Palmitoylation of the oncogenic RhoGEF TGAT is dispensable for membrane localization and consequent activation of RhoA

Rho guanine exchange factors (RhoGEFs) control many aspects of the cellular cytoskeleton, and thereby regulate and control processes such as cell migration, cell adhesion and proliferation. TGAT is a splice variant of the RhoGEF Trio, with oncogenic potential. Whether the subcellular location of TGAT is critical for its activity is unknown. Confocal microscopy of fluorescent protein tagged TGAT revealed co-localization with a Golgi marker. Because plasma membrane localized RhoGEFs are particularly effective at activating RhoA, plasma membrane localization of TGAT was studied. In order to quantitatively measure plasma membrane association we developed a novel, highly sensitive image analysis method. The method requires a cytoplasmic marker and a plasma membrane marker, which are co-imaged with the tagged protein of interest. Linear unmixing is performed to determine the plasma membrane and cytoplasmic component in the fluorescence signal of protein of interest. The analysis revealed that wild-type TGAT is partially co-localized with the plasma membrane. Strikingly, cysteine TGAT-mutants lacking one or more palmitoylation sites in the C-tail, still showed membrane association. In contrast, a truncated variant, lacking the last 15 amino acids, TGATΔ15, lost membrane association. The functional role of membrane localization was determined by measuring TGAT activity in single cells with a RhoA FRET-sensor and F-actin levels. Mutants of TGAT that still maintained membrane association showed similar activity as wild-type TGAT. In contrast, the activity was abrogated for the cytoplasmic TGATΔ15 variant. Synthetic recruitment of TGATΔ15 to membranes confirmed that TGAT effectively activates RhoA at the plasma membrane. Together, these results show that membrane association of TGAT is critical for its activity, but that palmitoylation is dispensable.


Introduction
Rho GTPases are a subclass of the Ras superfamily of small GTPases, best known for their regulation of the cytoskeleton in eukaryotes (1,2). Through remodeling of the F--actin network they regulate several important cellular processes like cell migration, cell adhesion, proliferation and cell shape (3--5). Rho GTPases function as molecular switches that cycle between an active GTP--bound form and an inactive GDP--bound form (6). There are several classes of regulatory proteins that influence Rho GTPase activation cycle. Rho guanine exchange factors (RhoGEFs) activate Rho GTPases by accelerating the exchange of GDP for GTP (7). Rho GTPase activating or accelerating proteins (RhoGAPs) are responsible for turning Rho GTPases off by promoting the hydrolysis of the bound GTP to GDP (8). RhoGDIs sequester Rho GTPases in the cytoplasm in their inactive GDP bound state by binding to their prenylated C--tails (9,10). Deregulation of the RhoGTPase cycle has been mainly investigated within the context of cancer (11) and metastasis (12), but is also implicated in other pathologies like neurodegeneration (13), hypertension (14) and hemopathies (15).
The RhoGEF TGAT (trio--related transforming gene in ATL tumor cells) was first identified as an oncogenic gene product in adult T--cell leukemia cells (16). TGAT is formed by alternative splicing of the gene from the RhoGEF Trio, consisting of 255 amino acids encoding the second C--terminal RhoA activating Dbl homology (DH) domain of Trio and a unique extra 15 amino acid extension at its C-terminus. It was found that both its RhoGEF activity and the 15 amino acid extension were required for its transforming potential in NIH3T3 cells in vitro and in vivo (16), for the activation of tumorigenic transcription factor NF--κB via the IκB kinase complex (17), and the stimulation of matrix metalloproteinases (MMPs) via the inhibition of RECK (18). RhoGEFs have been put forward as possible therapeutic targets for the treatment of cancer (19,20). TGAT is also considered as a possible therapeutic target for adult T--cell leukemia and several aptamer--derived inhibitors of TGAT were already developed (21). The 15 amino acid extension at the C--terminus of TGAT is referred to as the C--tail in this manuscript.
The molecular mechanisms underlying the contribution of the C--tail to the oncogenic potential of TGAT have not yet been investigated. Although Rho GTPases have been studied in detail for decades, the spatiotemporal aspects of their signaling and regulation are only starting to be uncovered (22). P63RhoGEF (23), a RhoA activating RhoGEF with a DH domain that has about 70% amino acid sequence homology to TGAT, is constitutively targeted towards the plasma membrane of cells by means of palmitoylation (24,25). The activity of p63RhoGEF towards RhoA at the plasma membrane is auto--inhibited by a Pleckstrin Homology (PH) domain. A synthetic p63RhoGEF variant without the PH domain is constitutively active towards RhoA at the plasma membrane (26).
Since TGAT consists of a DH domain, without an auto--inhibitory PH domain, it is likely that it has high basal RhoGEF activity. Because the C--tail is required for transforming activity, we hypothesized that the oncogenic potential of TGAT originates from subcellular targeting signals in the C--tail.
Here, we have used advanced fluorescence microscopy techniques in single living cells to investigate the role of several residues in the C--tail of TGAT on its subcellular location and function. A novel co--localization analysis based on confocal microscopy images shows that a TGAT mutant devoid of palmitoylation sites in its C--tail is still partially co--localized with the plasma membrane, whereas a TGAT mutant without the complete C--tail is exclusively located in the cytoplasm. Furthermore, we show that plasma membrane localization, but not palmitoylation of two cysteine residues in its C--tail, is necessary for actin polymerization and the activation of RhoA by TGAT. To confirm the plasma membrane as the subcellular site of action of TGAT GEF activity towards RhoA, we make use of a chemical heterodimerization system to target TGAT to several subcellular locations, and find that TGAT has the potential to activate RhoA on several endomembranes beside the plasma membrane.

The influence of cysteines in the C--tail of TGAT on subcellular localization
The 15 C--terminal amino acid residues of TGAT have been described as essential for the oncogenic activity of TGAT (16). We set out to investigate the possible causes for this oncogenic activity in the C--tail in more detail. Close inspection of the sequence of the last 15 amino acid residues reveal two cysteines at position 242 and 253 that are putative palmitoylation sites ( Figure 1A). This, in combination with the several basic and hydrophobic residues present in the C-tail, could potentially target TGAT to endomembrane structures, for instance the plasma membrane.
In order to investigate the subcellular localization of TGAT, we constructed and visualized fusions of TGAT with a fluorescent protein attached to its N--terminus.
HeLa cells transfected with YFP--TGAT and a golgi marker (CFP--Giantin) showed a strong co--localization of TGAT with the golgi apparatus, as previously observed for proteins that are palmitoylated (27,28) Figure 1C). The golgi apparatus localization of TGAT was clearly diminished in the YFP--TGAT Δ15 and YFP--TGAT C242S, C253S mutants, but still present to some extent in the YFP--TGAT C242S and YFP--TGAT C253S mutants. These results support the notion that palmitoylation of TGAT determines its subcellular localization.

A new method to detect plasma membrane co--localization
It was previously shown that the isolated catalytic DH domain of p63RhoGEF at the plasma membrane is sufficient to induce constitutive GEF activity towards RhoA and induce actin remodeling (26). Since TGAT also consists of a constitutive active DH domain, its oncogenic potential might originate from localization at the plasma membrane. Because plasma membrane localization is not immediately apparent from the confocal images of TGAT ( Figure  1B, C), we hypothesized that the fraction of TGAT at the plasma membrane might be very small compared to the unbound, intracellular component. In order to quantify the putative plasma membrane localization of TGAT and its mutants, we developed a novel co--localization method to analyze the confocal images of HeLa cells.
The method employs an untagged soluble RFP as marker for the cytoplasm and a lipid--modified CFP (Lck--CFP) as marker for the plasma membrane. The protein of interest with unknown localization is tagged with YFP, and its fluorescence is attributed to either the cytoplasm or the plasma membrane by linear unmixing.
We briefly describe the procedure here, a detailed description of the quantification method can be found in the Material and Methods. Images of cells coproducing the three fluorescent proteins were used as input.
CFP, YFP and RFP channels were spatially registered and background subtracted  Figure 2H). This procedure allowed for the detection of minute plasma membrane fractions.

C--tail palmitoylation is dispensable for plasma membrane localization of TGAT
To examine the plasma membrane localization of TGAT and its variants, we employed the novel co--localization analysis. First, we examined the dynamic range of our method by analyzing maximal and minimal plasma membrane association by employing a plasma membrane associated YFP (Lck--YFP) and a soluble YFP respectively. For the membrane bound positive control (Lck--YFP) we determined a PM--localized peak of 316 ± 20% ( Figure  3A), whereas for the cytosolic negative control the PM localized peak was 100--fold lower, 3 ± 1% ( Figure 3B).
Next, we examined the plasma membrane association of TGAT and its variants, i.e. YFP--TGAT, YFP--TGAT Δ15 , YFP--TGAT C242S , YFP--TGAT C253S or YFP--TGAT C242S, C253S . Wild type TGAT ( Figure 3C) clearly has a detectable PM component of about 19 ± 5%, which is significantly higher than the negative control. This indicates that TGAT is partly localized at the PM, albeit with a more then 10--fold lower level compared to the positive control Lck--YFP. The mutant TGAT Δ15 , has a PM peak value of 4 ± 1%, which is statistically indistinguishable from the negative control, implying a full cytoplasmic localization ( Figure 3D).
The mutants with the single point mutations, TGAT C242S and TGAT C253S ( Figure   3E,G) exhibit PM peaks of 16 ± 2% and 24 ± 5% respectively. These levels are comparable to TGAT and suggest that these mutants are still localized at the plasma membrane at similar levels as TGAT. The variant with the double point mutation, TGAT C242S, C253S exhibits a PM peak of 11 ± 3%, which is twofold lower than the TGAT level, but higher than the negative control at a 95% confidence level. This suggests that the double cysteine mutant of TGAT still associates with the plasma membrane.  Figure  3H). Soluble CFP exhibits the highest mobility (D = 12.4µm 2 /s, 95% CI [11.0 --13.9]) through the cell periphery.
From the mobility data we conclude that palmitoylation indeed has an effect on the plasma membrane residence time of TGAT, but the difference in diffusion time of TGAT C242S, C253S when compared to that of TGAT Δ15 suggests that a fraction of TGAT C242S, C253S is still interacting directly or indirectly with the plasma membrane.

The influence of cysteines in the C--tail of TGAT on activation of the small GTPase RhoA
We next set out to explore the influence of plasma membrane affinity and palmitoylation of the C--tail of TGAT on its functional properties. In order to investigate the possible activation of RhoA by TGAT and its mutants, we show that the plasma membrane localization conferred by the C--tail contributes to RhoA activation, but suggest that palmitoylation of the C--tail of TGAT does not influence the activation of RhoA in HeLa cells.

The influence of cysteines in the C--tail of TGAT on actin polymerization
Previously, we have shown that plasma membrane located RhoGEF activity results in increased actin polymerization, which was not observed for cytoplasmic located RhoGEF activity (26). In order to investigate the influence of

TGAT activates RhoA at the plasma membrane and mitochondria
In order to investigate our hypothesis that plasma membrane localization of TGAT results in the activation of RhoA, we decided to use a previously described (26,32)  Targeting TGAT Δ15 to the plasma membrane ( Figure 7A) or the CAAX--box of RhoA ( Figure 7B) resulted in a fast and sustained increase in RhoA biosensor activation. Interestingly, targeting TGAT Δ15 to mitochondria also resulted in a fast and sustained increase in RhoA activation ( Figure 7C), whereas targeting TGAT Δ15 to the golgi apparatus ( Figure  7D) only lead to a minimal response on the RhoA sensor. These results lead us to conclude that, beside the plasma membrane, TGAT has the potential to activate RhoA on several subcellular endomembrane locations as well.

Discussion
Despite the well--established observation that the C--tail of TGAT is essential for its oncogenic potential, it has been unclear what underlying molecular mechanism is responsible. Here, we showed that ectopically expressed TGAT is localized to endomembranes, especially the golgi apparatus. Using a novel quantitative co--localization analysis method for confocal images, we showed that a fraction of wild type TGAT is located at the plasma membrane, while TGAT Δ15 is not. Furthermore, we found that mutation of two possible palmitoylation sites in the C--tail, did not affect plasma membrane localization of TGAT. In addition, we observed a reduced mobility of the TGAT mutant devoid of both palmitoylation sites, when compared to mutant TGAT lacking its C--tail, which provides corroborating evidence that this mutant still exhibits membrane affinity. We hypothesized that the three basic residues and 4 hydrophobic residues in the C-tail still provide enough membrane affinity for TGAT to be localized to the plasma membrane. Functional analysis revealed that TGAT and also its cysteine mutants are capable of activating RhoA, resulting in increased actin polymerization.
In contrast, TGAT devoid of the C--tail, TGAT Δ15 , was severely impaired in the activation of RhoA, actin polymerization or translocation of the transcription factor MKL2. Together these results show that mutation of the cysteines in the C-tail of TGAT does not prevent TGAT from finding its target RhoA on cellular membranes. Synthetic recruitment of TGAT Δ15 from the cytoplasm to various endomembranes demonstrated that TGAT displays enhanced RhoGEF activity when it is present at membranes.
Analogous to our findings, it was previously shown that plasma membrane localization and function of the RhoGTPase Chp is also critically dependent on basic and hydrophobic residues in its C--tail, rather than palmitoylation or prenylation (33). Although the effects of palmitoylation, prenylation and basic or hydrophobic amino acid stretches on endomembrane affinity of proteins have been extensively studied (34--36), it is still unclear how lipidation exactly influences subcellular location and membrane affinity. Whether increase in post-translational lipidation modifications simply provide cumulative gradual increases to endomembrane and plasma membrane affinity, or that specific lipidation motif exist to target proteins to different subcellular endomembrane locations, is still unclear.
One striking outcome of this study is that localization of only a minor fraction of protein at membranes is sufficient to cause substantial changes in cell physiology. Importantly, this membrane localized fraction is almost undetected when employing confocal microscopy under optimal conditions. Only by virtue of a new unmixing--based image analysis strategy, the membrane association can be robustly detected. This method is generally applicable and should be of interest to studies where membrane association is hardly or not visible. Of note, combining the analysis method with high resolution imaging strategies (e.g. SIM or other super resolution techniques) will further lower the limit of detection for membrane association. We found that the mobility of the double mutant was higher than wild--type TGAT, as is expected when proteins spend less time sampling endomembranes due to reduced lipidation. Future site mutagenesis studies targeting the basic and hydrophobic residues in the C--tail could possibly shed more light on whether plasma membrane affinity is specifically involved in its oncogenic potential. Another option would be that the C--tail of TGAT contains unknown motifs for protein--protein interactions or targeting to scaffolds. In any case, our results imply that interfering with palmitoylation is not a viable strategy to reduce oncogenic activity of TGAT towards RhoA. This is in contrast to other GTPases like oncogenic RAS, where it has been postulated that interfering with the depalmitoylation machinery might provide therapeutic benefits by mislocalizing RAS activity (37).
The observation that TGAT can activate RhoA when targeted to mitochondrial sites is unexpected. It was previously shown that the DH domain of p63RhoGEF, which shares 70% sequence identity with the DH domain of TGAT at the protein level, does not activate RhoA at mitochondria in a similar assay. Currently, we do not have an explanation for the increased RhoGEF activity that is observed by mitochondrial localized TGAT--DH.
In summary, our results highlight a role for the C--terminal 15 amino acids in the subcellular location of TGAT and we propose that palmitoylation of the C--tail is dispensable for plasma membrane localization, and therefore activation of RhoA and actin polymerization, of TGAT. Furthermore, we introduce a novel co-localization analysis method for confocal images, which can be used to detect minimal fractions of proteins localized at the plasma membrane. This study provides a framework to further investigate the exact origin of the oncogenic potential in the C--tail from the RhoGEF TGAT.

Cell Culture & Sample Preparation
HeLa cells (American Tissue Culture Collection: Manassas, VA, USA) were cultured using Dulbecco's Modified Eagle Medium (DMEM) supplied with Glutamax, 10% FBS, Penicillin (100 U/ml) and Streptomycin (100µg/ml). Cell culture, transfection and live cell microscopy conditions were previously described (26). In the actin staining experiment, DAPI was excited with 420 nm light (slit width 30nm) and reflected onto the sample by a 455DCLP dichroic mirror and emission was detected with a BP470/30 filter, YFP was excited with 500nm light (slit width 30nm) and reflected onto the sample by a 515DCXR dichroic mirror and emission was detected with a BP535/30 filter. RFP was excited with 570 nm light (slit width 10nm) and reflected onto the sample by a 585CXR dichroic mirror and emission of RFP was detected with a BP620/60 filter.

Confocal microscopy
Experiments were performed using a Nikon A1 confocal microscope equipped with a 60x oil immersion objective (Plan Apochromat VC, NA 1.4). For the co-localization experiments the pinhole size was set to 1 Airy unit (<0.8µm) and images with 1.5x zoom of 1024x1024 pixels were acquired. For the MKL2 translocation experiments, the pinhole size was set to 1 Airy unit (<0.8µm) and images were acquired with 1x zoom using tile scans, resulting in images of 4660x4660 pixels. Samples were excited with 447nm, 514nm and a 561nm laser line, and reflected onto the sample by a 457/514/561 dichroic mirror. CFP emission was filtered through a BP482/35 emission filter; YFP emission was filtered through a BP540/30 emission filter; RFP emission was filtered through a BP595/50 emission filter. All acquisitions were corrected background signal. To avoid bleed--through, images were acquired with sequential line scanning modus.

Image Analysis
ImageJ (National Institute of Health) was used to analyze the raw microscopy images. Further processing of the data was done in Excel (Microsoft Office) and graphs and statistics were conducted using Graphpad version 6.0 for Mac, GraphPad Software, La Jolla California USA, www.graphpad.com. Figure 3, Figure 4, Figure 5 and Figure 6 were generated online, using the website http://boxplot.tyerslab.com/. For the MKL2 transcription factor experiments, MKL2 intensity in cytoplasm and nucleus were measured and their ratio was determined. The static FRET data in Figure 4 was processed using a custom made MatLab GUI, which was described before (26).

Boxplots in
The confocal image analysis of the plasma membrane localization for the different constructs and controls was performed by using a combination of wider than that of YFP, which has a smoothing effect on the mCherry profile. This is also consistent with the shape of the profiles in Figure 6B and the residuals (not shown). In order to estimate the confidence intervals on the unmixed CP and PM profiles a bootstrap was performed. Random sets (n = 1000) were drawn from the original set with replacement, and the same normalization and unmixing was performed on these sets, after which 95% confidence interval was calculated based on the standard error of the mean.   and (C) a cytoplasmic marker using a soluble RFP. The CFP channel was registered to the YFP channel based on the spatial shift determined with the positive control. The RFP channel was registered to the YFP channel based on the spatial shift determined with the negative control. Background fluorescence was subtracted from each image prior to processing. Using ImageJ many lines (10 px wide and 6--10 µm long) were carefully drawn perpendicular to the plasma membrane in regions with a well--defined cytoplasm--plasma membrane-extracellular space transition (yellow lines in panel A). For each channel line-scans were performed with the same lines using linear interpolation, the profiles were aligned and centered based on the peak in the Lck--CFP channel and placed on the same axis (panel D). The other channels were aligned and centered using the same positional shift and axis (panels E and F). Subsequently the average profiles were calculated (colored lines in panels D--G), and normalized to unity with respect to the cytoplasm fluorescence level (panel G). Because the Lck--CFP marker is also localized in the cytoplasm, the profile was corrected by subtracting the normalized RFP profile (dashed line panel G). In order to extract the cytoplasmic (CP) and plasma membrane (PM) component, the YFP profile was unmixed using the normalized RFP profile and the corrected Lck--CFP profile (panel H). The 95% confidence intervals (thin solid lines above and below the profiles) were estimated using bootstrapping (See material and Methods for details).