The present study was designed to widen the repertoire of phosphoinositide binding modules capable of providing information on PtdIns(4,5)P2 changes in mammalian cells. So far most such studies have utilized the PLCδ1PH-GFP construct, which does not detect PtdIns(4,5)P2in cellular location other than the PM and also could suffer from overestimating PtdIns(4,5)P2 decreases, due to a displacing effect of InsP3. The extent of this distortion may vary from cell-type to cell-type and also depends on expression levels  and, therefore, has been a matter of dispute [10, 31].
In a thorough recent study, several PH domains of S.cerevisiae have been described as capable of phosphoinositide binding, although only one, the Num1p PH domain, showed decent PtdIns(4,5)P2 binding specificity based on several in vitro binding assays . However, in vivo studies in yeast mutants that allowed specific manipulations of phosphoinositides concluded that PM targeting of several of these PH domains (Num1p, Cla4p, Skm1p, Slm1p, and Slm2p) showed PtdIns(4,5)P2 dependence, while the membrane recruitment of the Opy1p-PH domain was found to be independent of phosphoinositides. We extended these studies to mammalian cells and characterized the PtdIns(4,5)P2 dependence of the membrane association of these PH domains using controlled manipulations of PtdIns(4,5)P2. As described before, all of these PH domains showed PM localization and none was recruited to any intracellular membranes, although some (such as the Opy1p PH domain) also showed nuclear or nucleolar localization. The expression of some of these PH domains (namely the Num1p) was very low and required switching the GFP around to improve expression. Adding a nuclear export signal to the GFP-Opy1p-PH construct allowed a better assessment of its membrane binding properties. Remarkably, the behavior of these constructs did not completely match with that described in yeast cells. For example, the PH domains of Cla4p, Skm1p, and Slm2p showed no detectable decrease during PLC activation or after elimination of PtdIns(4,5)P2 via a 5-phosphatase. In contrast, the Opy1p PH domain followed the changes in PtdIns(4,5)P2 in spite of its apparent failure to do so in yeast. Unfortunately, the Num1p PH domain that binds PtdIns(4,5)P2 with the highest specificity in yeast cells did not show any features that would make it a great substitute for PLCδ1PH-GFP. First, its expression was modest and in cells showing higher expression, it created bright vesicular structures budding off the plasma membrane. Moreover, it showed only a modest translocation from the membrane to the cytosol after PLC activation. One of the probes, the Slm1p-PH appeared to be a mixed reporter of PtdIns(4,5)P2 and PtdIns4P in the membrane. This domain was only partially displaced from the membrane after 5-phosphatase recruitment, suggesting that it may still be binding to the PtdIns4P that is generated by the 5-phosphatase.
The discrepancy between the probes behavior in the yeast and mammalian cells is not unprecedented. For example, we did not observe Golgi localization of the PtdIns4P reporter OSH2-PH in mammalian cells  while it clearly showed the Golgi pool in yeast [16, 32]. This is suggestive of a more complex mechanism of membrane recruitment of the PH domains, involving not only phosphoinositides but probably other proteins (or anionic lipids) as interacting partners. Whether a mammalian protein can substitute for the yeast protein in the protein-protein interaction with a yeast PH domain probably varies from one domain to another, making the outcome difficult to predict. Nonetheless, these studies have concluded that none of the predicted PtdIns(4,5)P2 recognizing probes from the yeast PH domain collection have detected other pools of this lipid in intracellular compartments and none has shown any obvious advantage over the mammalian ones to be used in imaging studies. This is in contrast to PtdIns4P recognizing PH domains of the yeast that have been successfully used in mammalian cells [22, 32, 33].
Another PtdIns(4,5)P2 reporter characterized in this study was the Tubby domain of the mammalian transcription factor Tubby protein. The Tubby domain was described as a high affinity PtdIns(4,5)P2 binding module found at the C-terminus of the Tubby protein and being responsible for its lipid binding and membrane localization . It was also claimed as one that does not bind InsP3, although direct experimental evidence for this has not been available in published literature. The Tubby domain has already been used as a PtdIns(4,5)P2 reporter  and two recent studies have examined the usefulness of the full-length Tubby protein  or a mutant form of the Tubby domain  as a PtdIns(4,5)P2 probe in comparative studies similar to ours. Our results with the wild-type Tubby domain had several similarities, but also notable differences to the results described in those studies.
Firstly, both studies found that the Tubby protein as well as the Tubby domain has higher affinity to the PM than PLCδ1PH-GFP. In fact, the wild-type Tubby domain was found to have high enough affinity that it did not show agonist-induced responses in many cells prompting the authors to create a mutant (R332H) with a reduced affinity that was found useful in their studies . Interestingly, in the cells used in our studies the same mutation completely eliminated the membrane localization of the Tubby domain (Szentpetery and Balla unpublished observation) making it unsuitable for further studies. We did not have an explanation for this discrepancy other than the different fluorescent proteins used in the two studies and that the placing of the fluorescent protein relative to the Tubby domain was different. Quinn et al. used eYFP, while we used eGFP fusion constructs and we used GFP in front of the Tubby domain whereas the Quinn study found the mutant Tubby construct having YFP at its C-terminus a more suitable one. Since the dimerization tendency of YFP is larger than that of GFP  it is possible that the higher apparent affinity of the constructs in the Quinn study reflects a dimerization of the fusion proteins, which could explain these differences. Indeed, when we generated the same Tubby domain constructs fused to YFP (still the fluorescent protein in front) we found that the R332H mutant did show some membrane localization, especially in COS-7 cells but not as much as shown in the Quinn study (Szentpetery and Balla, unpublished observation). Although this slight localization does not match those described by Quinn et al., the higher dimerization tendency of YFP still may play into the apparent affinity of the mutant Tubby construct.
Nevertheless, a higher affinity of the full-length Tubby protein (fused to eGFP) to the membrane was also shown in the Nelson study and manifested as a rightward shift (compared to the PLCδ1PH-GFP response) in the dose-response curve with muscarinic agonist measuring the translocation of the Tubby protein from the membrane to the cytosol. Our results are in good agreement with these studies as we also found a significantly higher fraction of the Tubby domain at the PM than with PLCδ1PH, and documented a substantial difference between the agonist-sensitivities of the Tubby domain and PLCδ1PH-GFP translocation responses. Moreover, our FRAP analysis showed that the dissociation of the Tubby domain from the membrane is significantly slower than that of the PLCδ1PH-GFP.
All three studies showed that Tubby domain (or the full-length Tubby protein) displayed no sensitivity to InsP3 in agreement with the original claims . The present study also showed it with direct binding assays using recombinant Tubby domains. Quinn et al. showed that diffusion of InsP3 into the patch pipette caused no translocation of the Tubby domain mutant, while making the PLCδ1PH-GFP fully translocate to the cytosol . The Nelson study, on the other hand, used overexpression of an InsP3 3-kinase to limit InsP3 increases as described in their earlier studies [35, 36]. They found that in contrast to PLCδ1PH-GFP, the Tubby domain translocation after agonist stimulation was insensitive to InsP3 3-kinase overexpression . However, any reduction in InsP3 increase (whether converted to InsP4, or InsP2 or being buffered by an InsP3 binding domain) also reduces the Ca2+ signal and as a corollary will limit PLC activation. This was most likely the case in the Quinn study where overexpression of the InsP3 5-phosphatase eliminated the translocation responses of both PtdIns(4,5)P2 probes . This study also found that the Tubby domain translocation was slightly more sensitive to inhibition by Ca2+ depletion than that of the PLCδ1PH construct. This finding agreed with our observation that the Tubby domain had a slower response to the cytoplasmic Ca2+ increase than the PLCδ1PH-GFP, consistent with an increased "resistance" of the Tubby-domain-covered PtdIns(4,5)P2 to PLC-mediated hydrolysis. It is worth noting that both the Quinn study and ours used HEK293 cells.
In contrast to these findings, a striking reduction was observed in neuroblastoma cells in the translocation responses of the PLCδ1PH domain, but not those of the Tubby protein, when InsP3 3-kinase was expressed, with a strong reduction in the cytoplasmic Ca2+ response . This seemingly contradicted the higher Ca2+ requirement of the Tubby domain translocation also observed in that same study. One possible explanation for this apparent contradiction is, if the magnitudes of InsP3 increases are much larger in the neuronal cells used in the Nelson study than those observed in HEK293 cells. This is not an unreasonable assumption, since N1E-115 neuroblastoma cells require much higher InsP3 increases to induce Ca2+ signaling than other cell types . In search of an explanation for this unique feature, Watras et al. has identified a protein in neurons that binds the InsP3 receptor to significantly decrease its InsP3 sensitivity . This could make neuroblastoma cells (and perhaps other neuronal cell lines) less sensitive to InsP3 increases requiring them to generate a lot more InsP3 in response to agonists that could significantly displace the PLCδ1PH-GFP from the PM. Additionally, any change in the relative proportion of PtdIns(4,5)P2 vs. InsP3 can contribute to a larger InsP3 sensitivity of the PLCd1PH translocation as demonstrated in mathematical modeling studies .
The question remains whether the tighter binding of the Tubby domain to the membrane is related to its lower affinity to InsP3. A significant amount of free InsP3 present in the cytosol could indeed partially displace the PLCδ1PH-GFP from the membrane while not affecting the binding of the Tubby domain. However, overexpression of either the InsP3 kinase , the InsP3 5-phosphatase  or an InsP3 binding domains had little if any impact on the basal localization of the PLCδ1PH-GFP. Moreover, the slightest increases in free InsP3 are immediately detected by the InsP3-R in the form of Ca2+ release. Therefore, we do not favor an explanation that assumes significant free InsP3 levels in the cytoplasm of quiescent cells. We would rather assume that the tighter binding of the Tubby domain to the membrane reflects a genuinely higher affinity either to PtdIns(4,5)P2 itself, or to the lipid together with some additional membrane component(s). This is consistent with the slower dissociation rate as well as the stronger resistance of the PtdIns(4,5)P2 pool covered with the Tubby domain to respond to PLC activation. The fact that the 32P-phosphate or [3H]-myo-inositol labeled HEK293-AT1 cells (the same cells used in this study) lose about 80-90% of their labeled PtdIns(4,5)P2 within 30 sec of stimulation with 100 nM AngII , yet the Tubby domain responds only in a fraction of cells stimulated with the same dose of agonist, suggest that the Tubby domain probably underestimates the changes in PtdIns(4,5)P2 because of its interference with PLC activation (PLC breaks down a larger fraction of PtdIns(4,5)P2 that is not bound to the Tubby domain). Therefore, while the PLCδ1PH domain may overestimate PLC activation because of its InsP3 sensitivity, the Tubby domain may underestimate it because of its higher apparent PtdIns(4,5)P2 affinity. In fact, in the cells used in these studies (HEK293-AT1 and COS-7), the inhibitory effects of the two domains on Ca2+ signaling were very comparable: in one case due to the combined effects on InsP3 quenching and PtdIns(4,5)P2 binding (PLCδ1PH-GFP), while in the other case because of a tighter PtdIns(4,5)P2 binding (Tubby domain). The slight delay in the onset of Ca2+ rise caused by the expression of the PLCδ1PH-mRFP but not the mRFP-Tubby domain is certainly another indication that the PLCδ1PH domain binds InsP3. Importantly, we did not see a significant difference between the two domains in their sensitivity to expression of InsP3 buffering constructs, essentially both being affected probably due to interference with the Ca2+ signal.
An important observation of the present study was the delayed Tubby response relative to that of the PLCδ1PH domain during PLC activation but not when the 5-phosphatase was recruited to the PM. This finding may simply suggest that the InsP3 increase does indeed contribute to PLCδ1PH translocation. However, it is also possible that the endogenous PLC activation mechanism is more sensitive to masking of PtdIns(4,5)P2 by the Tubby domain than the catalysis by the recruited 5-phosphatase. The fact that the Tubby domain does not bind the soluble headgroup, InsP3, even though it binds PtdIns(4,5)P2 with higher affinity, already suggests that the binding force of the Tubby domain must receive significant contributions from interactions with the glycerol backbone. If this were indeed the case, the Tubby domain would obscure the phosphodiester group more and with that could block the access of PLC to this site.
Lastly, these studies have shown a perfect example of how the same lipid, PtdIns(4,5)P2 can regulate several effectors in the same cell simultaneously and yet differently. Just by having two different apparent affinities of two reporters used in the present study - mimicking two endogenous binding partners - is sufficient to elicit a differential response to the same PLC activation. This was also convincingly demonstrated in the experiments where the Kir6.2 KATP channel  or the KCNQ M-channel  (two known PtdIns(4,5)P2 regulated K+ channels) responses were correlated with PtdIns(4,5)P2 changes assessed by the fluorescent reporters.
There are several other conclusions highlighted by these studies. Firstly, all probes based on binding to PtdIns(4,5)P2 (or to any other lipid) show some bias depending on its affinity to the lipid. Secondly, the experimental system may also determine the extent to which the probe manifests its limitation. Thirdly, even the fluorescent protein used a fusion partner can modify the utility of a bioprobe. These should be all reminders that we need to be cautious of selecting a probe and consider it as "the" gold standard.