Experimental set-up and data extraction

TCS input and output activity can be monitored by using a promoter-reporter system that is tuned to the TCS. This means that one can study phosphocompetition in vivo by monitoring the gene transcription of cytokinin-responsive genes. In this case, a reporter construct encoding the non-coding, upstream intergenic region of ARR5 was cloned upstream of firefly Luciferase LUCm3. We previously showed that overexpression of the C-type ARR22 prevented the activation of our LUC reporter [49] consistent with that of transgenic studies [26]. We have expanded that methodology by deriving four metrics to capture the amount of cytokinin dependent and independent activity. We continually monitored the transcriptional activity of ARR5p::LUCm³ in the same cells over the experimental time period by measuring the LUC activity in Relative Light Units (RLU) and calculated the Area under the Curve (AUC) (Fig 1). The AUC takes into account the time between measurements and gives numerical values for testing differences via statistical procedures. After transfection, the cells were incubated for six hours before adding the fLUC substrate D-luciferin. Thereafter the light emission was monitored from living cells in a microplate reader every five minutes for one hour. This captured both basal overexpression effects and replicate efficacy before experimental treatments (see Fig 1: Basal Activity). Thereafter, a treatment was applied as mock (medium only) or cytokinin (to a final concentration 1 μM t-zeatin in medium) directly to the protoplasts and continually monitored every five minutes for ten more hours. Instead of taking a peak expression approach (see [49]), the total transcriptional activity over the measurement interval includes all kinetic and subtle differences that are present in the dataset, called the Total Response Level (Fig 1: Total Response Level). From this, treatment specific effects due only to cytokinin are derived by subtracting the mock AUC from the treatment derived AUC and is termed Treatment Specific Activity (Fig 1: Treatment Specific Activity), giving a more exact amount activity ascribable to treatment effects only, here the cytokinin t-zeatin. For example from Fig 2B, if we took the ratio of only ARR2-overepxression (ARR2-ox) to that of the promoter alone at the Total Response Level, we obtain a circa 3-fold induction level (~150/~50 = ~3). Yet, the quantification of the Treatment Specific Activity reveals that actually approximately 30-fold more cytokinin-specific activity (~90/~3 = ~30) was produced over that of the promoter alone, whereby the cytokinin Treatment Specific Activity of the promoter alone was set to 1 (Fig 2C). To simplify these types of comparisons, the experimental graphs retain the raw AUC scale on the right-side axis as well as the same values referenced to the promoter alone mock control on the left-side axis. In general transfection variance could be adequately controlled allowing representative experiments to be shown (see Methods). Where overall transfection levels still varied, but the biological readout was similar, the data could be normalized using the method of quantile normalization (QN). Where doubt was still present in transfection variance, pCF203 encoding GREEN FLUORESCENT PROTEIN (GFP) was co-transfected and the GFP intensity was quantified and the RLU emission rescaled according to the GFP emission level. Where needed, these techniques were combined. Comparative and exploratory data are described in the Methods.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Graphical summary of the parameters extracted from the light emission curves.

Transcriptional activity of a promoter::fLUC construct was quantified in a plate reader every 5 minutes as light emission from oxidative decarboxylation of D-luciferin by firefly luciferase (fLUC). The light emission is represented in Relative Light Units (RLU). The total luciferase activity was estimated by calculating the Area under the Curve (AUC). The AUC from the first hour, captures the Basal Activity, which is promoter-reporter expression before stimulation. The Basal Activity values enable controlling transfection and technical variance as well as assessing any treatment-independent, effector specific effects. After one hour, hormone or only medium was added to each respective half and LUC emission monitored for 10 more hours. The Mock and Treated sample Total Response Levels correspond to the AUC excluding Basal Activity. After blocking and ranking, the AUC of Treated minus Mock total response levels was calculated in a pair-wise manner deriving the Treatment-Specific Activity. The dark contour lines represent the mean of 4 to 6 replicates, red being mock and blue being t-zeatin treated. See the Methods for more details.

https://doi.org/10.1371/journal.pone.0212056.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Reporter gene activity directly reflects phosphorelay action in vivo.

(A) in vivo light emission curves obtained over 11 hours. Data is composed of four independent experiments, each containing 4 replicates per sample type. The experiments were combined after quantile normalization (QN, see Methods, S26 Fig). One replicate of this dataset with only ARR22 variants using a different metric was published in [49]. (B) AUC calculated after excluding the first hour, that is, beginning after treatment. (C) After ranking (see Methods), the AUC_cyt−AUC_mock was calculated for all four sets, which gives the area corresponding to the space above the mock curves and bordered by the cytokinin treated curves. Significance classes were determined using Fischer’s Least Significant Difference test, α = 0.01. Letters not connected by a letter are significantly different. 2 μg plasmid was used for each effector in 30 μL transfections at 6.6x10⁶ protoplasts mL^-1.

https://doi.org/10.1371/journal.pone.0212056.g002

ARR22 can block the transactivation and cytokinin-dependent activation of ARR2

In our previous report, overexpression of ARR22^WT hampered the cytokinin response of ARR5p::LUCm3 and this was dependent on the ability of ARR22^WT to accept TCS-phosphate [49]. In cytokinin treatments of whole plants or leaf material, ARR5 has a peak expression level 20–40 minutes after cytokinin treatment [31], whereas here, the firefly LUC reporter gene ARR5p::mLUCm³ had a peak emission level shift of about 1 hour after cytokinin treatment, which we presume reflects the lag in protein synthesis (Fig 2A). Conceptually, ARR22^WT is proposed to effectively dephosphorylate AHPs before B-types have a chance to obtain TCS~P from them [26, 45, 46, 49], therefore we asked the question what happens when we overexpress a B-type ARR, like ARR2, simultaneously along with ARR22^WT. To do this, ARR22^WT versions with mutations of its conserved phosphorylatable Asp74 to Ala, Asn or Glu (ARR22^D74N,A,E) were transfected alone or along with ARR2 (Fig 2). As can be seen in Fig 2, two immediate observations can be made: one, overexpression of ARR2 resulted in a major increase in the cytokinin responsiveness of the reporter and two, ARR22^WT was able to not only block the endogenous B-types but also overexpressed ARR2 (Fig 2A: ARR2 with ARR22^WT compared to ARR2 alone), with both proteins stable under these conditions (S1 Fig). Consistent with previous results [26, 45, 46, 49], this effect of ARR22^WT was dependent on the conserved Asp of the REC domain as mutation to any other residue prevented the dominant-negative phosphocompetition effects (Fig 2A). The Total Response Level (Fig 2B) thus follows and supports the curve data (Fig 2A) providing the desired statistical measure. Even though cytokinin-dependent activity was present and measureable (Fig 2A and 2C), quantification of the exogenous effect of cytokinin on the promoter alone at this stringent level did not show a mathematically significant difference (Fig 2B: promoter alone vs. promoter alone + t-zeatin) as the data are not noise free. We felt it was important to leave the biological and experimental noise in our results and therefore a stringent statistical quantification cutoff α = 0.01 was chosen in order to focus on the major effects, considering that the reporter clearly and reproducibly responded to cytokinin and dramatically, in a cytokinin-dependent manner with additional B-type ARR2 (see Figs 2–7).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Transactivation potential of various mutations of the phosphorylated Asp in ARR2.

Panels are the same as in Fig 2 with the following changes: Data is shown for one experiment, each containing 4 replicates per sample type. RLU is relative to the GFP emission level (see Methods, S27 Fig). 2 μg plasmid was used for each effector in 60 μL transfections at 3.5x10⁶ protoplasts mL^-1.

https://doi.org/10.1371/journal.pone.0212056.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Overexpression challenge of AHP5 against ARR22.

Panels are the same as in Fig 2 with the following changes: Data is shown for one experiment, each containing 4 replicates per sample type. 2 μg plasmid was used for each effector in 60 μL transfections at 3.5x10⁶ protoplasts mL^-1.

https://doi.org/10.1371/journal.pone.0212056.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Overexpression comparison of AHPs, A-types, and ARR22 against ARR2 on ARR5p.

Panels are the same as in Fig 2 with the following changes: Data is shown for one experiment, each containing 4 replicates per sample type. 2 μg plasmid was used for each effector in 30 μL transfections at 6.6x10⁶ protoplasts mL^-1.

https://doi.org/10.1371/journal.pone.0212056.g005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Action of ARR22 is upstream of B-types.

Panels are the same as in Fig 2 with the following changes: Data is composed of four independent experiments, each containing 4 replicates per sample type. RLU is relative to the GFP emission. The experiments were combined after quantile normalization (QN, see Methods). 2 μg was used for each effector in 60 μL transfections at 3.5x10⁶ protoplasts mL^-1.

https://doi.org/10.1371/journal.pone.0212056.g006

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Transactivation of ARR2 on WUS.

Panels are the same as in Fig 2 with the following changes: Data is composed of one experiment, containing 6 replicates per sample type. 2 μg was used for each effector in 30 μL transfections at 6.6x10⁶ protoplasts mL^-1.

https://doi.org/10.1371/journal.pone.0212056.g007

Quantification of the curve data by the AUC metrics (Fig 2B and 2C, S2 Fig) supports the visual interpretation of Fig 2A. ARR2-ox has repeatedly been reported to be transactive on its own, although the fold-levels vary from one experimental set up to another [33, 37, 40]. We typically observed that ARR2-ox was transactive with ~2-fold higher level already at the Basal Level (S2 Fig). This was also preserved at the Total Treatment Response in the mock treated samples (Fig 2B: ARR2-ox alone, mock). However, we observe that ARR2 is principally TCS dependent as it was more transactive when the TCS was activated by cytokinin, in this case ~30-fold over the promoter alone (Fig 2C). Both mock and cytokinin induction effects of ARR2-ox were blocked by ARR22^WT-ox (Fig 2B and 2C) whereas ARR22^D74N,A,E variants led to the same or only slightly higher Treatment Specific Activity (Fig 2C). This indicates that ARR22 does not negatively regulate the TCS independently of its phosphohistidine phosphatase function, consistent with transgenic studies over- and mis-expressing ARR22 [26, 45, 46]. Additionally, differences between ARR22^WT and its variants were not due to basal effects (S2 Fig). Interestingly ARR2 basal transactivity also might be TCS dependent as co-expression with ARR22^WT reduced ARR2-ox mock levels to that of the promoter alone (Fig 2B). Thus from our data, the TCS dependency of ARR2 is consistent with those studies which did not perturb development or response to cytokinin when ARR2 was overexpressed [37, 40] and is in contrast to other studies which did [33, 43, 55]. Furthermore, the high cytokinin response observed when ARR2 was overexpressed indicates that the endogenous B-type levels must be limiting and that there is more AHP~P likely present than endogenous B-types to receive it. Likewise, overexpression of ARR22^WT can outcompete an overexpressed B-type ARR2 for TCS~P, as we observed a reduction of ~30-fold (ARR2 alone) to ~10-fold (ARR2 + ARR22^WT) treatment specific activity. Thus ARR22^WT-ox likely hampers ARR2-ox from obtaining TCS~P from AHPs due to its phosphohistidine phosphatase effect which is lost when the conserved phosphorylatable Asp is mutated (this work, [26, 45, 46, 49]). However, in order to make this statement, one has to confirm that ARR2 activation is TCS dependent. Our data indicates that although ARR2 is transactive as previously observed [30, 33], it is clearly much more dependent on the TCS for its full-activation in a cytokinin-dependent manner. Nevertheless, Hwang & Sheen, 2001 proposed that ARR2 was not (strictly) TCS dependent as a mutation in the phosphorelay-Asp residue, ARR2^D80N, only led to a moderately reduced and not inert ARR2 variant. Transgenic studies using ARR2^D80N however suggested the opposite, as ARR2^D80N plants did not lead to a cytokinin-dependent delay of leaf senescence compared to the overexpression of the wild-type protein [40]. Therefore based on the previous results and the current understanding of RR function, we set-up a phosphocompetition experiment of ARR22 with ARR2^D80N to explore if ARR2 is TCS dependent or not.

Transactivity of the non-phosphorylatable ARR2 variant ARR2^D80N is TCS dependent

The B-type REC domain exerts a negative regulation on the DNA binding and transactivation ability [30]. This negative regulatory function is modulated by the phosphorylation status of a conserved Asp (D) in the catalytic acidic pocket of the REC domain ([3, 21, 56], S3 Fig). D->E substitutions in ARR2 result in a constitutive active protein [37] whereby D->N leads to a defective, but not null-protein [33, 40]. Neither ARR2^D80E nor ARR2^D80N is phosphorylated in a TCS dependent manner [37, 40], demonstrating that Asp80 is the critical Asp for ARR2. These two mutations are routinely used to demonstrate the TCS dependency of ARR2 [33, 37, 40, 55] and therefore the corresponding point mutations were also introduced into ARR2. These cDNA variants of ARR2 were transfected along with ARR5p::LUCm3 and challenged by co-expression with ARR22 or its inactive variant ARR22^D74E (see Fig 3).

ARR2-ox led to a much stronger reporter activity in the LUC emission curves (Fig 3A) and when quantified at the total levels (Fig 3B) is consistent with that shown in Fig 2. In comparison, this dataset produced a 20-fold more cytokinin specific activity over that of the promoter alone (Fig 3C) and this was counteracted by ARR22^WT co-overexpression but not ARR22^D74E (Fig 3). Before treatment, ARR2^D80E-ox was ~10x over mock (S4 Fig), demonstrating that the D->E substitution releases REC inhibition on transactivation capability, congruent with previous observations [33, 37, 40, 57]. This high level of expression was maintained during mock and cytokinin treatments (Fig 3B) and was principally independent of cytokinin, as ARR2^D80E only produced about 5-fold more cytokinin-specific activity when expressed alone (Fig 3C). Thus, overexpression of ARR2^D80E resulted in high LUC activity which nearly maximized during the basal measurement, and then stayed high, decaying slowly over time (Fig 3A). To investigate why the ARR2^D80E light emission curves continually decayed over time, the light emission pattern of ARR2^D80E was compared to that of a known constitutive promoter UBI4/2 [58]. As shown in S5 Fig, the expression profile of ARR2^D80E is similar to that of UBI4/2::LUCm3 and indicative of constitutive (high-transcriptional rate) promoter activity. The variant ARR2^D80E theoretically should be able to bypass ARR22 as ARR2^D80E does not need to obtain TCS~P from the AHP~P pool for its activity. This was confirmed since neither overexpression of ARR22^WT, ARR22^D74E nor ARR22^D74A were able to prevent the 10-fold transcriptional output of ARR2^D80E (Fig 3B, S6 Fig). The cytokinin-dependent activity of ARR2^D80E when overexpressed alone is likely due to the endogenous B-types present in the cell. We conclude that this cytokinin-dependent activity can be occasionally observed for ARR2^D80E-ox from time to time but varies (Fig 3, S5–S7 Figs). More importantly, ARR2^D80E led to high constitutive activity in all conditions confirming that ARR2^D80E is able to bypass the TCS activation requirement.

In contrast to observations in previous studies on the ARR6 promoter in mesophyll protoplasts [33, 40], in our assays ARR2^D80N-ox generally showed comparable levels to that of the promoter alone at the basal and mock levels (S4 Fig, Fig 3B). All of the ARR2 REC mutants were detectable by immunoblot analysis (S8 Fig). In the presence of cytokinin ARR2^D80N-ox resulted in a significant enhancement of cytokinin response that was weaker and dampened quicker than that of ARR2 (Fig 3A). This is analogous with previous reports where the D->N substitution led to reduced protein activity seen in transactivity or cytokinin-dependent inductive responses [33, 40, 57]. ARR2^D80N overexpression led to circa 9x more cytokinin specific-activity over the promoter alone, which corresponds to almost 50% of the ARR2^WT protein (Fig 3C). Unlike ARR2^D80E however, the cytokinin dependency was inhibited by ARR22^WT (Fig 3) but not variant ARR22^D74E, which was completely inert, even allowing the same specific-activity as that of ARR2^D80N alone (Fig 3C).

There are three conserved negatively charged amino acids of the REC domain that comprise the catalytic acidic pocket, with two of those being absolutely conserved Asp residues ([56], S3 Fig). Mutation of the two absolutely conserved Asp to a non-phosphorylatable residue prevents aspartyl-phosphate formation, whereas the third residue of the acidic pocket must be an Asp or Glu [30, 56]. That overexpression of ARR22 blocks the cytokinin induction of ARR2^D80N (Fig 3) suggests that either ARR2^D80N is still phosphorylatable or that it might interact with a component targeted by ARR22 needed for ARR2 to be transcriptionally active. In that regard, it is interesting that another B-type variant ARR18^D70N is also weakly transactive [57]. Based on work in bacteria, disturbance of the catalytic pocket does not always result in a completely dysfunctional response regulator protein [59, 60]. It is known that mutation of the Asp phosphorylation site can also led to aberrant phosphorylation and not always a null mutant, even if Asn or Glu are not phosphorylated [19, 56, 59]. Furthermore, constitutive active versions of the bacterial RRs CheY mutation E9Q and PhoB D13K were not mutations of the actual phosphorylated Asp [59, 60]. The acidic pocket of ARR2 is conserved with CheY and PhoB (S3 Fig) and such complex effects involving ARR2^D80N could also be occurring. Either way, these comparisons of ARR2 D80 variants clearly indicate that ARR2 is activated in a TCS-dependent manner. And, as suggested before, that ARR22 functions within the TCS in a dominant-negative fashion preventing ARR2 TCS-dependent activation. Taken together we can establish that ARR2 and ARR22 work in a TCS-dependent manner with respect to ARR5p::LUCm3 activation (Figs 2 and 3).

Overexpression of AHP5 cannot outcompete ARR22 overexpression

In the previous experiments we established that ARR22^WT-ox is likely able to deplete AHP~Ps even when ARR2 is overexpressed, reducing the output partially (Fig 2C) or completely (Fig 3C). We postulated that if ARR22^WT is controlling the rate-limiting access to AHP~Ps, perhaps overexpression of an AHP would be able to counterbalance ARR22^WT effects. AHP1, 2, 3, and 5 are found in the cytoplasm and the nucleus (AHP4 was reported only cytoplasmic [61]) and shuttle freely between the two compartments irrespective of their TCS phosphorylation state [24]. ARR22 is also located in cytoplasm and nucleus [26, 46] where it was also observed to interact with AHP2, 3, and 5 [46]. As mentioned, ARR22 is able to remove ~P from AHP5 in vitro [26] and likely in vivo [26, 45, 46, 49]. Therefore, a competition experiment was designed combining AHP5 with ARR22 or ARR22^D74E and ARR2 (Fig 4).

As shown before, ARR2-ox led to a large boost to cytokinin responsiveness of the reporter (Fig 4A) quantified here to about 7x more specific-activity (Fig 4C) fitting with the slightly lower overall basal levels compared to the previous experiments (S2, S4 and S9 Figs). Nevertheless the biological responses are consistent with the experiments shown before (Figs 2 and 3). Also as already described, ARR22^WT prevented ARR2 cytokinin-responsiveness and ARR22^D74E did not (Fig 4). Overexpression of AHP5 fostered a bit more reporter output at the endogenous B-type levels (Fig 4A) even being significantly different at the total response level due to an overall lower output at the mock level (Fig 4B) but not at the treatment specific-activity level, even though it was slightly higher than the promoter alone (Fig 4B and 4C). Since AHP5 was stably expressed (S10 Fig), this indicates that AHP5 is generally not limiting with respect to the cytokinin-dependent promoter activity in contrast to that observed by the overexpression of B-type ARR2. Thus, endogenous levels of B-types must be much less compared to the availability of AHP~Ps during cytokinin signaling. This positive effect of AHP5 under cytokinin treatment was preserved in the presence of ARR2 in these experiments (Fig 4B); here, co-overexpression of AHP5 and ARR2 significantly increased the cytokinin specific-activity (Fig 4C). Nevertheless, overexpression of ARR22^WT prevented ARR2 cytokinin-responsiveness at both the endogenous B-type level and when ARR2 was also co-overexpressed with AHP5, reducing the cytokinin specific-activity to that of the promoter alone (Fig 4C). Thus, although AHP5 had a slight positive effect on the reporter activity, its overexpression was not able to compensate for the dominant-negative phosphohistidine phosphatase activity of ARR22^WT. As observed in the previous experiments, ARR22^D74E did not alter the overexpression effects of AHP5, ARR2 or their co-expression (Fig 4) reaffirming that ARR22 functions strictly on TCS signaling.

Overexpression of neither A-types nor AHPs have the same rate-limiting effect as ARR22 on ectopic B-type levels

Previous studies showed that overexpression of AHP2 or AHP5 did not affect an ARR6::LUC reporter in mesophyll protoplasts irrespective of cytokinin [33] even though ectopic expression of AHP2 in plants resulted in cytokinin hypersensitivity [62]. Therefore a comparative phosphocompetition experiment was conducted in order to ascertain the effects of AHP2-ox in our biological system. Based on what is known about the TCS, A-types negatively regulate the TCS and their genes are induced by B-types as a negative-feedback mechanism [20, 53, 63]. This negative feedback is partially dependent on their ability to pick up TCS~P from AHPs as mutation of their conserved phosphorylatable Asp to another residue abolishes phosphorelay function in vitro [19] and in vivo [19, 20]. Therefore we asked if A-type overexpression could prevent ARR2 TCS-dependent activation as that observed for ARR22.

Inspection of Fig 5A reveals comparable results to those previously described for ARR2-ox and ARR22^WT-ox in this dataset. ARR2-ox led to about 2-fold more activity at the basal (S11 Fig) and mock levels (Fig 5B) whereby cytokinin treatment ushered in ~7.5x more specific activity (Fig 5C). Likewise, ARR22^WT prevented cytokinin responses at endogenous B-type and ARR2-ox levels (Fig 5A–5C). At the endogenous B-type levels, overexpression of AHP5 led to a visible increase in reporter activity (Fig 5A) of which significant differences could be ascertained at the total response level (Fig 5B) but not at the treatment specific level (Fig 5C). This is similar to that observed in the previous experiment (Fig 4). In contrast, in conjunction with ARR2, AHP5-ox and AHP2-ox lowered the basal and mock expression effects of ARR2-ox in Fig 5B. This effect was not observed in Fig 4, likely due to the lower overall basal expression in Fig 4 since both proteins were stably expressed (S10 Fig). More importantly, neither AHP2 nor AHP5 affected ARR2 specific activity (Fig 5C), indicating that AHP2 is not limiting and that AHP5 can cause cytokinin hypersensitivity under the right conditions (Fig 4), yet even when this effect is minor compared to overexpressing ARR2 with respect to gene activation.

A-type dominant-negative effects on the TCS due to overexpression are typically only observed when exogenous cytokinin is applied, and thus, without applying cytokinin appear like wild-type both phenotypically and at the molecular level [19, 20, 63]. By this accord, overexpression of A-types is predicted to block or dampen the reporter gene ARR5p::LUCm3 in a cytokinin dependent manner. While ARR7-ox lowered the promoter activity non-significantly at the basal and specific activity levels (S11 Fig, Fig 5C), it clearly prevented cytokinin induction at the endogenous levels similar to ARR22 (Fig 5A: ARR7-ox alone). Quantification of the AUC values confirmed a loss of induction at the total response level (Fig 5B) and the treatment specific response (Fig 5C), significantly for the former. Inspection at the basal levels (S11 Fig) suggests that without cytokinin, ARR7 was able to inhibit ARR2-ox to about 50% of its transactivity. Unlike ARR22^WT-ox however, under cytokinin treatment, overexpression of ARR7 with ARR2 was not able to prevent ARR2 from obtaining TCS~P and in this case resulted in surprisingly more treatment-specific activity (Fig 5C). This prompted testing two other A-types: ARR4 and ARR15. In addition, the conserved phosphorylatable Asp95 of ARR4 was mutated to ARR4^D95E or ARR4^D95N (S12 Fig).

As observed for ARR7, when expressed alone overexpression of ARR4 and ARR15 also prevented cytokinin induction of ARR5p::LUCm3 (Fig 5, S12 Fig). Still, while ARR4, ARR7 and ARR15 could block the endogenous driven reporter activity, they were not able to prevent ARR2-ox effects (S12 Fig) even though all proteins could be confirmed by immunoblotting (S13 Fig). The positive-induction response of ARR7-ox alongside ARR2-ox (Fig 5) was not reproducible (S12 Fig) and in general there was more variability in the A-types samples co-transfected with ARR2 such that negative, positive and neutral effects were observed (S12 Fig). A-types are regulated in very complex manners [3] such that it is possible that minor environmental differences compounded output effects when ARR2 was co-overexpressed. The mutations ARR4^D95E and ARR4^D95N abolished the ability of ARR4 to inhibit the endogenous cytokinin induction, yet had no dominant negative effect on ARR2-ox (S12 Fig). Remarkably, the reporter activity for ARR4^D95N and ARR4^D95E overexpression is very similar to that observed for ARR7^D85N transgenic plants [19], whereby the D mutations led to a reduced, but steady and sustained induction level over the measurement timespan compared to the promoter alone (S12 Fig). This demonstrated that our experimental conditions match results obtained with A-type ectopic expression in plants [19, 20] previously thought not possible. One interpretation is that A-types can hold certain concentration of B-types at a limiting equilibrium, and also implies that some B-types are still able to obtain TCS~P even when A-types are present. When cytokinin is given exogenously and drives the TCS in the forward direction, an excess of A-types can prevent the excess TCS signal from reaching endogenous level B-types, seen here as a loss of cytokinin-dependent reporter activity. In conclusion, none of the A-types were able to prevent significant amounts of TCS~P from reaching overexpressed ARR2 like that observed for ARR22^WT-ox.

Various B-types are equally perturbed by ARR22 overexpression

ARR22 therefore functions upstream of ARR2 at the AHP/A-type phosphorelay level (Figs 4 and 5, S12 Fig) and its dominant-negative effect could not be compensated for by AHP5-ox or AHP2-ox (Figs 4 and 5). Furthermore, A-types were only able to block TCS information flow dictated by the B-type concentration levels (Fig 5) since we observed that the ARR2 level was principally the rate-limiting step in gene activation and not the concentration of the AHPs or A-types (Figs 2–5). Taken together, ARR22 is therefore most likely preventing B-types from receiving the phosphorelay signal via AHPs at a rate significantly faster than that of A-types. If this is so, ARR22-ox should generally knock out the TCS and also prevent other B-types in a TCS-dependent manner. Alternatively, the effect of ARR22 could be directed somehow only to specific B-types, even indirectly through AHP specificity. Therefore three additional B1-subfamily B-types ARR1, ARR10 and ARR18 were cloned into to the same effector vector and challenged against ARR22 (Fig 6). All of these B-types have been shown to target ARR5 and/or ARR6 in various studies [33, 37, 51, 57, 64] and all can interact with at least 2 of the 3 AHPs known to interact with ARR22 (summarized in S14 Fig as a Cytoscape graph).

Even though basal levels (S15 Fig) were a bit lower, the experimental output is comparable to the previous experiments as judged by inspecting the promoter alone, ARR2, ARR22^WT and ARR2 & ARR22^WT (Fig 6). All of the B-types were stably expressed (S16 Fig). Similarly as described in previous experiments, ARR2 overexpression led to ~8x more specific activity (Fig 6C), matching the levels obtained in Figs 4 and 5. As seen in the total response levels, ARR2 led to ~8x more specific activity compared to ~5x for ARR1 (Fig 6A), reminiscent of ARR2 being more transactive and/or responsive to cytokinin than ARR1 in general [30, 33]. Thus, the higher transactivity of ARR2 and ARR1 was confirmed in this study at the total response mock level (Fig 6B) [30, 33]. ARR10-ox and ARR18-ox on the other hand did not perturb the mock levels much (Fig 6B). These results are consistent with a report that found ARR1 and ARR2 more transactive than ARR10 in mesophyll protoplasts [33]. ARR10 and ARR18 showed comparable specific activity as that of ARR1 (Fig 6C), which is analogous to the same study in mesophyll protoplasts, where ARR10 was shown to have low transactivity but high cytokinin responsivity [33]. Inspection of the literature showed that ARR18 led to more cytokinin-responsive activity than ARR2 on an ARR5p::GUS reporter assay [57] yet less than ARR1 when tagged as c-myc-ARR18, and yet greater than ARR1 when ARR18 was not epitope tagged [64]. Thus differences in protein output and stability therefore can be traced to epitope tagging of B-types [43, 55, 64, 65] and apparently, the position and the type of epitope tag (or lack thereof) might affect B-type signaling capacity. That said, in our experiments, all of these B-types were able to activate the ARR5p::LUCm3 reporter in a cytokinin-dependent manner (Fig 6) confirming the hypothesis that endogenous B-type levels are limiting.

Co-overexpression of ARR22^WT with ARR1, ARR2, ARR10 and ARR18 blocked their cytokinin-dependent induction effects (Fig 6). ARR22^WT typically did not absolutely eliminate all of the cytokinin-dependent activity when expressed alone in combination with the ARR2 or the other B-types (ARR18 being the exception) which could be explained as coming from the phosphorelay mediated by the AHPs AHP1 and AHP4 that are not targeted by ARR22, or just that the B-types are still able to find a phosphorylated AHP before ARR22 does. As ARR22^D74E (S17 Fig) did not exert the dominant-negative effect of ARR22^WT on these B-types, we can conclude that the phosphohistidine phosphatase activity of ARR22 is sufficiently rapid to severely dampen gene output that is dependent on the TCS. In conclusion, it appears that over/mis-expression of ARR22 in any environment is likely to lead to a severe and non-discriminatory reduction of B-type output.

Combined ARR2 and ARR22 overexpression can determine TCS dependency of difficult targets

The various mutant variants of ARR2 and ARR22 can be used to determine if a promoter is a target of the cytokinin-mediated TCS signaling pathway or not. Therefore a recently identified, notoriously difficult target, WUSCHEL (WUS) was tested with this method. B-types ARR1, ARR2, ARR10 and ARR12 have recently been shown to directly bind to WUS in vivo [12–14] fulfilling a long standing prediction [66] concerning the complex regulation of the meristem’s organizing center [1, 66–68]. Therefore, the intergenic promoter region (-1741 to +1) of WUS was cloned upstream of LUCm3 and co-transfected against the ARR22 and ARR2 variants: ARR2, ARR22, ARR2^D80N and ARR2^D80E (Fig 7). During these experiments, WUSp::LUCm3 typically required a slightly longer incubation of 8 hours to provide an informative response range compared to the six hours required for the ARR5p::LUCm3 promoter. In addition, the ARR5p::LUCm3 promoter alone was also transfected alongside the WUS set to provide a known-target reference; relative values are referenced to the WUSp::LUCm3 promoter alone.

As before, a weak induction of ARR5p::LUCm3 by cytokinin was observed (Fig 7A and 7B), indicating that the experimental conditions are comparable to those previously described. The WUSp::LUCm3 promoter occasionally showed very marginal, slightly higher expression when treated with cytokinin. It is known that WUS is only weakly induced by cytokinin [2, 13] and we could occasionally detect it here, about 1.25-fold (Fig 7B). ARR2-ox led to higher total response under mock and a boost in transcriptional activity upon the addition of cytokinin of WUSp::LUCm3 (Fig 7A and 7B) resulting in a two-fold increase in cytokinin specific activity for WUSp::LUCm3 (Fig 7C). Interestingly, the emission peak of ARR2-ox on WUSp::LUCm3 was reproducibly delayed with respect to that of ARR5p::LUCm3, here 300 min compared to ARR5p::LUCm3 at 175 min (Fig 7A and previous Figs). This suggests that the transcriptional complexes involving the two promoters are likely different. This was the first indication that careful inspection of the kinetic information may be exploitable in the future.

Moreover, ARR2-ox effects were nullified by the co-transfection of ARR22 implicating that a functional phosphorelay is responsible for ARR2 in activating WUSp::LUCm3. Likewise, overexpression of the constitutive active ARR2^D80E variant led to high WUSp::LUCm3 reporter activity at the basal (S18 Fig) and total response levels independent of cytokinin (Fig 7B). The disabled ARR2^D80N variant was essentially inert under all conditions (Fig 7). As observed before, the ARR2 variants were stably expressed and showed no obvious differences during these experiments either as judged by immunoblot analysis (S19 Fig). In conclusion, we observe that the WUSp::LUCm3 promoter responded similar to that of ARR5::LUCm3 to the ARR2 variants (Fig 3) and ARR22 (Figs 2 and 3) co-overexpression with the exception of the peak emission timing. Taken together, this provides confirmatory evidence that WUS responds to cytokinin signaling in a TCS-dependent manner and that this response is, at least partially, classically mediated through B-type RRs such as ARR2, consistent with the recent reports of B-types affecting WUS expression [12–14].

FACS-RT-qPCR corroboration analysis on the endogenous genes ARR5, ARR7, and WUS

A FACS-RT-qPCR experiment was conducted to corroborate that the cloned promoters pARR5 and pWUS accurately captured the native regulatory output in our particular cell culture protoplasts. To do this, cells were transfected with effector-GFP fusions and GFP fluorescent cells were collected by Fluorescence activated cell sorting (FACS), which is a technique to isolate particular cells one at a time from a heterogeneous mixture based on their fluorescence profile. After collecting the GFP positive cells, RNA was reverse transcribed (RT) and quantified by quantitative-PCR (qPCR). qPCR primers were designed to monitor endogenous ARR2, ARR1, ARR5, ARR7, WUS, ACTIN 2 (ACT2) and ELONGATION FACTOR 2 (EF2). As mentioned, ARR5 and ARR7 are known B-type targets [51]; ARR1 and ARR2 are reported to be unaffected by cytokinin treatment [48]. EF2 served as the normalization reference control, with ACT2 serving as the error reference control [69]. qPCR primers for WUS were designed to capture the 5’ or 3’ end of the gene. Additionally, where possible, two different C-terminal GFP fusions to wild-type ARR2 and ARR2^D80E were constructed (see Methods). Free-GFP was transfected to test the nominal response of the cells to cytokinin. Unlike the transfection experiments, the FACS experiments were incubated overnight (~16hrs) before experimental begin due to machine-time access. On each day, independent transfection samples for each construct were pooled and then split, one composing mock and the other t-zeatin treated samples. Cytokinin treated cells were incubated for 30 min before beginning each sort, which took on average 20 min comprising the peak expression level for most A-type genes [31]. The qPCR data are presented in Cq values, median normalized to reference EF2, and therefore stay in base2 (Ct, Cq changes).

Although it was clear that the GFP fusions were expressed based on the FACS data, the localization of free-eGFP, ARR2-GFP and ARR2^D80E-GFP was also examined by conventional epifluorescence microscopy. For all ARR2 variant fusions, the majority of the GFP signal was observed in the nucleus although some very weak GFP emission was also observed in the cytoplasm for ARR2^WT (S20 Fig). Inspection of Fig 8 confirmed that neither ARR1, ARR2 nor ACT2 was affected by cytokinin treatment. Endogenous ARR5 and ARR7 were induced roughly 5 to 6 fold by cytokinin very similar to that observed in studies with intact plants [26, 36, 47]. ARR7 followed the same pattern as endogenous ARR5 except that a more dramatic difference was observed for ARR7 compared to ARR5, traceable to the overall much lower basal expression level of ARR7 (+9 Ct relative to EF2) compared to ARR5 (+3 Ct relative to EF2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Endogenous gene regulation of ARR5, ARR7 and WUS in response to ARR2 variants overexpression in cell culture protoplasts.

(A) Overexpression of eGFP, ARR2-eGFP, ARR2^D80E-eGFP were captured by FACS 16 hrs after transfection including cytokinin stimulus. (B) Ct(Cq) values of all samples were normalized to the median of EF2 with differences shown in Ct(Cq) with respect to EF2 (log₂). Smaller values indicate a higher transcript rate. ‘+ tZ’ indicates 1 μM t-zeatin; target gene for qPCR is indicated as the principle column header. The data presented here are composed of three, complete experimental replicates. 4 μg was used for each effector per 60 μL at 3.5x10⁶ protoplasts mL^-1; a total of eight 60 μL transfections was pooled and split for each experiment into mock and treated samples.

https://doi.org/10.1371/journal.pone.0212056.g008

Overexpression of ARR2-GFP led to transactivation of both ARR5 and ARR7 without cytokinin, either matching or exceeding their cytokinin-inducible levels (Fig 8). This is different compared to the LUC assays without exogenous cytokinin, as ARR2-ox either matched or exceeded the mock levels (see Basal and Total Response Levels from Figs 2–7) and led to 7 to 30 times more specific activity the presence of exogenous cytokinin (compare Fig 8 with Figs 2–6). ARR2^D80E-GFP-ox also drove ARR5 and ARR7 to levels either matching or exceeding their cytokinin-inducible levels (Fig 8) whereby ARR2^D80E-GFP led to slightly higher levels than ARR2^WT-GFP (Fig 8) consistent with its constitutive active confirmation (Fig 3, S6 Fig). Thus the high level of induction for endogenous ARR5 and ARR7 due to ARR2^WT-GFP-ox appears to have left very little room for further induction by cytokinin, resulting in a modest boost of only 2 Ct more activity with t-zeatin (Fig 8). Similarly, ARR2^D80E-ox led to a slightly higher expression than ARR2^WT-GFP and a smaller (1 Ct) positive effect in the presence of cytokinin (Fig 8), similar to that observed during the LUC assays (Fig 3, S6 Fig). Thus, one can conclude that the cloned promoter region of ARR5 behaved analogously to the native gene in the same system, with the caveat that without cytokinin, overexpression of wild-type ARR2 drove endogenous ARR5 and ARR7 to higher levels compared to the LUC assays, resulting in less cytokinin inducibility.

WUS on the other hand gave a slightly different answer. Endogenous WUS was actually not expressed in the GFP and ARR2-GFP overexpression samples, with or without cytokinin: the Ct was ≥ 36 Ct cycles, where the water control also started to give noisy results and many technical replicates produced no value at all (S1 Table). Thus, the normalized log expression level of ≥ +10(log₂) was more indicative of “absence of transcript”. Consequently, endogenous WUS was not induced by ARR2 overexpression with or without cytokinin (Fig 8). With hindsight, this is no surprise as the WUS chromatin environment keeps the gene transcriptionally silent in cell culture [68] unless placed on Shoot-Induction-Medium [12–14, 70, 71]. ARR2^D80E-GFP-ox was able to activate endogenous WUS by bringing the cycle numbers up from the non-noise range (≥ 36 Ct) to clearly present (to 33 Ct) (S1 Table) which corresponds to ~ +8(log₂), an increase of 2(log₂). As WUS was only transcribed in the presences of constitutive-active ARR2^D80E we conclude that it is highly likely that this constitutive-active version of ARR2 was able to eventually open the chromatin around the endogenous WUS enabling some transcription to take place.

Altogether, the FACS-RT-qPCR and the LUC reporter assay indicate that the cloned promoter regions and ARR2 overexpression reflect the protein networks that are natively present in the cell culture protoplasts. In comparison between the two approaches, the high dynamic response gained by the reporter assay suggests that it gives us a better ability to track the TCS phosphorelay activity compared to the endogenous readout. Overnight LUC assays lost their dynamic response when compared to shorter incubation times (S21 Fig), but performing a FACS-RT-qPCR 6 hours after transfection yielded data with identical results to the 16 hour incubations (S29 Fig). Differences between the endogenous ARR5 and the reporter could be due to differences in the vector packing compared to the native chromatin, changes in the three-dimensional structure of the vector, unknown regulatory regions 3’ of the 5’ intergenic space. Alternatively, only a few copies of the endogenous ARR5 promoter region are present per cell, whereas an unsaid number of ARR5 promoter molecules are present in the transfection experiment, quite possibly diluting out the endogenous B-types over the number of promoter targets. The endogenous ARR5 responded both to cytokinin treatment, and overexpression of ARR2 or ARR2^D80E corroborating the LUC assay generated data. That the endogenous WUS was weakly activated by ARR2^D80E suggests that the constitutive active form of ARR2 was eventually able to open the chromatin allowing transcription of the WUS gene. This might explain some of the strange, inexplicable phenotypic effects previously observed in transgenic plants expressing ARR2^D80E [37]. Likewise, ARR2 was able to activate the cloned reporter WUSp::LUCm3 in an open chromatin context in a cytokinin dependent manner (Fig 7) demonstrating the utility and sensitivity of the LUC assay for identifying TCS targets.

Transfection setup and LUC assay

Protoplasts and transfections were performed as previously described from an A. thaliana dark-grown cell culture cultivated at 23 ± 1°C [49]. Assays were transfected and treated simultaneously in a 96-well format using the 96-well manual pipette ‘Liquidator⁹⁶’ (Steinbrenner) with the exception that the DNA was filled to 10 μL instead of 9 μL with ddH₂O. Transfections and plate measurements were performed at 25 ± 1°C. Plasmid DNA was purified using Xtra-midi preparation kits (Macherey-Nagel) and kept at 4°C during storage. For every transfection, the total DNA load was kept absolutely the same to minimize competitive dilution effects. Six hours after transfection, the luciferase assay was started. 150 μL of each transfection was pipetted into two wells of a white ELISA plate. 15 μL of K3 with D-luciferin (K+ salt) was added to yield a final concentration of 200 μM using the Liquidator⁹⁶ pipette. Luciferase activity was monitored in vivo with plate reader FluoroSkan Ascent FL (0.5 s integration, 5 min intervals for 600 min, PMT 812). Basal expression levels were measured for 60 min and then t-zeatin (in K3 Buffer) to a final concentration of 1 μM, or K3 Buffer alone, was added (also with the Liquidator⁹⁶) to the treated or untreated samples, respectively. Linearity of the LUC emission was tested by transfecting UBI4/2::LUCm3 and performing serial dilutions (S22 Fig).

The experiments were conducted 6 hours after transfection as this yielded the best signal/noise ratio. Waiting overnight resulted higher basal levels, but severely limited cytokinin responsiveness (S21 Fig) presumably due to accumulated protein that masked the inductive response. Saturation for the promoter alone was tested by varying the amount of reporter plasmid and the amount of D-luciferin, whereby here 3 μg ARR5p::LUCm3 and 200 μM was saturating (S21 Fig). The reporter/effector relative concentration was tested to find a stable range of reporter activity. Transfection controls demonstrated that the method is extremely sensitive and amounts down to 125 ng of effector were detectable. In addition, we found that 0.5 to 2.0 μg ARR2 to 3 μg ARR5p::LUCm3 showed stable and consistent treatment specific activity (S23 Fig). Either 1 μg or 2 μg effector DNA was used as indicated in each respective figure legend.

PEG-mediated transfection assays are known to carry the highest variance between batches [77]. To account for this, all comparative components were restricted to the same batch of cells and transfected at one sitting and evaluated from a single 96-well plate in a maximum of 12 blocks for reporter and effector combinations, including plus/minus cytokinin. In addition, at least two biological replicates were performed and from each biological replicate two technical replicates were made, yielding a minimum of four measurements (examples in S24 Fig). Protoplasts were randomly distributed to wells to ensure the random sampling hypothesis before the actual transfection event. Stable transfection efficacies across a plate were confirmed by cytometry (S25 Fig).

It was observed that the total relative output was reduced when the reporter plasmid pool was kept constant when challenged with increasing amounts of another plasmid (S23 Fig). We theorize, but did not exhaustively investigate here, that all plasmid/DNA species are uptaken as one pool (the DNA pool) and not as independent events, as such it is expected that the more other species is added then it will reduce the relative amount the others are uptaken. When effector combinations were used, the total DNA load was calculated and leveled with the empty effector vector in those samples with fewer combinations. Therefore, the total DNA amount (in μg) was always maintained to ensure an equal uptake probability. Therefore, 3 μg of ARR5::LUCm3 was used with 2 μg or 1 μg for each effector plasmid and the empty effector cloning plasmid pHBTL-3xHA [78] was substituted for lacking effector plasmids ensuring an equal, total DNA load. For example for Fig 2, we have 3 μg of ARR5::LUCm3 + 2 μg ARR2 + 2 μg ARR22 or 3 μg of ARR5::LUCm3 + 2 μg ARR2 + 2 μg empty pHBTL-3xHA plasmid or 3 μg of ARR5::LUCm3 + 4 μg empty pHBTL-3xHA plasmid and so on. Overall, the transfection variance could be adequately measured and representative experiments are shown. Replicates that clearly failed based on visual inspection were excluded from the analyses (e.g. D-luciferin not added). We also explored other normalization techniques throughout this paper. Where overall transfection levels still varied, but the biological readout was similar but not the same, the data were normalized using the method of quantile normalization (QN) (Fig 2, see Math, S26 Fig). Where doubt was still present in transfection variance, pCF203 encoding eGFP was co-transfected and the GFP intensity for the population was obtained by flow cytometry or in a microtiter fluorescence plate reader allowing the LUC emission levels to be rescaled according to the GFP emission level (e.g. S27 Fig). Where needed, these techniques were combined.

Determination of GFP expression intensity in a population

Cells expressing GFP were identified by excitation with a 488 nm laser and comparing peak (FL1) to shoulder (FL2) emission with either an MoFlo XDP (BC; FL1 = 534/30, FL2 = 585/29) or CytoFLEX (BC; FL1 = 525/40, FL2 = 585/42) cytometer; percentage and mean intensities were calculated using the respective software for each machine, i.e. Summit 5.4 for XDP and CytExpert for CytoFLEX. The gating strategy was the same as described in Fig 8. Flow rates were kept to the absolute minimum. Alternatively, GFP expression was monitored in a TECAN Safire fluorescence plate readers using 488 nm ± 5 nm excitation and 515 ±12 nm emission with all other parameters at default levels; Z-gain and detector gain was taken from the brightest well. A comparison between the two methods is provided in (S28 Fig); a linear relationship to the number of cells and total mean emission was obtained for both methods.

Math

Replicates were checked visually to ensure reproducibility and confirm proper experimental execution. Thereafter, experiments were discarded and repeated if the biological and technical replicates were significantly different before treatment. Experiments determined biologically reliable were evaluated and each experimental data set was performed 3 to 5 times or as needed in order to validate the observations; the most representative dataset was chosen for presentation or quantile normalized. The data were analyzed in JMP (SAS) after labeling with a homemade program (LUC_Organizer.jar, KWB; available upon request).

Significance tests were always performed with the non-normalized data using Fischer’s Least Significance Test (LSD) and an alpha level of 0.01 as indicated directly in the graph; significance classes not sharing a letter are significantly different at that level. Note, the LSD test uses the pooled variance for class comparisons. Plots were printed as PDFs from JMP graph output and organized (but not modified) in Adobe Illustrator. Box plots show the median (horizontal line) with the box ends representing the 75th and 25th quantiles; whiskers extend from the box ends to the outermost data point of either quantile side as 1.5*(interquartile range). Data points outside these ranges are outliers.

Specific Activity was calculated by ranking LUC activates and taking the rank-difference of treated to un-treated. Large variance in transfection rates between plates could be normalized by quantile normalization (QN) in R (http://www.r-project.org/) using the normalize.quantiles function of the preprocessCore Package. The raw and QN values used for Fig 1 are shown in S26 Fig. The Area under the Curve (AUC) was calculated as shown in Fig 1 directly in JMP (using a JMP formula) followed by exporting the cumulative value to a separate table for analysis. Alternatively, LUC activity/GFP-mean intensity was explored in Fig 3, S27 and S28 Figs as described elsewhere in the Methods and figure legends.

Cloning

The full-length promoter, defined as from ATG of the respective coding region until the next upstream genic region. ARR5p (-2252 to +6) is previously described [49]. WUSp (-1741 to +1) was cloned into pBT8-LUCm³ via EcoRI and NcoI restriction sites; both are marked in S1 Text. Effector cDNAs for ARR1, ARR2, ARR10, ARR18, AHP2, AHP5, ARR7, ARR4^DXX, ARR15 and ARR22^DXX were obtained or cloned in pENTR (see S2 Table and Acknowledgements). cDNAs for ARR4^DXX were previously described [6]. Site-directed mutagenesis of ARR2 cDNA had been used to generate ARR2^D80N, ARR2^D80E. cDNAs were cloned into pHBTL-3xHA-GW [78] by Gateway^™ recombination. ARR2^DXX cDNAs were recombined into pABind-GFP [79] and pUBQ10(pUGT1-GW) for C-terminal GFP fusions.

Immunoblotting

To confirm the protein expression of HA tagged proteins, cells were pooled and pelleted at 1000g for 5 min at RT after adding five to ten volumes of W5 [154 mM NaCl, 125 mM CaCl₂ x 2 H₂O, 5 mM KCl, 5 mM glucose, pH 5.9]. Glass beads (∼0.2 mm in diameter) were added, followed by 80 to 200 μl of 1x SDS-Laemmli buffer. The cells were vortexed for 2 min and then incubated at 65 °C for 10 min or 95°C for 1.5 min. Debris was removed by centrifugation at 13,000 rpm for 5 min and the supernatant transferred to a new Eppendorf tube. Proteins were then separated using standard SDS-PAGE on a 12% poly:bis-acrylamide (37.5:1) gel and transferred to a PVDF membrane. Membranes were blocked with 5% milk powder in TBS-T followed by incubation with monoclonal antibodies 1:1000 α-HA (rat 3F10; Roche) and 1:5000 α-Rat-AP (Sigma A-9654) or α-HA (mouse ab-24779; abcam) and α-mouse-AP (Bio-Rad Cat.170-6520). AP activity was monitored with NBT+BCIP. POD was detected with anti-HA-POD (1:1000, rat, Roche) and chemiluminescence was imaged with a PEQLAB Fusion Fx7 Imaging System.

FACS-RT-qPCR