Silencing PHOSPHOENOLPYRUVATE CARBOXYLASE1 in the Obligate Crassulacean Acid Metabolism Species Kalanchoë laxiflora causes Reversion to C3-like Metabolism and Amplifies Rhythmicity in a Subset of Core Circadian Clock Genes

Unlike C3 plants, Crassulacean acid metabolism (CAM) plants fix CO2 in the dark using phosphoenolpyruvate carboxylase (PPC; EC 4.1.1.31). PPC combines PEP with CO2 (as HCO3−), forming oxaloacetate that is rapidly converted to malate, leading to vacuolar malic acid accumulation that peaks phased to dawn. In the light period, malate decarboxylation concentrates CO2 around RuBisCO for secondary fixation. CAM mutants lacking PPC have not been described. Here, RNAi was employed to silence CAM isogene PPC1 in Kalanchoë laxiflora. Line rPPC1-B lacked PPC1 transcripts, PPC activity, dark period CO2 fixation, and nocturnal malate accumulation. Light period stomatal closure was also perturbed, and the plants displayed reduced but detectable dark period stomatal conductance, and arrhythmia of the CAM CO2 fixation circadian rhythm under constant light and temperature (LL) free-running conditions. By contrast, the rhythm of delayed fluorescence was enhanced in plants lacking PPC1. Furthermore, a subset of gene transcripts within the central circadian oscillator were up-regulated and oscillated robustly. The regulation guard cell genes involved controlling stomatal movements was also altered in rPPC1-B. This provided direct evidence that altered regulatory patterns of key guard cell signaling genes are linked with the characteristic inverse pattern of stomatal opening and closing during CAM.


INTRODUCTION
Crassulacean acid metabolism (CAM) is a pathway of photosynthetic CO2 fixation found in species adapted to low rainfall and/ or periodic drought, such as the Madagascan-endemic succulent, Kalanchoë laxiflora Baker (Family: Crassulaceae; Order: Saxifragales) (Hartwell et al., 2016). CAM species open their stomata for primary atmospheric CO2 fixation in the dark, when their environment is cooler and more humid, and close stomata in the light, when the atmosphere is at its hottest and driest (Osmond, 1978). The increased water use efficiency (WUE) and CO2 fixation efficiency of CAM species has led to the proposal that productive CAM crop species, including certain Agave and Opuntia species, represent a viable approach to generate biomass for biofuels and renewable platform chemicals for industry through their cultivation on under-utilised, seasonally-dry lands that are not well-suited to food crop production (Borland et al., 2009;Cushman et al., 2015). Furthermore, efforts are underway to engineer CAM into key C3 crops (Borland et al., 2014;DePaoli et al., 2014;Borland et al., 2015;Lim et al., 2019).
During CAM, primary nocturnal CO2 assimilation is catalyzed by phosphoenolpyruvate carboxylase (PPC), generating oxaloacetate that is rapidly converted to malate and stored in the vacuole as malic acid (Borland et al., 2009). At dawn, malic acid is transported out of the vacuole and malate is decarboxylated in the cytosol, with the released CO2 re-fixed by RuBisCO behind closed stomata (Borland et al., 2009). Strict temporal control prevents a futile cycle between the enzymes and metabolite transporters driving malate production in the dark, and those driving malate decarboxylation in the light (Hartwell, 2006). PPC regulation is central to this temporal control. PPC is activated allosterically by glucose 6-phosphate (G6P) and inhibited by malate, aspartate and glutamate (O'Leary et al., 2011). The circadian clock optimizes the timing of the CAM carboxylation and decarboxylation pathways to prevent futile cycling (Wilkins, 1992;Hartwell, 2005).
Temporal optimization of PPC involves protein phosphorylation in the dark period, catalysed by a circadian clock-controlled protein kinase, namely phosphoenolpyruvate carboxylase kinase (PPCK) (Carter et al., 1991;Hartwell et al., 1996;Hartwell et al., 1999;Taybi et al., 2000;Boxall et al., 2005). Phosphorylated PPC is less sensitive to feedback inhibition by malate, which in turn ensures sustained CO2 fixation as malic acid accumulates throughout the dark period (Nimmo et al., 1984;Carter et al., 1991;Boxall et al., 2017). In the light, PPC becomes more sensitive to inhibition by malate due to dephosphorylation by a protein phosphatase type 2A (PP2A), which is not known to be subject to circadian control (Carter et al., 1990).
In CAM species such as Kalanchoë, the PEP substrate required for nocturnal atmospheric CO2 fixation by PPC is generated through starch breakdown and glycolysis . In the model C3 model species Arabidopsis thaliana, nocturnal degradation of leaf starch begins with the phosphorylation of glucan chains by GLUCAN WATER DIKINASE (GWD) and PHOSPHOGLUCAN WATER DIKINASE (PWD) (Ritte et al., 2006). The phosphorylated glucan chains are then further degraded by ALPHA-AMYLASES (AMYs) and BETA-AMYLASES (BAMs) to maltose and glucose. Nocturnal starch hydrolysis by BAMs is the predominant pathway in C3 leaves, with chloroplastic BAM3 being the major BAM isozyme driving nocturnal starch degradation in photosynthetic leaf mesophyll cells of Arabidopsis (Fulton et al., 2008). BAM1, BAM2 and BAM5-9 are not required for nocturnal starch degradation in Arabidopsis leaf mesophyll cells (Santelia and Lunn, 2017). Maltose and glucose are exported from chloroplasts by MALTOSE EXCESS 1 (MEX1) and PLASTIDIC GLUCOSE TRANSPORTER (pGlcT), respectively, with MEX1 being the predominant C3 route for carbon export (Smith et al., 2005). In the facultative CAM species M. crystallinum, chloroplasts possess transporters for triose phosphate (TPT), G6P (GPT), glucose (pGlcT) and maltose (MEX1) (Neuhaus and Schulte, 1996;Kore-eda et al., 2005;Kore-eda et al., 2013). Chloroplasts from C3 M. crystallinum exported maltose during starch degradation, whereas chloroplasts isolated from CAM-induced leaves exported predominantly G6P, supporting the proposal that starch was broken down via plastidic starch phosphorylase (PHS1) during nocturnal CO2 fixation (Neuhaus and Schulte, 1996).
A further defining characteristic of CAM relates to the characteristic nocturnal stomatal opening and CO2 uptake, and light period stomatal closure during malate decarboxylation and peak internal CO2 supply (Males and Griffiths, 2017). CAM stomatal control is the inverse of the stomatal regulation observed in C3 species (Borland et al., 2014). The opening and closing of stomata is driven by the turgor of the guard cell (GC) pair that surround the stomatal pore. High turgor drives opening and reduction in turgor leads to closure. The increase in turgor during opening is driven by the accumulation of K + , Cland malate 2ions plus sugars in the GCs (Jezek and Blatt, 2017). Closure of stomata is driven by a reversal of GC ion channels and metabolism, with K + and Clbeing transported out, and metabolites being turned over within the GCs. Stomatal aperture responds to changing light, CO2, ABA, solutes and water availability (Zhang et al., 2018).
In addition to its role in CAM and C4, PPC performs an anapleurotic function by replenishing tricarboxylic acid cycle intermediates utilized for amino acid biosynthesis (Chollet et al., 1996). PPC also functions in the formation of malate as a counter anion for light period opening in C3 GCs, and supports nitrogen fixation into amino acids in legume root nodules (Chollet et al., 1996). The major leaf PPCs in Arabidopsis, which are encoded by PPC1 (AT1G53310) and PPC2 (AT2G42600), were found to be crucial for leaf carbon and nitrogen metabolism (Shi et al., 2015). The double ppc1/ppc2 null mutant accumulated starch and sucrose, and had reduced malate, citrate, and ammonium assimilation, and these metabolic changes led to a severe, growth-arrested phenotype (Shi et al., 2015).
Although PPC catalyses primary CO2 fixation in CAM and C4 plants, the only reported PPC mutants are for the C4 dicot Amaranthus edulis and the C4 monocot Setaria viridis (Dever et al., 1995;Alonso-Cantabrana et al., 2018). Loss of the C4 PPC isogene in A. edulis caused a severe and lethal growth phenotype in normal air, with the homozygous mutant plants only managing to reach flowering and set seed when grown at highly elevated CO2 (Dever et al., 1995). In S. viridis, the C4 PPC was reduced to very low levels using RNAi in transgenic lines (Alonso-Cantabrana et al., 2018). These lines grew very slowly even at 2 % CO2 (normal air is 0.04 %), and developed increased numbers of plasmodesmatal pit fields at the mesophyll-bundle sheath interface (Alonso-Cantabrana et al., 2018). By contrast, no CAM mutants lacking PPC have been described, but transgenic lines of Kalanchoë lacking the lightperiod, decarboxylation pathway enzymes mitochondrial NAD-malic enzyme (NAD-ME) and pyruvate orthophosphate dikinase (PPDK) displayed a near complete loss of dark CO2 fixation and failed to turnover significant malate in the light period (Dever et al., 2015). In addition, a CAM mutant of the inducible CAM species Mesembryanthemum crystallinum has been described lacking the starch synthesis enzyme plastidic phosphoglucomutase (Cushman et al., 2008).
In addition, the CAM-associated PPCK1 gene in Kalanchoë was also silenced using RNAi, and this led not only to a reduction in dark period CO2 fixation, but also perturbed the operation of the central circadian clock . However, even the strongest PPCK1 RNAi line was still able to acheive ~33 % of the dark period CO2 fixation observed in the wild type . Those findings led us here to develop transgenic lines of K. laxiflora in which the CAM-associated isogene of PPC itself (isogene PPC1) was down-regulated using RNAi. The most strongly silenced line, rPPC1-B, lacked PPC1 transcripts and activity, and this resulted in the complete loss of dark CO2 fixation associated with CAM, and arrhythmia of the CAM CO2 fixation rhythm under constant light and temperature (LL) free-running conditions. Growth of rPPC1-B plants was reduced relative to wild type in both well-watered and droughtstressed conditions. The plants reverted to fixing CO2 in the light, especially in their youngest leaf pairs. Although the circadian rhythm of CO2 fixation dampened rapidly towards arrhythmia in the rPPC1-B line, the distinct circadian clock output of delayed fluorescence, and the oscillations of the transcript abundances of a subset of core circadian clock genes, were enhanced. Furthermore, the temporal phasing of a wide range of GC specific signaling genes involved in opening and closing was perturbed relative to the wild type in rPPC1-B.
8 PPCK3 being up-regulated by 5-to 8-fold in rPPC1-B ( Figure 1F and 1G), any protein produced from these transcripts did not phosphorylate any PPC2 protein that may have resulted from the induced PPC2 transcripts ( Figure 2B); at least not within the limits of detection with this immunoblotting technique. Rapidly desalted leaf extracts were used to measure the apparent Ki of PPC for L-malate. The Ki was higher in the dark than the light for wild type and rPPC1-A, but no change in the Ki was detected for rPPC1-B (Supplemental Figure 1). Furthermore, PPC activity was not detected in rapidly desalted extracts from rPPC1-B leaves, whereas rPPC1-A displayed reduced but detectable PPC activity, at 43 % of the wild type level ( Figure 2C).
Pyruvate orthophosphate dikinase (PPDK), which functions in concert with NAD-ME during the light-period as part of the CAM malate decarboxylation pathway (Dever et al., 2015), was measured on immunoblots using specific antibodies against PPDK and phospho-PPDK (Chastain et al., 2000;Chastain et al., 2002). The blots showed a similar amount of PPDK protein over the diel cycle in wild type, rPPC1-A and rPPC1-B ( Figure 2D). In Kalanchoë, PPDK is inactivated in the dark by phosphorylation by PPDK-regulatory protein (PPDK-RP), which also activates PPDK in the light through dephosphorylation (Dever et al., 2015). In the wild type, immunoblotting of phospho-PPDK revealed that PPDK was dephosphorylated, and therefore likely to be fully active, between 02:00 and 06:00 in the light when it is required for conversion of pyruvate, from malate decarboxylation, to PEP, thereby facilitating recycling of carbon through gluconeogenesis to starch ( Figure 2E). Line rPPC1-A showed the same pattern of PPDK phosphorylation/ de-phosphorylation as the wild type ( Figure 2E). However, PPDK was phosphorylated throughout the 24 h cycle in rPPC1-B ( Figure 2E), and, therefore, likely to be inactive. Consistent with this prediction, loss of the light period dephosphorylation of PPDK in rPPC1-B correlated with a significant decrease in PPDK activity in the light, whereas the wild type and line rPPC1-A showed strongly light-induced levels of PPDK activity that correlated with the detected level of PPDK dephosphorylation in the light ( Figure 2F).

Levels of Malate, Starch and Soluble Sugars
During CAM in Kalanchoë, primary nocturnal fixation of atmospheric CO2 (as HCO3 -) results in vacuolar malic acid accumulation throughout the dark period. Starch accumulates in the light period and is broken down during the dark to provide PEP as the substrate for carboxylation by PPC. Starch is also broken down in a rapid burst at dawn to form soluble sugars (Wild et al., 2010;Boxall et al., 2017). As the lack of CAMassociated PPC1 was predicted to prevent primary nocturnal carboxylation, metabolites including starch, malate and soluble sugars were measured every 4 h over the 24 h cycle (Figure 3).
Wild type plants accumulated 130 µmols gFW -1 malate by dawn, whereas rPPC1-A and rPPC1-B accumulated 75 µmol gFW -1 and 19.5 µmols gFW -1 , respectively ( Figure 3A). The D-malate values for wild type, rPPC1-A and rPPC1-B were 124.0, 64.3 and 16.2 µmol gFW -1 ( Figure 3A). During the diel cycle, rPPC1-A and rPPC1-B synthesized 100 % and 41 % of the amount of starch accumulated by the wild type ( Figure 3B). The D-starch values for wild type, rPPC1-A and rPPC1-B were 8.5, 8.5 and 3.5 mg starch gFW -1 ( Figure 3B). Lines rPPC1-A and rPPC1-B accumulated 51 % and 15 % of the amount of sucrose accumulated by the wild type ( Figure 3C), and 83 % and 69 % of the level of glucose ( Figure 3D). Glucose accumulated 4 h after the sucrose peak in the wild type, whereas glucose levels peaked at the same time as sucrose in lines rPPC1-A and rPPC1-B ( Figure 3C and CD). Finally, lines rPPC1-A and rPPC1-B accumulated, respectively, 113 % and 61 % of the amount of fructose compared with the wild type ( Figure 3E). In rPPC1-B, the daily maxima for fructose and glucose occurred coincident with that of sucrose at 2 h after dawn ( Figure 3C to 3E).

Growth Analysis in Well-Watered versus Drought-Stressed Conditions
CAM is widely regarded as an adaptation to drought, and so it was important to compare the growth performance of wild type (full CAM) with that of rPPC1-A (small reduction in CAM) and rPPC1-B (no CAM) under both well-watered (WW) and drought-stressed (D-S) conditions ( Figure 3F and Supplemental Figure S2). rPPC1-B were significantly smaller than wild type in both WW and D-S conditions ( Figure 3F).
In WW conditions, the shoot dry weight of rPPC1-B was 21 % less than the wild type ( Figure 3F and S2), but rPPC1-A was not significantly different to the wild type. Under D-S, the shoot dry weight of line rPPC1-B was 12 % lower than that of the wild type ( Figure 3F). The visible size of representative 4-month-old plants demonstrated that rPPC1-B was smaller than wild type and rPPC1-A ( Figure 3G), consistent with the shoot dry weight data ( Figure 3F). Shoot fresh weight was also lower in WW rPPC1-B (Supplemental Figure S2). Furthermore, rPPC1-B displayed a reduced degree of leaf in-rolling in response to drought relative to the wild type (Supplemental Figure S3).
Apart from opening briefly for phase II, just after lights-on, the wild type closed its stomata in the light and opened them throughout the dark, when gs tracked CO2 uptake ( Figure 4B). In rPPC1-B, stomata stayed open in the light, closed briefly at dusk, and opened slightly throughout the dark, with a small peak prior to dawn ( Figure   4B). The dark conductance of rPPC1-B correlated with the release of respiratory CO2 ( Figure 4A). The internal partial pressure of CO2 inside the leaf (Ci) was highest for the wild type in the light period, when stomata were closed, whereas in rPPC1-B, Ci peaked during the dark period, when stomata were slightly open, but the leaves failed to fix atmospheric CO2, and respiratory CO2 escaped ( Figure 4C).

Impact of Drought on CO2 uptake in Plants lacking PPC1
To test the importance of PPC1 for carbon assimilation during drought, CO2 uptake was measured continuously in whole, young plants (9-leaf-pairs) over 22 days of drought, followed by re-watering ( Figure 4D). CAM develops with leaf age in K. laxiflora (Supplemental Figure S4). The leaves gradually reduce light period CO2 fixation and increase nocturnal CO2 fixation as they develop (Supplemental Figure   S4). A young wild type plant with 9-leaf-pairs thus includes young leaves fixing CO2 mainly in the light via the C3 pathway, and older leaves fixing the majority of their atmospheric CO2 in the dark via PPC and CAM.
When W-W on day 1, wild type, rPPC1-A and rPPC1-B fixed, respectively, 7 %, 6 % and -26 % of their CO2 during the dark, and 93 %, 94 % and 126 % during the light period (Table 1). This indicated that, in well-watered conditions, the young leaves of these young plants performed the majority of the 24 h CO2 uptake (Table 1). On day 1, wild type, rPPC1-A and rPPC1-B fixed 1093, 1008 and 1078 µmol of atmospheric CO2 m -2 in total over 24 h (Table 1). After 7 days without watering, 24 h CO2 fixation was 1345, 1367 and 1506 µmol m -2 . It should be noted that total leaf area was measured at the end of the experiment, so leaf growth and expansion during the experiment could not be accounted for.
After 7 days without water, there was a substantial increase in nocturnal CO2 uptake in wild type and rPPC1-A relative to day 1 (52 % and 40 % of CO2 fixation occurred in the dark, respectively; Table 1). rPPC1-B respired less (-14 %) after 7 days of drought relative to the -26 % dark respired CO2 on day 1 (Table 1). After 13 days of drought, wild type, rPPC1-A and rPPC1-B fixed, respectively, 795, 802 and 145 µmol CO2 m -2 over the 24 h cycle (Table 1). Thus, plants performing CAM (wild type and rPPC1-A) were able to fix over 5-fold more CO2 after 13 days of drought compared to rPPC1-B. Furthermore, after 22 days of drought, wild type, rPPC1-A and rPPC1-B fixed 76, 130 and -30 µmol m -2 during the 24 h light/ dark cycle.
On day 23 without water, the drought-stressed plants were rewatered (see photos in Supplemental Figure S3). After re-watering, CO2 fixation increased rapidly for all plants ( Figure 4D). In addition, wild type and rPPC1-A, displayed pronounced phase III of CAM after re-watering ( Figure 4D). Following soil rehydration, the wild type fixed 931 µmol m -2 CO2 in the dark, compared to 768 µmol m -2 for rPPC1-A, whereas prior to drought rPPC1-A fixed more atmospheric CO2 (Table 1). rPPC1-B fixed CO2 throughout the light period following re-watering, and it also resumed respiratory CO2 loss throughout the dark ( Figure 4D).
As there was an increase in C3 photosynthesis in rPPC1-B , chlorophyll a and b were assayed for LP2 through LP7 ( Figure 4E and 4F). Line rPPC1-A contained significantly more chlorophyll a and b than wild type in LP2, and rPPC1-B contained significantly less chlorophyll a and b than the wild type in LP3 to LP6 ( Figure 4E and 4F).

Characterisation of CAM Gene Transcript Abundance in rPPC1 Lines
Having established that rPPC1-B lacked nocturnal CO2 fixation ( Figure 4A), it was important to investigate the temporal regulation of other CAM-associated genes in the rPPC1 RNAi lines. The transcript abundance of CAM genes in CAM leaves (LP6) was investigated using samples collected every 4 h over a 12-h-light/ 12-h-dark cycle ( Figure 5). PPDK was unchanged relative to the wild type levels in either of the rPPC1 lines ( Figure 5A), but its regulator PPDK-RP was up-regulated in line rPPC1-B ( Figure   5B), consistent with the continuous phosphorylation and inactivation of PPDK ( Figure   2E and 2F). b-NAD-ME was only slightly different from the wild type in the rPPC1 lines, but was lower in rPPC1-B at 22:00 ( Figure 5C).
In light of the marked reduction in starch to half the wild type level in rPPC1-B ( Figure 3B), transcripts associated with starch turnover were also measured. In rPPC1-B, dusk-phased starch breakdown-associated genes (GWD, AMY3a and 3b, PHS1; Figure 5D to 5F and 5J ), and sugar transporters (MEX1, pGlcT; Figure 5L and 5M) were up-regulated, whereas dawn-phased starch breakdown genes (BAM1, BAM3, BAM9; Figure 5G to 5I) and sugar transporters (GPT2; Figure 5K) were downregulated in comparison to the wild type.
Finally, a CHLOROPHYLL A/B BINDING PROTEIN (CAB1) gene was upregulated in rPPC1-B at dawn ( Figure 5N), whereas a potential sucrose sensor connecting growth and development to metabolic status, TREHALOSE 6-PHOSPHATE SYNTHASE7 (TPS7) (Schluepmann et al., 2003), was down-regulated relative to the wild type at 6 and 10 h into the 12-h-dark period ( Figure 5O).

Characterisation of Diel Regulation of Circadian Clock Genes in rPPC1 Lines
Recent studies using PPCK1 RNAi lines of Kalanchoë reported that reduced sucrose 2 h after dawn correlated with perturbation of the central circadian clock . Of the two CIRCADIAN CLOCK ASSOCIATED1 (CCA1) genes in K. laxiflora, only CCA1-2 was down-regulated in rPPC1-B ( Figure

Conditions
Under constant light and temperature (LL) free-running conditions, detached wild type CAM leaves (LP6) of K. laxiflora displayed a robust circadian rhythm of CO2 uptake with a period of approximately 20 h ( Figure 6A). This rhythm was entirely consistent with CAM rhythms reported previously for K. fedtschenkoi and K. daigremontiana (Lüttge and Ball, 1978;Anderson and Wilkins, 1989). The rhythm dampened rapidly in line rPPC1-B ( Figure 6A), whereas rPPC1-A maintained a rhythm that was very similar to that of the wild type ( Figure 6A).
When LL CO2 uptake was measured for well-watered whole plants, rPPC1-B fixed more CO2 than the wild type and line rPPC1-A; 11269, 7414, 8383 µmol CO2 m -2 , respectively ( Figure 6B). Wild type maintained robust oscillations of CO2 exchange under LL conditions, whereas rPPC1-A dampened to arrhythmia after 3 days, and rPPC1-B was arrythmic ( Figure 6B).

Delayed Fluorescence Rhythms were Amplified in rPPC1-B Despite the Loss of the CAM-Associated CO2 Uptake Rhythm
Delayed fluorescence (DF) is rhythmic in Kalanchoë and provides a measure of a chloroplast-derived clock-output that can be used for statistical analysis of circadian period, robustness and accuracy (Gould et al., 2009;Boxall et al., 2017). DF was measured under LL ( Figure 6C), and analysed to calculate rhythm statistics using Biodare (Moore et al., 2014;Zielinski et al., 2014).
Wild type DF oscillations were very similar to those reported previously for K. fedtschenkoi ( Figure 6C; Gould et al., 2009;Boxall et al., 2017). rPPC1-A and rPPC1-B had more robust oscillations ( Figure 6C). The rhythm amplitude increased slightly with time in rPPC1-A and rPPC1-B, but remained relatively constant in the wild type.
The relative amplitude error (RAE) plot showed a wider spread of period for the wild type than for the rPPC1 lines ( Figure 6D). Mean periods were between 21.5 h and 22.1 h when calculated using spectral resampling or fast Fourier transform (nonlinear least squares) methods, respectively ( Figure 6E). Average periods were similar between wild type and the rPPC1 lines. A lower mean RAE was calculated for rPPC1-A and rPPC1-B compared to the wild type ( Figure 6F), supporting statistically the visibly robust and high-amplitude DF rhythm in the rPPC1 lines ( Figure 6C).

Rhythm Characteristics of Core Circadian Clock and Clock-Controlled Genes
Having established that circadian control of CO2 fixation was dampened under LL, and that the circadian control of DF was enhanced under LL in plants lacking PPC1 ( Figure   6A to 6F), it was important to investigate the regulation circadian clock-controlled genes in the rPPC1 lines under LL (Figure 7).
Wild type displayed a rhythm in the transcript abundance of PPC1 that was absent in rPPC1-B ( Figure 7A). PPC2 was rhythmic in wild type and rPPC1-B, but its abundance was lower in rPPC1-B ( Figure 7B). The PPCK1 rhythm was of greater amplitude in rPPC1-B, and the daily transcript peaks occurred 4 to 8 h later than the wild type after the first 24 h of LL ( Figure 7C). PPCK2 and PPCK3 were induced in line rPPC1-B and oscillated with higher amplitude ( Figure 7D and 7E). PPCK2 peaked after the wild type on the second and third 24 h cycles under LL ( Figure 7D and 7E), which was consistent with the induction detected under light/ dark cycles ( Figure 1F and 1G). GPT2, involved in the transport of G6P across the chloroplast membrane, was down-regulated and had lower amplitude in rPPC1-B ( Figure 7F), whereas CAB1 was induced and oscillated robustly ( Figure 7G).
In rPPC1-B, core clock gene CCA1-1 was down-regulated, phase delayed and had lower amplitude than the wild type ( Figure 7H), whereas CCA1-2, TOC1-1 and TOC1-2 were all up-regulated and phase delayed relative to the wild type for the latter two peaks of LL ( Figure 7I, 7J and 7K). PRR7 was up-regulated and more robustly rhythmic in rPPC1-B ( Figure 7L), whereas the rhythms of PRR3/7 and PRR9 dampened in rPPC1-B ( Figure 7M and 7N). Finally, JMJ30/ JMJD5, ELF3, CDF2, RVE1-like, EPR1, FKF1 and GI were up-regulated in rPPC1-B, displaying higher amplitude, phase delays and/ or period lengthening during the LL time course ( Figure   7P to 7V), whereas LNK3-like was down-regulated but remained rhythmic with a lengthening period ( Figure 7O).

Diel Regulation of GC-Signalling and Ion Channel Genes in GC-Enriched
Epidermal Peels of rPPC1-B PHOTOTROPIN1 (PHOT1) encodes a protein kinase that acts as a blue light (BL) photoreceptor in the signal-transduction pathway leading to BL-induced stomatal movements (Kinoshita et al., 2001). In rPPC1-B, PHOT1 transcripts were up-regulated relative to the wild type and peaked 8 h into the 12-h-light period ( Figure 8A). CRYPTOCHROME2 (CRY2) is a photoreceptor that regulates BL responses, including the entrainment of endogenous circadian rhythms (Somers et al., 1998), and stomatal conductance via an indirect effect on ABA levels (Boccalandro et al., 2012).
In rPPC1-B epidermal peels, CRY2 transcript levels were up-regulated and peaked at dusk, whereas CRY2 peaked 8 h into the dark in the wild type ( Figure 8B).
In Arabidopsis, GC localised b-CARBONIC ANHYDRASE1 (CA1) and CA4 are involved in CO2 sensing in GCs, with the ca1ca4 mutant displaying impaired stomatal control in response to CO2 (Hu et al., 2010). In rPPC1-B, a b-carbonic anhydrase gene was much more strongly induced at dawn relative to the wild type ( Figure 8C). PATROL1 controls the tethering of the proton ATPase AHA1 to the plasma membrane, and is essential for opening in response to low CO2 and light (Hashimoto-Sugimoto et al., 2013). The cycle of PATROL1 in epidermal peels was slightly different between rPPC1-B and wild type, with rPPC1-B peaking 4 h into the dark, but the wild type peak at 4 h into the light ( Figure 8D).  Park et al., 2009;Santiago et al., 2009). In rPPC1-B epidermal peels, RCAR3 transcripts were induced spanning dusk and the first half of the dark period, and peaked at least 4 h earlier than the wild type ( Figure 8J).

CONVERGENCE OF BLUE LIGHT
In rPPC1-B, the GUARD CELL OUTWARD RECTIFYING K + CHANNEL (GORK), which is known to be involved in the regulation of stomata according to water status (Ache et al., 2000), peaked 4 h earlier than wild type ( Figure 8K). STELAR K + OUTSIDE RECTIFIER (SKOR) is a selective outward-rectifying potassium channel (Gaymard et al., 1998). It peaked at dawn in the wild type, whereas in rPPC1-B, SKOR transcript levels reached their daily minimum at dawn ( Figure 8L). The plasmamembrane localised ALUMINIUM ACTIVATED MALATE TRANSPORTER12 (ALMT12), which is involved in dark-, CO2-, abscisic acid-and water deficit-induced stomatal closure, was elevated relative to the wild type at dawn in rPPC1-B ( Figure   8M). An E3 UBIQUITIN-PROTEIN LIGASE (RMA1) promotes the ubiquitination and proteasomal degradation of aquaporin PIP2:1, which is known to play a role in GC regulation (Grondin et al., 2015). RMA1 was induced 3-fold at dawn in rPPC1-B compared to the wild type ( Figure 8N). EMBRYO SAC DEVELOPMENT ARREST39 (EDA39) is a calmodulin binding protein that promotes stomatal opening (Zhou et al., 2012). In rPPC1-B, EDA39 was up-regulated relative to the wild type and peaked at dusk ( Figure 8O).

SALT OVERLY SENSITIVE2 (SOS2) is a Calcineurin B-Like (CBL)-interacting
protein kinase involved in the regulatory pathway for intracellular Na + and K + homeostasis and salt tolerance (Liu et al., 2000). SOS2 has also been demonstrated to interact with and activate the vacuolar H + / Ca 2+ antiporter CAX1, thereby functioning in cellular Ca 2+ homeostasis; an important function during stomatal opening and closing (Cheng et al., 2004). SOS2 transcript levels were elevated relative to the wild type at all time points, and in particular, it was induced ~5-fold at its peak 4 h into the dark period in epidermal peels of rPPC1-B relative to the wild type ( Figure 8P). The Ca 2+ -ATPase2 (ACA2) and ENDOMEMBRANE-TYPE Ca 2+ -ATPASE4 (ECA4) catalyze the hydrolysis of ATP coupled with the translocation of calcium from the cytosol into the endoplasmic reticulum and/ or an endomembrane compartment (Jezek and Blatt, 2017). ACA2 and ECA4 were induced by between ~4-fold to ~8-fold in rPPC1-B epidermal peels relative to the wild type, particularly when they reached peak levels at 4 h into the 12-h-dark period ( Figure 8Q and 8R).
Finally, transcription factor MYB60 is involved in stomatal opening in response to light and also promotes GC deflation in response to water deficit (Cominelli et al., 2005). In rPPC1-B, MYB60 was induced ~5-fold relative to the wild type at dawn, although it also maintained a dusk phased peak of transcript abundance like the wild type ( Figure 8S). MYB61 functions as a transcriptional regulator of stomatal closure (Liang et al., 2005). It was down-regulated in rPPC1-B relative to wild type, and peaked 8 h later in rPPC1-B, 4 h into the dark period, whereas wild type MYB61 peaked in the light 4 h before dusk ( Figure 8T).

Fixation in an Obligate CAM Plant
PPC1 loss-of-function mutants of the obligate CAM species K. laxiflora were generated using RNAi (Figure 1 and 2). Although PPC2 transcripts were induced almost 7-fold in the most strongly silenced line, rPPC1-B ( Figure 1B), this did not compensate for the loss of the CAM-associated PPC1 ( Figure 1A and 2A). Total extractable PPC activity was below the level of detection in rPPC1-B ( Figure 2E), and PPC-catalyzed dark period fixation of atmospheric CO2 was abolished ( Figure 4A and 4D). Furthermore, transcript abundance of the PPC1 regulatory protein kinase, PPCK1, was reduced at the time of its circadian clock-mediated nocturnal peak ( Figure   1E), and PPCK2 and PPCK3 were up-regulated ( Figure 1F and 1G), but no phosphorylation of PPC was detected in line rPPC1-B ( Figure 2B). This outcome was most likely due to the fact that the major substrate for PPCK1, namely the CAMassociated PPC1, had been down-regulated such that there was no substrate for any PPCK activity that resulted from translation of the available PPCK1, PPCK2, or PPCK3 transcripts ( Figure 1). This suggests that the induction of PPCK2 and PPCK3 transcripts may have been a futile attempt to compensate for the loss of PPC1 through attempting to enhance the nocturnal activation of any remaining PPC protein.

Physiological Consequences of Silencing CAM Isogene PPC1 in K. laxiflora
In mature leaves (LP6) and young whole plants (9-leaf-pairs stage) of rPPC1-B, the timing of primary CO2 fixation switched from the dark period to the light ( Figure 4A and 4D), and the main period of stomatal conductance shifted to the light period ( Figure   4B). This revealed that primary atmospheric CO2 fixation was occurring in the light via Gas exchange measurements using whole plants revealed that rPPC1-B fixed 82% of the total amount of atmospheric CO2 fixed over 24 h by the wild type (Day 1, Table 1). This emphasised that well-watered wild type fixed only 18 % of their daily CO2 in the dark via PPC (Table 1). However, this changed dramatically during drought stress ( Figure 4D). The wild type and intermediate line rPPC1-A increased nocturnal CO2 fixation throughout 22 days of drought ( Figure 4D); fixing 100 % of their atmospheric CO2 in the dark via PPC after both 13 and 22 days of drought (Table 1).
By contrast, rPPC1-B failed to achieve net dark period atmospheric CO2 fixation throughout 22 days of drought stress ( Figure 4 and Table 1). Furthermore, the lightperiod CO2 fixation of rPPC1-B collapsed rapidly after day 7 of water-with-holding, dropping from 1759 µmoles CO2 m -2 over the 12-h-light period on day 7, to 161 µmoles m -2 on day 13, and net respiratory CO2 loss of -22 µmoles m -2 on day 22 (Table 1).
These data demonstrate very clearly the importance of a fully functional CAM system for continued atmospheric CO2 fixation throughout a period of drought lasting just over 3-weeks.
Despite the inability of rPPC1-B to induce CAM during drought, the domed shape of its respiratory CO2 release throughout the 12-h-dark period ( Figure 4D) indicated active refixation of respiratory CO2 that peaked around the middle of each 12-h-dark period. This revealed that these plants were still capable of performing a version of 24-h photosynthetic physiology approximating CAM-idling (Winter, 2019).
This was further supported by the low level of malate accumulation at dawn that was achieved by rPPC1-B ( Figure 3A), and the decline and/ or midday dip in atmospheric CO2 fixation in the light period ( Figure 4A and 4D).
Our data also revealed that the intermediate line rPPC1-A was less able to adapt to drought by inducing CAM when compared to the wild type. After only 7 days of drought, the wild type had up-regulated dark period CO2 fixation and downregulated light period CO2 fixation such that 52 % (698 µmoles m -2 ) of daily atmospheric CO2 fixation was occurring in the 12-h-dark period; a sharp rise from 7 % (78 µmoles m -2 ) on day one (Table 1). By contrast, line rPPC1-A had only progressed to performing 40 % (553 µmoles m -2 ) of its daily atmospheric CO2 fixation in the dark period on day 7 of drought (Table 1) and decreased the proportion of atmospheric CO2 fixed in the light period, in response to drought ( Figure 4D and Table 1).
Upon re-watering on day 23, wild type and rPPC1-A returned to performing CAM, including pronounced phase II and IV at the start and end of the light period, respectively ( Figure 4D). It was noteworthy that the wild type performed more dark period CO2 fixation than rPPC1-A following re-watering ( Figure 4D), suggesting that the wild type had an advantage in terms of achieving more dark CO2 fixation, both in the well-watered young plants on days 1 through 10 of drought ( Figure 4D), and when subsequently recovering higher levels of dark CO2 fixation upon re-watering on day-23 ( Figure 4D). Line rPPC1-B also bounced back to its normal, pre-drought physiology after re-watering, which was consistent with the re-watering response reported previously for a range of facultative, weak-CAM species, including M. crystallinum, Portulaca oleracea, P. umbraticola, T. triangulare, and various Calandrinia species (Winter and Holtum, 2014;Holtum et al., 2017;Winter, 2019).
Although the rPPC1-B plants fixed all of their CO2 in the light, especially during a pronounced phase II in the hours after dawn, the CO2 uptake pattern was not constant over the light period ( Figure 4A). For example, in LP6 of a well-watered plant, CO2 fixation dropped from ~8 µmol m -2 s -1 during phase II after dawn, to ~2 to 3 µmol m -2 s -1 around 4 h after lights on ( Figure 4A), and stomatal conductance reduced to a similar extent ( Figure 4B). The observed stomatal closure could not have been due to a high malate concentration and associated internal release of CO2 in the mesophyll ( Figure 3A). These data are consistent with the current hypothesis that the circadian clock output may drive the closure of stomata and the associated decline in atmospheric CO2 fixation, even when a CAM leaf has not produced malate during the preceding dark period (Von Caemmerer and Griffiths, 2009).
Overall, with respect to CAM physiology, the data presented here provide strong support for the long-held view that an ability to use CAM, and rely increasingly on CAM in response to drought stress, provides a genuine adaptive advantage in terms of prolonging net atmospheric CO2 fixation during drought progression ( Figure   4D) (Kluge and Fischer, 1967;Osmond, 1978). The data demonstrated that the wild type and line rPPC1-A achieved, respectively, net atmospheric CO2 fixation of 18,568 and 19,071 µmoles m -2 over the 22-day drought progression, whereas line rPPC1-B only achieved 13,983 µmoles m -2 net atmospheric CO2 fixation (Supplemental Table  S2). Thus, the wild type, with its fully functional CAM system, was able to fix 33 % greater more CO2 over the entire drought treatment period. line redirected more of its starch degradation through starch phosphorylase. However, this would require export of G6P via GPT, and yet GPT2 transcript levels were repressed ( Figure 5K). In addition, GWD was also induced ( Figure 5D), suggesting that the mesophyll cells also increased starch breakdown via GWD and b-amylase, although all measured BAM transcripts were repressed ( Figure 5G, 5H and 5I).

Consequences of the loss of
As MEX1 and pGlcT were induced with a peak phased to dusk ( Figure 5L and 5M), and GPT2 was repressed at 2 h into the light period ( Figure 5K), we favour the proposal that the rPPC1-B line redirects more of its nocturnal starch degradation via the amylolytic route, with MEX1 and pGlcT acting to export the products of starch breakdown from the chloroplast in the dark as maltose and glucose, respectively. This proposal is also consistent with the overall return of rPPC1-B to a more C3-like pathway of CO2 fixation and associated metabolism, especially given that the accepted C3 pathway of leaf starch breakdown in the dark uses amylolytic route involving the MEX1 and pGlcT transporters (Zeeman et al., 2010).

Core clock gene regulation under LD cycles in the absence of PPC1
In CAM Kalanchoës, the circadian clock optimises the timing of the daily cycle of dark CO2 fixation via PPC, and light period malate turnover and CO2 refixation via RuBisCO (Hartwell, 2006;Boxall et al., 2017). In line rPPC1-B, some, but not all, core circadian clock-associated genes displayed altered temporal profiles of transcript abundance under LD cycles ( Figure 5O to 5DD). CCA1-1, FKF1 and GI transcript levels were remarkably consistent between all three lines ( Figure 5O, 5CC and 5DD), revealing that individual genes belonging to the core and evening loops of the core oscillator

Impact of PPC Silencing on the CAM Circadian Rhythm of CO2 Fixation
Under continuous light and temperature (LL) conditions, the CAM-associated circadian rhythm of CO2 fixation was dampened towards almost complete arrhythmia in detached leaves and whole, young plants of rPPC1-B, and CO2 was fixed continuously ( Figure 6A and 6B). The collapse of the CO2 circadian rhythm suggests that CO2 fixation by RuBisCO in rPPC1-B was not subject to robust and high amplitude circadian control; certainly not in comparison to the rhythm of atmospheric CO2 fixation during CAM in the wild type ( Figure 6B). In K. daigremontiana, it has been suggested that RuBisCO made a large contribution to the observed rhythm of CO2 assimilation under LL conditions, as the level of malate did not oscillate (Wyka and Lüttge, 2003 The rhythm of CO2 fixation in LL conditions is regulated via phosphorylation of PPC by circadian clock controlled PPCK1 (Hartwell, 2006;Boxall et al., 2017). In rPPC1-B, PPCK1 was slightly up-regulated and displayed a robust, high amplitude rhythm ( Figure 7C). However, as PPCK1 had no substrate PPC to phosphorylate ( Figure 2B), the data presented here provide further support for the proposal that the circadian control of CAM via PPCK1 is dependent on PPC phosphorylation, and not in some way linked directly to the circadian oscillations of PPCK1 transcript and activity levels .
Furthermore, in addition to the robust rhythm of PPCK1, transcript levels of PPCK2 and PPCK3 were also induced and became dramatically more rhythmic under LL (Figure 7). The direct functions of PPCK2 and PPCK3 are currently unknown, but they may be involved in phosphorylating other PPCs in different tissues or cell types, with one possible location being the GCs, as supported by the epidermal peels RT-qPCR data in Figure 8. Overall, the induction and robust rhythmicity of PPCK2 and PPCK3 transcripts was futile with respect to CAM circadian physiology, as they failed to drive a robust rhythm of CO2 assimilation under LL conditions ( Figure 6A and 6B).

Regulation of core clock genes in rPPC1-B
Similar to the LL CO2 fixation phenotype reported here for rPPC1-B, transgenic K. fedtschenkoi lines lacking PPCK1 also lost the CO2 fixation rhythm, and the transcript oscillations of many core clock genes were altered . However, different core clock genes were perturbed in rPPC1-B compared to those whose rhythmic regulation changed in the rPPCK1 lines of K. fedtschenkoi  cf. Figure 7). In rPPC1-B, CCA1-1 transcript oscillations dampened, and both TOC1-1 and TOC1-2 transcripts were up-regulated and oscillated with delayed peak phase relative to the wild type ( Figure 7H, 7J and 7K). By contrast, in line rPPCK1-3, CCA1-1, CCA1-2 and TOC1-2 dampened rapidly towards arrhythmia under LL free-running conditions, whereas TOC1-1 was up-regulated and rhythmic . Overall, these differences between the rhythms of core clock genes in rPPC1-B and rPPCK1-3 revealed that the clock responded very differently to the silencing of two interconnected genes that lie at the heart of nocturnal CO2 fixation. Elucidating the mechanistic basis for these differences should be a fruitful avenue for further investigation, especially in light of the proposed cross-talk between CAM-associated metabolites and regulation within the core clock , which was further supported by the clock gene phenotypes reported here (Figure 7).

Interactions between sugars linked to CAM and the core circadian clock
In both LD and LL conditions, TOC1-1, TOC1-2, PRR7 and PRR3/7 were up-regulated in rPPC1-B, although only at certain time points in the subjective light and dark periods in the case of PRR3/7 under LL ( Figure 5I to 5L, and 7J to 7M). In Arabidopsis, PRR7 is required for sensing metabolic status and coordinating the clock with photosynthesis (Haydon et al., 2013). In rPPC1-B, the low sucrose at 2 h after dawn ( Figure 3C) may be sensed via a mechanism involving PRR7 . However, in Kalanchoë, a related-PRR gene, PRR3/7, was the more abundant transcript and displayed a transcript peak 2 h before dusk under LD ( Figure 5T). In the wild type, PRR3/7 may function as part of a signal transduction pathway that senses metabolic status at dusk when primary CO2 fixation begins (Haydon et al., 2013;Boxall et al., 2017).
A further insight into the perturbation of sugar sensing and signalling in line rPPC1-B resulted from the discovery that the regulation of TREHALOSE 6-PHOSPHATE SYNTHASE7 (TPS7) was altered in rPPC1-B ( Figure 5O). TPS7 is a member of the class II TPS genes that have both inactive TPS and TP-phosphatase (TPP) domains, but which are proposed to play a role in signalling metabolic status of the cell in Arabidopsis (Ramon et al., 2009). In rPPC1-B, TPS7 was down-regulated relative to the wild type ( Figure 5N). The TPS7 ortholog in K. fedtschenkoi was also down-regulated in transgenic lines that had reduced CAM and less robust rhythms of CAM-associated CO2 fixation due to the silencing and down-regulation of either PPCK1, b-NADME or PPDK (Dever et al., 2015;Boxall et al., 2017). Thus, TPS7 in Kalanchoë may function in the sensing and/ or signalling of the cellular metabolic status. However, there are currently no data relating to a potential role for TPS7 in relation to sugar sensing leading to metabolic entrainment of the core circadian oscillator in either Kalanchoë or other plant species (Ramon et al., 2009).

Robust rhythms of DF in the absence of rPPC1-B, but arrhythmic DF in the absence of the PPC1 circadian phospho-regulator PPCK1
One of the most noteworthy circadian phenotypes related to the contrast between the current findings of robust and increased amplitude DF rhythmicity in rPPC1-A and rPPC1-B ( Figure 6C), and the arrhythmic DF reported previously for the K. are the most likely candidates for driving the induction and robust LL rhythmicity of DF in rPPC1-B. Conversely, the decline in the amount and rhythmicity of these transcripts in rPPCK1-3 may play a role in the dampening of the DF rhythms in that transgenic line .
These results also allow us to propose that robust rhythmicity of the CAMassociated CO2 fixation rhythm in wild type Kalanchoë is most likely to rely on CCA1-1, PRR3/7, PRR9, and LNK3-like. However, it must be stressed that the limited set of core clock and clock-associated genes profiled under LL to date in rPPC1-B and rPPCK1-3 means that there are likely to be other clock-associated genes involved in driving robust CAM CO2 fixation rhythms, and robust DF rhythms.

Perturbation of diel rhythms of gene transcript oscillations in GCs
A key gap in current understanding of the molecular-genetics and physiology associated with CAM centres on the cell signalling mechanisms that mediate the inverse pattern, relative to C3, of stomatal opening and closing (Borland et al., 2014; Males and Griffiths, 2017). It has been proposed that GCs of CAM leaves and stems respond directly to the internal supply of CO2. This theory has recently led several groups to use whole leaf RNA-seq datasets to investigate alterations, relative to C3 leaves, in the temporal phasing of known GC regulatory genes and membrane transporters (Abraham et al., 2016;Wai and VanBuren, 2018;Yin et al., 2018;Heyduk et al., 2019;Moseley et al., 2019). However, these studies did not use enriched GCs as the source of RNA, and therefore the data that were mined represented all leaf cell types, including palisade and spongy mesophyll, GCs, subsidiary cells, phloem, phloem companion cells, xylem, bundle sheath, and water storage parenchyma in Agave and pineapple. Thus, the re-scheduling of the temporal patterns of candidate GC genes in these previous datasets may have been complicated by transcripts from the same genes that were functional in other cell types of the leaves. Separated epidermal peels from Kalanchoë CAM leaves are enriched for intact guard cells. We leveraged this feature in order to compare the temporal pattern of transcript regulation between GC-enriched epidermal peels of the wild type and rPPC1-B (Figure 8). A wide range of genes known to be involved in stomatal opening and closing displayed alterations in both transcript abundance, and/ or the timing of the daily transcript peak ( Figure 8A to 8U). In particular, the temporal re-phasing and/ or down-regulation/ up-regulation of GC transcripts including HT1, OST1, SLAC1, RCAR3, GORK, SKOR, ALMT12, RMA1, EDA39, SOS2, ACA2, ECA4, and MYBs 60, and 61, revealed that the temporal control of the regulatory circuits that mediate the opening and closing of stomata was altered in rPPC1-B (Figure 8). It must however be emphasised that transcript changes alone are often difficult to interpret as they do not provide direct evidence concerning the encoded protein's abundance and activity.
Despite this caveat, the GC transcript abundance alterations in rPPC1-B (Figure 8) correlated with the 12-h-shift in stomatal opening to the light period in rPPC1-B at the level of whole leaf physiology ( Figure 4B). Taken together, these results therefore support the theory that the measured changes in transcript abundance and temporal patterns for a range of guard cell regulatory genes were connected to the measured changes in stomatal opening and closing.
Thus, the up-regulation of HT1, SKOR and MYB61 in the wild type relative to rPPC1-B, and the down-regulation of a wide range of genes, including SLAC1, PP2C, SOS2, ACA2, ECA4, and MYB60, in the wild type, are likely to be important regulatory changes that facilitate nocturnal stomatal opening and light period closure (Figure 4 and 8). As MYB60 is required for stomatal opening in response to light in C3 Arabidopsis (Cominelli et al., 2005), it is particularly noteworthy that its transcript abundance in wild type K. laxiflora GC-enriched epidermal peels peaked at dusk, when stomata open in the wild type ( Figure 8LL). Furthermore, in rPPC1-B, MYB60 also had a dramatic 3-to 4-fold induction at dawn relative to the wild type, suggesting that high MYB60 at the start of the light period may play a key role in the observed light period stomatal opening of the rPPC1-B line. It was also notable that rPPC1-B continued to have a peak of MYB60 transcripts phased to dusk, as observed for the wild type, which correlated with the fact that rPPC1-B did open its stomata slightly throughout the dark period ( Figure 4B). However, this nocturnal stomatal opening was futile in terms of atmospheric CO2 fixation, as rPPC1-B released respired CO2 from its leaves throughout the dark period ( Figure 4A). Overall, MYB60, as well as several other misregulated guard cell signalling, ion channel and metabolite transporter genes ( Figure   8 and Supplemental Figure 5S), represent key targets for future genetic manipulation experiments in transgenic Kalanchoë aimed at understanding the important regulators underpinning the inverse stomatal control associated with CAM.

Informing biodesign strategies for engineering CAM into C3 crops
Efforts are underway to engineer CAM and its associated increased WUE into C3 species as a means to develop more climate-resilient crop varieties that can continue to fix CO2 and grow in the face of drought, whilst using water more wisely than C3 varieties (Borland et al., 2014;Borland et al., 2015;Lim et al., 2019). The data presented here for the rPPC1 loss-of-function lines of K. laxiflora confirm experimentally the proposed core role of PPC1 for efficient and optimised CAM. Our results also provide encouragement that the level of over-expression of a CAMrecruited PPC1 gene introduced into an engineered C3 species may not need to be as high as the level found in extant obligate CAM species, because reducing PPC1 levels to only 43 % of wild type activity in line rPPC1-A led to plants that were still capable of 29 full CAM, and that fixed more CO2 than the wild type over 3 weeks of drought ( Figure   4D).
The results presented here also emphasise that only certain sub-components of the core circadian clock are essential for the temporal optimisation of CAM in Kalanchoë, which in turn further simplifies the challenge of achieving correct temporal control of an engineered CAM pathway introduced into a C3 species. Transgenic manipulation of the expression and regulation of CCA1-1, PRR3-7, PRR9 and LNK3like in Kalanchoë using RNAi, over-expression and/ or CRISPR-Cas mediated gene editing, will allow the further refinement of the evolving model for the subset of core clock genes that form the transcription-translation feedback loop that underpins the temporal optimisation of CAM.

Plant materials
Kalanchoë laxiflora were propagated clonally from leaf margin adventitious plantlets using the same clonal stock originally obtained from the Royal Botanic Gardens (

Generation of transgenic K. laxiflora lines
An intron containing hairpin RNAi construct was designed to target the silencing of both copies of the CAM-associated PPC1 gene in the tetraploid K. laxiflora genome (JGI Phytozome accession numbers: Kalax.0018s0056.1 and Kalax.0021s0061.1, which are the K. laxiflora orthologues of previously characterised K. fedtschenkoi CAM-associated PPC1, GenBank accession: KM078709). A 323 bp fragment was amplified from CAM leaf cDNA using high fidelity PCR with KOD Hot Start DNA Polymerase (Merck, Germany). The amplified fragment spanned the 3' end of the PPC1 coding sequence and extended into the 3' untranslated region to ensure specificity of the silencing to both of the aforementioned CAM-associated PPC1 gene copies. Alignment of the 323 bp region with the homologous regions from the six other plant-type PPC genes in the K. laxiflora genome demonstrated that none of the other PPC genes shared any 21 nucleotide stretches that were an exact match for the 323 bp PPC1 RNAi fragment. Thus, the RNAi construct in the hairpin RNA binary vector used to generate the stable transgenic lines was predicted to silence only the two copies of the CAM-associated PPC1, and to be equally specific and efficient at silencing both copies. This specificity and equality of silencing efficiency was confirmed by the RT-qPCR data in Figure 1, as the qPCR primers targeted both copies of the CAM-associated PPC1, and so the quantitative signal is representative of the averaged signal for the transcript abundance of the two gene copies.  Skoog, 1962;Gamborg et al., 1968) and 3 % sucrose. 4-to 6-week-old seedlings raised in tissue culture were chopped into explants for transformation in a sterile laminar flow bench using a sterile scalpel. The explants were dipped in the Agrobacterium suspension carrying the PPC1 RNAi binary vector, and co-cultivated and regenerated as described previously (Dever et al., 2015).

High-throughput leaf acidity and starch content screens
Leaf acidity (as a proxy for leaf malate content) and leaf starch content were screened with leaf disc stains using chlorophenol red and iodine solution at both dawn and dusk as described by Cushman et al. (2008). For each transgenic line, leaf discs were sampled from LP6 in triplicate at 1 hour before dawn and 1 hour before dusk and stained in a 96 well plate format.

Net CO2 exchange measurement
Gas exchange measurements were performed using a custom-built, 12-channel IRGA system (PP Systems, Hitchin, UK), which allowed individual environmental control (CO2/ H2O) and measurement of rates of CO2 uptake for each of 12 gas exchange cuvettes, with measurements collected every ~18 min. The system was described in full by Dever et al. (2015), but was expanded here with the addition of a further 6 cuvettes, thereby doubling scope for replication and throughput. All experiments were repeated at least 3 times using 3 separate individual young plants (9-leaf-pairs), or detached LP6 from three separate clonal plants of each line. Representative gas exchange traces are shown. Replicated wild type and rPPC1 lines were compared in neighbouring gas exchange cuvettes during each experimental run, such that the data are directly comparable between each line. As the entire gas exchange system was housed in a Snijders Microclima MC-1000 growth cabinet, all 12 gas exchange cuvettes were under identical conditions in terms of light intensity and temperature.

Net CO2 exchange using the LI-COR 6400XT system
The gas exchange of mature CAM leaves (LP6) was measured over a 12-h-light, 25˚C, 60 % humidity: 12-h-dark, 15˚C, 70 % humidity cycle using an infra-red gas analyser (LI-6400XT, LI-COR, Inc) attached to a large CO2 gas cylinder. Data was logged every 10 minutes using an auto-program that tracked the light and temperature regimes of the growth cabinet.

Leaf malate, starch and sucrose content
LP6 (full CAM in wild type) from mature plants were sampled into liquid nitrogen at the indicated times and stored at -80˚C until use. The frozen leaf samples were prepared and assayed for malate and starch as described by Dever et al. (2015) using the published methods for assaying malate in an enzyme-linked spectrophotometric assay (Möllering, 1985), and starch (Smith and Zeeman, 2006). Sucrose, glucose, and fructose were assayed according to the manufacturer's protocol (Megazyme Technologies).

Chlorophyll Assays
Chlorophyll was assayed from mature greenhouse grown plants from leaf pairs 2-7.
Chlorophyll was extracted twice from 0.5 cm diameter leaf discs in 2 ml 80 % (v/v) acetone and homogenised in a bead beater (PowerLyzer 24; Mo-Bio, Inc). The tubes were centrifuged at full speed in a bench top microfuge at 4˚C for 2 min, and the supernatants were combined and transferred to a new tube and protected from the light. Absorbance was read at 663 nm and 645 nm, and chlorophyll contents were calculated according to the published method (Arnon, 1949).

Total RNA isolation and RT-qPCR
Total RNA was isolated from 100 mg of frozen, ground leaf tissue using the Qiagen and Kalax.1110s0007.1; Arabidopsis orthologue AT2G30720.1). Gene expression in a pool of RNA generated from LP6 samples collected every 4 hours over a 12-h-light/ 12-h-dark cycle was set to 1. Primers for RT-qPCR analyses are listed in supplemental table S1.

Immunoblotting
Total protein extracts of K. laxiflora leaves were prepared according to Dever et al.
(2015). One-dimensional SDS-PAGE and immunoblotting of leaf proteins was carried out following standard methods. Blots were developed using the ECL system (GE Healthcare, UK). Immunoblot analysis was carried out using antisera to PPC raised against purified CAM leaf PPC from K. fedtschenkoi, kindly supplied by Prof. H.G.
Nimmo, University of Glasgow (Nimmo et al., 1986), and the phosphorylated form of PPC raised against a phospho-PPC peptide from barley, and kindly supplied by Prof.

Growth measurements
Mature plants of wild type and the two rPPC1 lines were grown from developmentally synchronized clonal leaf plantlets in greenhouse conditions for 4 months. At the start of the drought treatment all the plants were watered to full capacity. Water was withheld from 10 replicate plants of each line for 28 days, and 10 plants were maintained well-watered over the same period. The plants were harvested as separated aboveground (shoot) and below-ground (root) tissues, weighed to determine fresh weight, and then dried in an oven at 60˚C until they reached a constant dry weight.

PPC assays
Frozen leaf tissue was ground in liquid nitrogen with a small quantity of acid washed sand and the relevant enzyme specific extraction buffer (approximately 1 g tissue to 3 ml of extraction buffer). Extracts were prepared and PPC assays were performed according to the extraction, desalting and assay buffer conditions described previously (Dever et al., 2015).

Determination of the apparent Ki of PPC for L-malate in rapidly desalted leaf extracts
The apparent Ki of PPC for L-malate was determined using leaf extracts that were rapidly desalted as described by Carter et al. (1991). LP6 were collected at 10:00 (2 h before the end of the 12-h-light period), and 18:00 (middle of the 12-h-dark period) from three biological replicates of wild type, rPPC1-A and rPPC1-B. The apparent Ki of PPC activity for feedback inhibition by L-malate was determined according to the method described by Nimmo et al. (1984), with the modifications to the range of Lmalate concentrations added to the assays as described by Boxall et al. (2017).

PPDK assays
PPDK assays were performed using the extraction and assay buffers described previously (Kondo et al., 2000;Dever et al., 2015), with the addition of NADH and G6P (Salahas et al., 1990), and Cibercron Blue (Burnell and Hatch, 1986). Briefly, 0.3 g of powdered leaf tissue that had previously been ground to a fine powder in liquid nitrogen was extracted in 1 ml of ice-cold extraction buffer containing 100 mM Tris pH 8.0, 10 mM DTT, 1 mM EDTA, 1 % Triton, 2.5 % w/v PVPP, 2 % PEG-20000, 10 mM MgCl2, 1 mM PMSF, 2 µM orthovanadate, and 10 µM Cibercron Blue, by grinding with a small quantity of acid washed sand in a pestle and mortar.
Extracts were vortexed for 30 s and the pH was adjusted to pH 8.0. Extracts were then placed on ice for 10 mins before spinning them at full speed in a benchtop microfuge at 4˚C. The supernatant (500 µl) was desalted using PD minitrap Sephadex-G25 columns (GE Healthcare). The desalting buffer contained: 100 mM Tris-HCl pH 8.0, 10 mM MgCl2, 10 mM DTT, 0.1 mM EDTA, 10 µM Cibercron Blue.

Delayed fluorescence (DF) measurements
The imaging system for DF was identical to the luciferase and delayed

Accession numbers
Sequence data associated with this article are available via the JGI Phytozome portal for the K. laxiflora genome: https://phytozome.jgi.doe.gov Kalanchoë laxiflora v1.1. and the specific gene IDs for each gene measured in this work are provided in Supplemental Table 1.

rPPC1-B
In rPPC1-B, stomatal opening shifted to the light period ( Figure 4A to 4C). It was thus important to investigate whether the temporal pattern of GC-specific regulatory genes was rescheduled in rPPC1-B. RNA was isolated from epidermal peels separated from mature leaves (LP6 to LP8). Peels were enriched with intact stomatal GCs engaged in the CAM pattern of opening and closing in the wild type. PPC1 and PPC2 transcripts were low in epidermal peels (Supplemental Figure   S5A and S5B) when compared to whole leaves ( Figure 1A). PPC3 and PPC4 were more abundant transcripts in wild type epidermal peels ( Figure S5C and S5D) relative to whole leaves ( Figure 1C and 1D). Furthermore, PPC3 and PPC4 were up-regulated in epidermal peels of rPPC1-B compared to the wild type, and their transcript abundance peaked 4 h into the dark, whereas in the wild type both transcripts peaked at the end of the dark period or dawn ( Figures S5A to S5D).
CAM-specific PPCK1 was up-regulated in rPPC1-B epidermal peels compared to wild type, whereas PPCK2, which does not currently have a proposed role in the regulation of CAM PPC1 in the mesophyll, was unchanged ( Figure S5E and S5F).
Both PPCK2 and PPCK3 were more abundant than PPCK1 in epidermal peels, suggesting their encoded proteins may function to regulate the activity of the GC PPC(s) ( Figure S5F and S5G). Most strikingly, PPCK3 was up-regulated ~5-fold at 4 h after lights-on in rPPC1-B epidermal peels relative to the wild type ( Figure S5G).
In C3 plants, starch is degraded before dawn to fuel stomatal opening (Blatt, 2016). In Arabidopsis GCs, BAM1 encodes the major starch degrading enzyme (Valerio et al., 2010;Prasch et al., 2015;Horrer et al., 2016;Santelia and Lunn, 2017) that functions with AMY3 to mobilize starch at dawn, releasing maltose in the chloroplasts (Horrer et al., 2016). In rPPC1-B epidermal peels, GWD was halved relative to wild type at its peak, 8 h into the 12 h light period ( Figure S5H). AMY3a and AMY3b were up-regulated at dusk and in the second half of the light period, respectively ( Figure S5I and S5J). BAM1 was up-regulated at dawn compared to wild type, and peaked with a very similar transcript abundance to wild type at 20:00 h, 8 h into the dark period ( Figure S5K), whereas BAM3 was ~3-fold lower in rPPC1-B relative to the wild type at the time of its nocturnal peak, which also occurred 8 h into the dark ( Figure S5L). BAM9 in rPPC1-B rose later than the wild type, lagging behind the wild type at dusk and 4 h into the dark, and it stayed ~5-fold higher at dawn and 4 h into the light, when the wild type reached its daily trough ( Figure S5M). BAM9 in Arabidopsis is predicted to be catalytically inactive and has no known function to date (Monroe and Storm, 2018). The peak of PHS1 was delayed by 4h peaking at the light/ dark transition ( Figure S5N), and MEX1 was down-regulated in epidermal peels of rPPC1-B, peaking at 08:00 in the light, 4 h after the wild type peak ( Figure S5O).

Diel Regulation of Core Circadian Clock genes in Epidermal Peels of rPPC1-B
The circadian clock genes PRR3/7 and PRR7 were up-regulated in epidermal peels of rPPC1-B relative to wild type ( Figure S5P and S5Q), with a similar diel pattern to that measured in whole CAM leaves ( Figure 5S and 5T). PRR3/7 peaked 4 h earlier in rPPC1-B, whereas PRR7 transcript levels were greater throughout the light period in rPPC1-B ( Figure S5P and S5Q). LUX is a component of the morning transcriptional feedback circuit within the clock, and acts as a transcription factor that directly regulates the expression of PRR9 by binding specific sites in its promoter (Helfer et al., 2011). In epidermal peels, LUX was down regulated >7-fold in rPPC1-B compared to the wild type at its peak, 4 h into the light, and also peaked 4 h later than the wild type ( Figure S5R).

rPPC1-B
GCs perceive CO2 and regulate stomatal aperture via ABA signalling and reactive oxygen species, with low CO2 mediating stomatal opening and high CO2 causing closure (Chater et al., 2015). NADP oxidoreductase encodes an NADP-binding Rossman-fold super-family protein that was only detected in epidermal peels in Kalanchoë (Boxall, Dever, Kadu and Hartwell, unpublished observation). The enzyme encoded by this gene may function to generate the H2O2 burst induced by ABA as part of the stomatal closure signalling pathway (Daszkowska-Golec and Szarejko, 2013).
In Arabidopsis, transcription factors MYB94 and MYB96 function together in the activation of cuticular wax biosynthesis under drought stress (Lee et al., 2016). MYB96 53 may also be involved in the response to drought stress in Arabidopsis through ABA signalling that mediates stomatal closure via the RD22 pathway (Seo et al., 2011).
Both MYB94 and MYB96 were induced by as much as 3-fold in rPPC1-B compared to wild type ( Figure S5T and S5U). Specifically, MYB94 transcript levels rose 8 h earlier than they did in the wild type and were already close to peak levels by the light-to-dark transition, whereas wild type MYB94 levels peaked sharply 8 h into the dark period ( Figure S5T). MYB96 transcript levels were induced relative to the wild type at both dawn and dusk, but the peak of MYB96 at dawn represented the largest fold-change relative to the wild type ( Figure S5U).  , Impact of drought on K. laxiflora wild type and transgenic lines with reduced PPC1. Gas exchange profile for shoots of whole young plants (9-leaf-pair stage) measured throughout 27-days. The plants were watered to full capacity on day 0 and allowed to progress into drought until day 23, when they were re-watered to full capacity (black arrow). The mean of 4 individuals is presented for each line. (E) and (F) Impact of silencing PPC1 on chlorophyll a (E) and chlorophyll b (F) content in leaf pairs 2 through 7. In (E) and (           Mature leaves (leaf pair 6), which performed CAM in the wild type, were sampled every 4 h under constant conditions (100 µmol m -2 s -1 at 15 ˚C) for wild type and rPPC1-B. RNA was isolated and used for real-time RT-qPCR. A thioesterase/thiol ester dehydrase-isomerase superfamily gene (TEDI) was amplified as a reference gene from the same cDNAs. Gene transcript abundance data represents the mean of 3 technical replicates for biological triplicates and was normalized to loading control gene (TEDI); error bars represent the standard error. In all cases, plants were entrained under 12-h-light/ 12-h-dark cycles prior to release into LL free-running conditions. Black data are for the wild type and red data rPPC1-B.     Plants were entrained under 12-h-light/ 12-h-dark for 7 days prior to sampling. Epidermal peel samples were separated from leaf pairs 6, 7 and 8, with samples collected every 4h starting at 02:00, 2 h into the 12-h-light period. Each biological sample represents a pool of epidermal peels taken from 6 leaf pair 6 leaves from 3 clonal stems of each line. Each peel was frozen in liquid nitrogen immediately after it was taken and pooled later. RNA was isolated and was used in RT-qPCR. A thioesterase/thiol ester dehydrase-isomerase superfamily gene (TEDI) was amplified as a reference gene from the same cDNAs. Gene transcript abundance data represent the mean of 3 technical replicates for biological triplicates, and was normalized to the reference gene (TEDI); error bars represent the standard error. In all cases, plants were entrained under 12-h-light/ 12-h-dark cycles prior to release into LL free-running conditions. Black data are for the wild type       (1,7,13,22 and 27) throughout the 27-day drought-stress and re-watering experiment. Results expressed as percentages were calculated separately for the 12-h-light and 12-hdark periods by determining the percentage of the total diel/ 24 h CO 2 uptake that occurred in that period. The actual CO 2 uptake over the 12-h-light and 12-h-dark period is also presented for each line.  (A), shoot fresh weight, (B), root fresh weight, (C), shoot dry weight, and (D), root dry weight. Fresh and dry weight of above ground biomass (shoot) and below-ground tissues (roots) at maturity following 138 d of growth under glassshouse conditions. Wild type (WT), rPPC1-A and rPPC1-B were measured under both well-watered and drought-stressed conditions. n = 6 plants; error bars represent standard errors. Black data are for the wild type, blue data rPPC1-A, and red data rPPC1-B. Asterisks indicate significant difference from the wild type based on Student's t-test: *Shoot FW p = 0.026; Shoot DW p= 0.026, n = 10 developmentally synchronised clonal plants of each line. Wild type plants with normal levels of PPC1 had visibly curled leaves after 23 d drought, whereas rPPC1-B leaves look less curled. Plants were removed from the gas exchange cuvettes after 23 d of drought, prior to re-watering. Plants were returned to the gas exchange cuvettes after re-watering for additional CO 2 uptake measurements to be collected. The gas exchange data for these plants is presented in Figure 4D.      Plants were entrained under 12-h-light/ 12-h-dark for 7 days prior to sampling. Epidermal peel samples were separated from leaf pairs 6, 7 and 8, with samples collected every 4h starting at 02:00, 2 h into the 12-h-light period. Each biological sample represents a pool of epidermal peels taken from 6 leaf pair 6 leaves from 3 clonal stems of each line. Each peel was frozen in liquid nitrogen immediately after it was taken and pooled later. RNA was isolated and was used in RT-qPCR. A thioesterase/thiol ester dehydrase-isomerase superfamily gene (TEDI) was amplified as a reference gene from the same cDNAs. Gene transcript abundance data represent the mean of 3 technical replicates for biological triplicates, and was normalized to the reference gene (TEDI); error bars represent the standard error. In all cases, plants were entrained under 12-h-light/ 12-h-dark cycles prior to release into LL free-running conditions. Black data are for the wild type and red data rPPC1-B.