derived from seeds originally collected from natural populations of Nicotiana attenuata Torr. ex S. Watson from the Desert Inn Ranch near Santa Clara, UT, USA . Seed germination and plant growth in the glasshouse were carried out as described by Kr€ ugel et al. (2002) and undertaken in a glasshouse in Jena, Germany, under a 16 h : 8 h, light : dark photoperiod. Screening of the EV line (pSOL3NC, line no. A-04-266-3) for a single homozygous transgene insertion and absence of off-target effects is described by Bubner et al. (2006). Screening of TOC1-silenced (ir) lines (pSOL8_16844, A-11-205-4) via RNAi is described by Yon et al. (2012), with additional screening described in Valim et al. (2019). Unless specified in figure captions, the irTOC1-205 line was used for all experiments described in this study. Importation and release of transgenic plants were carried out under Animal and Plant Health Inspection Service (APHIS) import permit nos. 07-341-101n (EV) and 11-350-102m (irTOC1) and release permit no. 16-013-102r. Field growth conditions were described by McGale et al. (2018). Briefly, seedlings were germinated on Gamborg's B5 media (Duchefa, Harleem, the Netherlands) under illumination from fluorescent lights with a 16 h : 8 h, light : dark photoperiod (GE Plant & Aquarium 40 W and GE Warm White 18 W, Cleveland, OH, USA) at ambient temperatures at the field station. One week after germination, seedlings were grafted (see section 2.3, Micrografting, below, and Fragoso et al. (2011). One to two weeks after grafting, seedlings with four visible leaves were transferred into previously hydrated 50-mm peat pellets (Jiffy 703; Jiffy Products of America Inc., Lorain, OH, USA) treated with borax (sodium tetraborate; Sigma-Aldrich Chemie Gmbh, Munich, Germany) to provide boron, an essential micronutrient (1 : 100 dilution of a 1.1 g l À1 stock solution), and adapted over 2 wk to the field conditions of high light intensity (over 2.0 mmol m À2 s À1 of photosynthetically active radiation; see Joo et al., 2017) and low relative humidity (often 8% relative air humidity during the day during planting; see Nicotiana Attenuata Data Hub: http://nadh.ice.mpg.de/NaDH/fielddata/ weather) by keeping seedlings first in shaded, closed translucent plastic 34-quart boxes (Sterilite, Townsend, MA, USA), then opening the boxes, and subsequently transferring open boxes to partial sunlight in mesh tents (Tatonka, Dasing, Germany). Adapted size-matched seedlings were transplanted into an irrigated field plot at the Lytle Ranch Preserve, Santa Clara, Utah, in April 2016, and fitness correlates and biomass were counted at harvest in June 2016.

Simulated herbivory treatments
Manduca sexta L. oral secretions (OS) were collected as described in Voelckel et al. (2001). Briefly, larvae were reared from eggs on N. attenuata plants, and OS from 3 rd -to 5 th -instar larvae were collected with gentle vacuum suction and stored under argon at -80°C. Manduca sexta feeding was simulated by applying 20 µl of eliciting solution of 1 : 5 (v/v) water-diluted oral secretions immediately to fresh leaf wounds (W + OS) generated by three lines of pattern wheel punctures on either side of a leaf and gently dispersed across the leaf surface with a gloved finger. Plants were elicited during the rosette stage, 10-13 d after potting, and between 13:00 h and 14:00 h.

Micrografting
Seven-day-old seedlings were micro-grafted as described by Fragoso et al. (2011). EV (E) and irTOC1 (T) genotypes were used as scions or rootstocks yielding EE, TT, and ET graftswhere the first letter refers to the scion genotype and the second letter to the genotype of the rootstock.

Transcript abundance
Leaf (150 mg) or root tissues (300 mg) were harvested, and RNA was extracted with TRIzol reagent (Invitrogen) according to the manufacturer's instructions. Total RNA was quantified using a NanoDrop (Thermo Scientific, Wilmington, DE, USA) and cDNA was synthesized from 500 ng of total RNA using RevertAid H Minus reverse transcriptase (Fermentas, Vilnius, Lithuania) and oligo (dT) primer (Fermentas). Reverse transcription quantitative PCR (RT-qPCR) was performed in an Mx3005P PCR cycler (Stratagene, San Diego, CA, USA) using SYBR GREEN1 kit (Eurogentec, Li ege, Belgium). The N. attenuata Translation initiation factor 5a-2 (IF5a-2) gene and the N. attenuata elongation factor 1a (EF1a) were used as reference genes for normalization, with IF5a-2 being used in the experiment represented in Supporting Information Fig. S2 (see Results section on 'Blocking ethylene signaling'), and EF1a being used in the experiment represented in Fig. 1. EF1a is utilized as a standard reference gene for time series analyses in N. attenuata, while the data in Fig. S2 were collected as part of an experiment that involved roots and shoots; for these analyses, IF51a, which remains consistent across tissue types in N. attenuata, was utilized as a reference gene. The sequences of primers used for RT-qPCR are provided in Table S1. All qRT-PCR data were normalized using the delta-Ct method. To minimize the influence of phenology, irTOC1 plants were germinated 1 wk before EV plants, as they display a delay in flowering of 1 wk (Yon et al., 2012;Valim et al., 2019). After potting, plants did not receive additional fertilizer for 10 d, after which rosette-stage plants were pulse labelled with 5.1 mg of 15 N isotope administered as K 15 NO 3 3 d before W + OS elicitation. Following 15 N pulse labeling, plants were watered with an unlabeled nutrient solution every day. Incorporation of 15 N into N-containing metabolites was determined as described by Ullmann-Zeunert et al. (2012. Briefly, secondary metabolites were extracted in pre-cooled acidified methanol, and analysis was performed using an electrospray ionization-time of flight mass spectrometer (Bruker Daltonic, Bremen, Germany) as described previously by Gaquerel et al. (2010). Average mass spectra were extracted and 15 N-incorporation analysis was performed using the EXCEL spreadsheet PROSIPQUANT (Taubert et al., 2011), modified for small metabolites based on compound sum formulae.

Research
New Phytologist MCP) and then re-acclimated to glasshouse conditions for 7 h before W + OS elicitation at 13:00 h. Plants (five total) along with the activated solution of 1-MCP were placed within growth chambers fitted with Plexiglas lids and a 14.5 cm fan for air circulation. Following Kahl et al. (2000), 500 mg of Ethylblock (0.43% 1-MCP (van der Sprong, Postbus, the Netherlands)) was dissolved in a vial containing 50 ml of alkaline solution (0.75% KOH + NaOH in a 1 : 1 ratio) to release the active substance, 1-MCP, and immediately placed within the chamber. To control for any potential stress responses associated with confinement in chambers and exposure to the solution, control plants were kept in identical chambers supplied with an equivalent amount of alkaline solution.

Phytohormone and secondary metabolite extraction and quantification
Phytohormone analysis of leaf material was performed by ultra performance liquid chromatography-mass spectrometry (EvoQ Elite Triple quad-MS; Bruker Daltonik GmbH, Bremen, Germany) following extraction in pre-cooled acidified methanol and column purification as described by Sch€ afer et al. (2016). Secondary metabolites were extracted similarly in pre-cooled acidified methanol, and analysis was performed by an electrospray ionization-time of flight mass spectrometer (Bruker Daltonik) as described previously by Li et al. (2017Li et al. ( , 2018.

Statistical analyses
All data were analyzed using R v.3.4.2 (R Core Team, 2018) and RSTUDIO v.1.0.153 (RstudioTeam, 2016). After fitting the data to a linear model, the model residuals were graphically examined for homoscedasticity and normality, and potential outliers were indicated based on Cook's distance as well as a Cook's distance vs leverage analysis (Cook & Weisberg, 1982). Points that were considered highly divergent from the model, based on a cutoff of 0.3 for Cook's distance (Cook's Di of observation i) as well as above Di of 1 for Cook's distance vs leverage, were further investigated to identify any divergent phenotypes, such as unhealthy or dying plants, divergent developmental characteristics, or else inconclusive data processing results, such as poorly separated or missing peaks in the secondary metabolite analysis. Details regarding which outliers were removed, as well as further details of statistical analyses for each dataset, are given in Notes S1.
Pairwise post-hoc comparisons were made using the R package EMMEANS (Lenth et al., 2018) using Tukey HSD tests after significant results were observed in a two-way ANOVA. When a two-way ANOVA revealed significance in only one factor, and not in the interaction, the interaction was excluded from the model and the pairwise comparisons were extracted within the nonsignificant factor by performing pairwise t-tests. All ANOVAs and relevant contrasts are presented in Notes S1. For regression analyses, estimates of the slopes of the covariate trends were generated for each level of the factor using the function emtrends in the R package EMMEANS (Lenth et al., 2018).

Results
TOC1 silencing in N. attenuata leads to a shift in JA signaling, and abundance of defense-related gene transcripts after simulated herbivory ZTL and TOC1 interact directly within the circadian clock (M as et al., 2003;Yon et al., 2012), and silencing ZTL has been shown to decrease nicotine but not phenolamide levels in N. attenuata in a JA-independent manner after elicitation with oral secretions from the generalist herbivore Spodoptera littoralis (Li et al., 2018). We therefore investigated whether these phenotypes were similar under simulated herbivory after elicitation with oral secretions of a nicotine-tolerant specialist herbivore, M. sexta. Transcript abundance of JA biosynthetic genes was not strongly affected by either ZTL or TOC1 silencing (Fig. 1a-d), while JA and JA-Ile concentrations were strongly reduced in ZTL-silenced plants only (Fig. 1e,f). Although the peak of JA-Ile induction was shifted in irTOC1 plants, we did not observe a significant change in irTOC1 or irZTL relative to EV in transcript levels of Threonine Deaminase, which is strongly elicited by herbivory and contributes to the conversion of threonine into isoleucine, a necessary step in the production of bioactive JA-Ile (see Kang et al., 2006). irTOC1 TD transcript abundance was 50% higher than irZTL transcript abundance by 180 min after induction, after JA and JA-Ile concentrations had already decreased (Fig. 1g). The peak of transcript abundance following simulated herbivory was significantly decreased in both irTOC1 and irZTL plants for MYC2 and MYB8, two important coordinators of defense responses following herbivory (Fig. 1h,j), while basal transcript abundance of COI1 was reduced in ZTL-silenced plants only (Fig. 1i).
Silencing TOC1 leads to increases in nicotine accumulation and decreases in accumulation of both constitutive and inducible phenolamides after simulated herbivory Given the dampened elicitation of MYC2 and MYB8 in both irTOC1 and irZTL plants, we examined the concentrations of three of N. attenuata's most abundant secondary metabolites, the alkaloid nicotine and the two phenolamides caffeoylputrescine (CP) and dicaffeoyl spermidine (DCS). Nicotine, the signature compound of Nicotiana spp., functions as both a constitutive and elicited defense, but is not elicited by oral secretions of the specialist herbivore M. sexta Halitschke et al., 2001). While CP and DCS are both phenolamides, CP is strongly elicited by M. sextaherbivory, while DCS is produced constitutively in the plant (Kaur et al., 2010). All three compounds are N-rich and share the precursor molecule putrescine, which is produced as part of the urea cycle (Fig. 2a).
Under control conditions, CP and nicotine levels did not differ significantly among irZTL, irTOC1 and EV plants. At 72 h after induction (hai), irZTL levels of nicotine and CP were similar to the levels reported in Li et al. (2018), with a 50% decrease in nicotine accumulation and a nonsignificant tendency towards lower levels of CP compared to EV plants (Fig. 2b,d). irTOC1 levels of nicotine were 45% higher and CP levels were 55% lower compared to EV plants 72 hai (Fig. 2b,d). CP levels were not significantly different Ó 2020 The Authors New Phytologist Ó 2020 New Phytologist Foundation New Phytologist (2020) 228: 1227-1242 www.newphytologist.com between irZTL and irTOC1, while nicotine levels were 60% higher in irTOC1 relative to irZTL plants. Under control conditions, DCS levels were 72% lower in irTOC1 and 50% lower in irZTL relative to EV plants, while neither irZTL nor irTOC1 levels 72 hai were significantly different from EV levels (Fig. 2c).
Allocation of N is shifted towards nicotine in TOC1-silenced plants Nicotine, CP, and DCS are N-rich compounds which are found in high abundance both constitutively and after herbivore induction and share an N-rich precursor, putrescine (Fig. 2a). We examined the allocation pattern of N among these three compounds using a 15 N-pulse labeling approach described previously in N. attenuata (Ullmann-Zeunert et al., 2012, with both control and M. sexta W + OS-elicited plants being sampled at various time intervals over the next 4 d (Fig. 3a). To minimize the influence of phenology, irTOC1 plants were germinated 1 wk before EV plants, as they display 1-wk-delayed flowering (Yon et al., 2012;Valim et al., 2019). Final measured nicotine concentrations were 40% higher in irTOC1 relative to EV elicited leaves, while nicotine concentrations under control conditions While irZTL shows reductions in all compounds compared to EV, irTOC1 produces relatively more nicotine and less of the two phenolamides, leading to significant changes in the (e, f) trait correlations among these compounds for irTOC1 plants. Data in boxes refer to results of ANOVA tests; differing letters and asterisks represent significant differences in estimated marginal means of linear trends, extracted through Tukey-adjusted pairwise contrasts between genotypes, *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant. Shading represents 95% confidence level intervals along correlation axes for (e) and (f). CoA, Coenzyme A.

Research
New Phytologist were largely unchanged (Fig. 3b). CP and DCS concentrations followed an inverse trend: CP concentrations were similar under control conditions between EV and irTOC1, while final measured CP concentrations were 30% higher in EV elicited plants (Fig. 3c); final measured DCS concentrations were 40% higher in EV control plants, and 25% higher in EV elicited plants (Fig. 3d).
Incorporation of 15 N (measured as % of atomic incorporation of 15 N, hereafter %At 15 N) was temporally shifted in irTOC1 and EV plants by the W + OS treatment for nicotine (Fig. 3e). Final measured %At 15 N for CP was 50% lower in irTOC1 control plants, and 30% lower in irTOC1 elicited plants relative to EV (Fig. 3f). Final measured %At 15 N for DCS was 30% lower in irTOC1 control plants and 15% lower in irTOC1 elicited plants relative to EV (Fig. 3g). Given the large differences in nicotine and phenolamide pools between EV and irTOC1 plants, mirroring differences in their concentrations (Fig. S1), we calculated the investment of 15 N into these metabolites by comparing the fractions of the nicotine and phenolamide pools synthesized from 15 N. These results showed that a larger fraction of 15 N-labeled nicotine accumulated in irTOC1 elicited plants (Fig. 3h), while larger fractions of 15 Nlabeled CP accumulated in EV elicited plants (Fig. 3i). The fraction of 15 N-labeled DCS was larger in EV relative to irTOC1 elicited and control plants, although elicitation increased fractions in both genotypes (Fig. 3j). In summary, the patterns of 15 N incorporation as a fraction of metabolite concentration follow the same pattern as metabolite concentrations; however, irTOC1 plants preferentially invest 15 N into nicotine over phenolamides to a greater extent than predicted solely by the differences in concentration of these metabolites.
To explore allocation differences between nicotine and phenolamides in irTOC1 plants, we compared the normalized differences between nicotine and CP or DCS (e.g. nicotine À CP/ nicotine + CP), for concentration and 15 N incorporation rates. This produced a normalized score where a value of +1 corresponds to only nicotine being detected; a value of À1 corresponds to only phenolamide being detected; and a value of 0 corresponds to equal abundances or incorporation rates being measured (Fig. 4b). Given the larger contribution of nicotine than phenolamide to the plant's biomass, normalized concentration differences for both EV and irTOC1 are heavily biased towards nicotine, with values between +1.00 and +0.85 for all time points, genotypes, and treatments for nicotine vs CP (Fig. 4c), and between +1.00 and +0.50 for nicotine vs DCS (Fig. 4d). Against this nicotine-biased background, elicitation shifted the proportion of nicotine and phenolamide pools towards CP or DCS, and both control and elicited EV plants were significantly shifted towards CP or DCS relative to irTOC1. The bias toward nicotine was less marked in normalized 15 N incorporation, with plants shifting towards greater incorporation into CP for elicited EV plants, and approaching 0 for elicited irTOC1 and control EV plants; however, control irTOC1 plants remained strongly nicotine-biased, with values above +0.5 until 45 hai (Fig. 4e). Patterns were similar for nicotine vs DCS 15 N incorporation, with control and elicited EV as well as elicited irTOC1 plants approaching 0, while elicited irTOC1 plants remained above +0.4 (Fig. 4f).
Blocking ethylene signaling by 1-methylcyclopropene treatment abolishes differences in nicotine accumulation between EV and TOC1-silenced plants The JA-induced increase in nicotine accumulation in N. attenuata is suppressed by oral secretions of M. sexta larvae (Winz & Baldwin, 2001;Von Dahl et al., 2007), and the observed increase in nicotine accumulation in irTOC1 plants after simulated herbivory mirrors the release from suppression in ethylene-deficient plants. However, the herbivory-elicited ethylene burst is not affected in TOC1-silenced plants (Fig. S2), and the transcript abundances of various components of the ethylene signaling pathway are not significantly affected by TOC1 silencing (Fig. S3). To determine whether increased levels of nicotine in irTOC1 plants were related to ethylene perception, rosette stage plants (10 d after potting) were treated with 1methylcyclopropene (1-MCP) inside of a Plexiglas chamber for 8 h overnight and re-acclimated to glasshouse conditions for 7 h before elicitation at 13:00 h (Fig. 5a). As for the 15 N-labeling experiment, irTOC1 plants were germinated 1 wk early. 1-MCP is an ethylene agonist that binds to the ethylene receptor ETR1 and prevents inactivation of downstream ethylene-sensitive transcription factors ( Fig. 5b; see Wang et al., 2002;Blankenship & Dole, 2003;Watkins, 2006;Ju & Chang, 2015). Differences in nicotine accumulation after W + OS elicitation between EV and irTOC1 plants were abolished by 1-MCP treatment (Fig. 5c), while accumulation of CP increased proportionally after 1-MCP treatment, remaining lower in irTOC1 plants (Fig. 5e). DCS levels were slightly increased in EV plants after 1-MCP treatment and elicitation, and remained significantly higher than irTOC1 plants (Fig. 5g). We analyzed root metabolite levels and also observed a decrease in the variation of nicotine levels between roots of EV and irTOC1 plants treated with 1-MCP (Fig. 5d). Consistent with previous findings, CP and DCS were not consistently detected in roots, and the low levels which were detected did not change significantly as a result of treatment (Fig. 5f,h).
To further test TOC1's role in regulating ethylene sensitivity, we performed an ethylene triple response assay, a classic tool for exploring ethylene responses (Merchante & Stepanova, 2017). We compared the responses of irTOC1, EV, and 35S-etr1 seedlings germinating in a gradient of 1-aminocyclopropane-1carboxylic acid (ACC), an ethylene precursor, under dark conditions. 35S-etr1 seedlings express the Arabidopsis etr1-1 mutant receptor, rendering plants unresponsive to ethylene (von Dahl et al., 2007). In EV or wild-type seedlings, exposure to even minute concentrations of ACC (0.2 lM) is sufficient to observe the expected response, consisting of a shortening and thickening of the hypocotyl as well as an exaggeration of apical hook curvature in Arabidopsis. In N. attenuata, although the effect of ethylene on the thickness of the hypocotyl was less obvious, hypocotyl shortening provided a reliable metric for quantifying ethylene sensitivity (Fig. S4a) of EV, which is highly sensitive and displays hypocotyl shortening at concentrations of 0.2 lM. Two independent irTOC1 lines were not responsive to 0.2 lM ACC, and displayed more elongated hypocotyls under 0.5 lM ACC concentrations compared to EV; as expected, 35S-etr1 was unresponsive to any tested concentration of ACC (Fig. S4b).
TOC1 in roots limits nicotine accumulation, but changes to reproductive allocation depend on TOC1 in shoots Given that nicotine is synthesized in the roots, and that changes in the allocation of N among metabolically expensive compounds New Phytologist (2020) 228: 1227-1242 Ó 2020 The Authors New Phytologist Ó 2020 New Phytologist Foundation www.newphytologist.com

Research
New Phytologist might affect the plant's ability to defend itself against herbivores, we examined the effect of tissue-specific silencing of TOC1 by using a micrografting approach under field conditions to generate a heterograft silenced in TOC1 in roots only (EV/irTOC1, hereafter ET heterografts) or in whole plants (irTOC1/irTOC1, hereafter TT homografts) as well as control grafts with EV roots and shoots (EV/EV, hereafter EE homografts; Fig. 6a). The increased nicotine accumulation of the heterograft (EE -ET: Concentrations and 15 N incorporation rates were normalized between nicotine and the two phenolamides CP and DCS, using the equation (nicotine À phenolamide)/(nicotine + phenolamide). These produced normalized differences that can be interpreted as exemplified in the following figure, such that a value of > 0 denotes greater values for nicotine, a value = 0 denotes equal values for nicotine and the phenolamide, and a value of < 0 denotes greater values for the phenolamide. (c, d) CP and DCS normalized concentration differences showed marked biases towards nicotine, although significant differences between EV and irTOC1 plants were observed (n = 5 per genotype, treatment and time point; means AE SEM are shown). (e, f) Normalized 15 N incorporation differences showed more equivalence between nicotine and CP/DCS, with significant differences between EV and irTOC1 plants also observed (n = 5 per genotype, treatment and time point; means AE SEM are shown). Asterisks represent significant differences in estimated marginal means of linear trends, extracted through Tukey-adjusted pairwise contrasts, within one timepoint: *, P < 0.05; **, P < 0.01; ***, P < 0.001.
Ó 2020 The Authors New Phytologist Ó 2020 New Phytologist Foundation New Phytologist (2020) 228: 1227-1242 www.newphytologist.com P < 0.04) mirrored the levels of nicotine in TT homografts ( Fig. 6b; EE -TT: P < 0.04, pairwise t-tests across treatments, Holm-Bonferroni P-value adjustment). The reduction in W + OS-elicited CP levels observed in TT homografts was abolished in ET heterografts, and levels between treatments were significantly different across genotypes ( Fig. 5c; Control -Elicited: P = 0.029, pairwise t-tests across treatments, Holm-Bonferroni Pvalue adjustment). Curiously, DCS levels showed an increase in ET heterografts, but not in TT homografts relative to EE homografts ( Fig. 6d; EE -ET: P = 0.028, EE -TT: P = 0.087, pairwise t-tests across treatments, Holm-Bonferroni P-value adjustment). A similar pattern was observed for the accumulation of the N-free polyphenol chlorogenic acid, with levels being attenuated in irTOC1 plants and TT homografts (Fig. S5b). ET plants did not exhibit fitness reductions associated with increased nicotine accumulation, rather having a 33% increase in seed capsule production 40 d after planting (dap) compared to EE control plants (Fig. 6e, EE -ET: P = 0.0005, Tukey contrasts by genotype). Unlike TT plants, both ET and EE elicited plants experienced a 50-60% decrease in seed capsule production by 40 dap (Fig. 6e, EE Control -EE W+OS : P < 0.0001; EE Control -EE W+OS : P = 0.0051; Tukey contrasts by treatment). Flower production was similarly affected by elicitation in EE and ET plants (Fig. 6f). By contrast, TT aboveground and belowground biomass were decreased 45-50% by elicitation, with TT control roots being more massive than EE control shoots (Fig. 6g, EE Control -TTControl : P = 0.0114, ET Control -TT Control : P = 0.0708) and with TT control roots more massive than EE and ET control roots (Fig. 6h, EE Control -TT Control : P = 0.0067, ET Control -TT Control : P = 0.0358; Tukey contrasts by treatment). Elicited plants had similar above-and belowground biomass.

Discussion
Selective allocation by plants, both of resources to defense vs growth and reproduction, and of defense traits among tissues, is a common theme in theories of plant defense (McKey, 1974;Herms & Mattson, 1992;Meldau et al., 2012). Although plant defense theories generally do not consider the temporal dimension of selective allocation, two frameworks, apparency theory and optimal defense theory, incorporate different probabilities of herbivore attack based on plant phenotypes. In particular, optimal defense theory predicts that tissue defense should be correlated with the likelihood of attack and tissue value (Zangerl & Rutledge, 1996;. This could also be considered in terms of diurnal changes in likelihood of attack, in addition to the traditional view of changing likelihood over ontogeny. Previous work in N. attenuata is consistent with the Nicotine accumulation in (c) leaves and (d) roots of EV (black) and irTOC1 (blue) plants is presented for plants treated with and without 1-MCP, in addition to Manduca sexta wounding and oral secretions (W + OS and 1-MCP + W + OS, respectively). Accumulations of the inducible phenolamide caffeoylputrescine in (e) leaves and (f) roots and the constitutive phenolamide dicaffeoyl spermidine in the (g) leaves and (h) roots are presented on the same scale; insets provide visualization of significant differences at low abundances (e, f, h; n = 4-5 per genotype and treatment, mean + SE for all panels). Data in boxes refer to results of ANOVA tests; asterisks represent the significance of irTOC1 values to those of EV extracted using Tukey-adjusted pairwise contrasts, within one timepoint: *, P < 0.05; **, P < 0.01; ***, P < 0.00; ns, not significant.

Research
New Phytologist predictions of optimal defense theory: inducible defenses in N. attenuata are regulated by levels of cytokinins, a class of developmental growth hormones (Br€ utting et al., 2017), and developmental transitions such as flowering alter the ethylene/JAmediated defense signaling at the heart of inducible plant defenses (Diezel et al., 2011). However, the role of the circadian clock in optimizing defenses across development has not been previously investigated.
Our results provide new mechanistic insight into the circadian clock's role in optimizing the allocation of resources in N. attenuata's defense chemistry, building on the evidence outlined in Li et al. (2018) that the clock component ZTL interacts with jasmonate signaling to promote synthesis of the abundant, N-rich defense compound nicotine. Here, we find that TOC1 interacts with ethylene signaling to dampen nicotine accumulation, and that TOC1 additionally mediates the allocation of N between nicotine and another group of abundant N-containing defense compounds, the phenolamides. The increase in elicited nicotine levels in particular fit the current model of TOC1-ZTL interaction, given that TOC1 indirectly inhibits ZTL by repressing GI ( Fig. 1a; Makino et al. 2001), and ZTL in turn targets TOC1 for degradation (Fujiwara et al. 2008).
Testing the functional consequences of these allocation differences under field conditions demonstrates that these trade-offs in N allocation do not explain the large effect of TOC1 function on plant fitness, which furthermore seems to be specific to the function of TOC1 in shoots, while nicotine is synthesized in the root and then transported to the shoot. Furthermore, correcting for . Data in boxes refer to results of ANOVA tests; asterisks represent significant differences (P < 0.05) in estimated marginal means of linear trends, extracted through Tukey-adjusted pairwise contrasts: *, P < 0.05; **, P < 0.01; ***, P < 0.00; ns, not significant.
Ó 2020 The Authors New Phytologist Ó 2020 New Phytologist Foundation New Phytologist (2020) 228: 1227-1242 www.newphytologist.com phenological differences between irTOC1 and EV plants did not affect these patterns of defense compound accumulation, indicating a potentially tissue-specific function of TOC1 in defense metabolite production. Given that the circadian clock has been shown to be mechanistically distinct (Endo et al., 2014) and to vary in functional outputs across tissue types (Bordage et al., 2016), there should be a greater effort to explore the functional consequences of individual circadian clock components across tissues for defense responses. Another theory of plant defense that has been applied in understanding resource allocation, the carbon/nutrient balance theory, predicts that resources such as carbon or macronutrients such as N, phosphorus, or potassium may be rate-limiting for growth or defense compound production (Bryant et al., 1983;Massad et al., 2012). Nitrogen pools in particular provide a valuable testing ground for questions of resource allocation. They can be highly limited and are subject to changing needs and intense intraspecific competition in N. attenuata populations, and many of N. attenuata's highly abundant defense metabolites, including nicotine and phenolamides, are N-rich van Dam & Baldwin, 2001;Ullmann-Zeunert et al., 2013). Resource allocation can be manipulated directly by the circadian clock, as has been shown for carbon metabolism (Sanchez-Villarreal et al., 2013;M€ uller et al., 2014;K€ olling et al., 2015). Our results suggest that TOC1 modulates N allocation indirectly as an output of defense signaling (Figs 3, 4), given that levels of chlorogenic acid (CGA) are also attenuated in irTOC1 plants, and CGA is a polyphenolic compound also regulated by MYB8 much like CP and DCS (Fig. S5). The different effects of ZTL and TOC1 on the accumulation of N-rich defense metabolites (Fig. 2) as well as TOC1's apparent role in ethylene-mediated defense signaling (Fig. 5) indicate that individual members of the circadian clock have specific roles in regulating plant responses to herbivory.
Previous data suggest that herbivory disrupts normal clock function and its outputs . It may be that, whatever the adaptive value of circadian-regulated inducibility of JA or defense metabolite production, plant resources are preferentially if temporarily mobilized for defense after elicitation occurs. Our results are consistent with this expectation. We find that that simulated herbivory elicitation shifts %At 15 N towards phenolamides and away from nicotine in both EV and irTOC1 plants, although differences remain significant between the two genotypes: only EV plants incorporate greater amounts of 15 N into CP rather than nicotine (Fig. 4e). This highlights an important difference between investigating variation in accumulation vs allocation. The effect of TOC1 is harder to observe when comparing concentrations and total secondary metabolite accumulation after elicitation (Figs 2, 3b-d, S1b-d).
Although circadian clock outputs are not always rhythmic, rhythmic phenotypes may point towards novel functions for clock components. Ethylene emissions are rhythmic and ethylene signaling is regulated by the circadian clock in Arabidopsis (Thain et al., 2004;Song et al., 2018). Previous work has shown that ethylene-insensitive mutants overproduce ethylene but do not display altered rhythmicity in ethylene production (Thain et al., 2004). Despite this, our results provide evidence that for a well-known ethylene responsethe dampening of herbivoreelicited nicotine accumulationthe circadian clock component TOC1 seems to play a key role in signal transduction. The induction of ethylene after elicitation with M. sexta oral secretions is not diminished by TOC1 silencing, but rather tends to increase, consistent with the hypothesis that TOC1 abrogation may lead to impaired ethylene sensitivity (Figs 5, S2). This is supported by a triple response assay demonstrating reduced ethylene sensitivity of irTOC1 plants (Fig. S4).
The ecological significance of ethylene-mediated nicotine dampening under field conditions remains unclear, given that nicotine levels remained high even without experimental elicitation in the field (Fig. 6b). This could be due to the presence of generalist herbivores, which may elicit plants under natural conditions independently of experimental manipulations (even when herbivore exclusion is attempted e.g. with insecticide; see McGale et al., 2018). Ethylene signaling prevents a surge of nicotine in plants elicited by the nicotine-tolerant M. sexta (von Dahl et al., 2007), but is likely not relevant for plants already using nicotine to defend against nicotine-susceptible generalists. It is thus possible that further field studies would yield different degrees of nicotine production in response to experimental elicitation, with the elicited peak in irTOC1 plants more obvious under conditions with fewer generalist herbivores. We note that the more M. sexta-specific induction of CP was observed (Fig. 6c). Thus, our field experiments demonstrate that even under less controlled and more realistic conditions, the shifted ratio between nicotine and phenolamides in TOC1-silenced plants persists (Fig. 6). Micrografting of TOC1-silenced roots to EV shoots demonstrates that TOC1's role in regulating herbivore-elicited phenolamide production is likely shoot-specific, while TOC1's role in control of nicotine biosynthesis is root-specific.
TOC1 is known to be involved in flowering time regulation, and this is observed under field conditions for whole-plant TOC1 silencing (Valim et al., 2019). In N. attenuata, the onset of flowering attenuates the elicitation of defense compounds (Diezel et al., 2011), and whole-plant TOC1-silencing may thus change the responsiveness of plants to elicitation over their lifetime. However, developmental timing did not explain the shift in defenses between nicotine and phenolamides observed in irTOC1 plants, as both plants that were developmentally matched by offsetting germination (Figs 3-5), as well as plants that were germinated synchronously (Figs 1, 2, 6), exhibited the same shifts in nicotine and phenolamide production. Furthermore, using a micrografting approach under field conditions, root TOC1-silenced heterografts with wild-type phenology produced greater amounts of nicotine (Fig. 6b), despite no apparent fitness consequences (Fig. 6e,f). As discussed in the previous paragraph, nicotine did not respond to induction in the field, unlike CP, indicating that all grafted genotypes remained sensitive to induction (and that nicotine was already responding to ambient herbivory). Together, our data suggest that phenological functions of TOC1 affect plant fitness, but do not explain differential production of N-rich defense compounds.
Given that the fitness metrics of root-only TOC1-silenced heterografts remained similar to EV plants under field New Phytologist (2020) 228: 1227-1242 Ó 2020 The Authors New Phytologist Ó 2020 New Phytologist Foundation www.newphytologist.com

Research
New Phytologist conditions, the fitness consequences of whole-plant TOC1 silencing may have less to do with a differential allocation of N generally, and more to do with a specific shift towards overproduction of a single defense metabolite such as nicotine at the expense of a more diverse defensive metabolite profile. It should be noted that our field data cannot fully disentangle phenological from metabolic effects on plant fitness, as ET grafts differ from TT grafts in both phenology and phenolamide production. Future work should dissect to what extent the fitness disadvantages of irTOC1 and other clock-altered plants under field conditions are due to allocation costs, or to the opportunity costs of mistimed defenses. It should be noted that TOC1 has drastic effects on plant drought responses, both in seedlings under laboratory conditions (Legnaioli et al., 2009) as well as under field conditions (Valim et al., 2019). However, there is little evidence under field conditions for an interactive effect between drought and herbivory; a recent study using 16 soybean genotypes failed to find any interactive effects of drought and herbivory on plant performance (Grinnan et al., 2013). The results of the 15 N-labeling experiment (Figs 3, 4), together with previous work by Joo et al. (2018), suggest that the circadian machinery plays a greater role in the metabolic state of the plant preceding and immediately at the beginning of herbivory events, and that herbivory-elicited metabolic reorganization decouples the clock from its normal function under control conditions. The use of inducible promoters (see e.g. Sch€ afer et al., 2013) may help to dissect the role of the clock in basal or preparatory stages before herbivory, as well as the role of circadian clock components such as TOC1 and ZTL in the mobilization of defense responses during herbivory.
One test of the herbivory disruption hypothesis would be to measure whether herbivores feeding at different times of the day perform differently on clock-shifted plants, with or without prior elicitation. Performance differences are observed for rhythmically feeding Trichoplusia ni on clock-disrupted vs wild-type Arabidopsis plants without pre-elicitation (Goodspeed et al., 2012). The disruption hypothesis predicts that pre-elicitation would reduce this difference. Alternatively, clock components may function in the mobilization of defense apart from their timekeeping function, in which case knocking out or down these components (including TOC1 and ZTL) may affect plant susceptibility and herbivore performance regardless of either elicitation, or whether herbivores display rhythmic feeding. Performing such experiments under field conditions, while more realistic, makes them difficult to interpret given the many effects of the clock on phenology. Performing experiments at different times of the season should further elucidate the effect of phenology on the clock's role in plant defense, as it has done for yield (Izawa et al., 2011). This could elucidate interactions of phenology with environmental cues in defense, similarly to the coincidence model of photoperiodic flowering, which requires both functioning circadian rhythms and appropriate seasonal stimuli.

Supporting Information
Additional Supporting Information may be found online in the Supporting Information section at the end of the article.     Notes S1 Expanded statistical information and outlier removal.

Table S1
Primer sequences used in this study.
Please note: Wiley Blackwell are not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.
New Phytologist is an electronic (online-only) journal owned by the New Phytologist Foundation, a not-for-profit organization dedicated to the promotion of plant science, facilitating projects from symposia to free access for our Tansley reviews and Tansley insights.
Regular papers, Letters, Research reviews, Rapid reports and both Modelling/Theory and Methods papers are encouraged. We are committed to rapid processing, from online submission through to publication 'as ready' via Early View -our average time to decision is <26 days. There are no page or colour charges and a PDF version will be provided for each article.
The journal is available online at Wiley Online Library. Visit www.newphytologist.com to search the articles and register for