The chicken B-cell line DT40 has two isoforms of phosphatidylinositol 5-phosphate 4-kinase (PI5P4K), α and β, which are likely to exist as a mixture of obligate homo- and hetero-dimers. Previous work has led us to speculate that an important role of the β isoform may be to target the more active PI5P4Kα isoform to the nucleus. In the present study we expand upon that work by genomically tagging the PI5P4Ks with fluorochromes in the presence or absence of stable or acute depletions of PI5P4Kβ. Consistent with our original hypothesis we find that PI5P4Kα is predominantly (possible entirely) cytoplasmic when PI5P4Kβ is stably deleted from cells. In contrast, when PI5P4Kβ is inducibly removed within 1 h PI5P4Kα retains its wild-type distribution of approximately 50:50 between cytoplasm and nucleus even through a number of cell divisions. This leads us to speculate that PI5P4Kα is chromatin-associated. We also find that when cells are in the exponential phase of growth PI5P4Kβ is primarily cytoplasmic but translocates to the nucleus upon growth into the stationary phase or upon serum starvation. Once again this is not accompanied by a change in PI5P4Kα localization and we show, using an in vitro model, that this is possible because the dimerization between the two isoforms is dynamic. Given this shift in PI5P4Kβ upon nutrient deprivation we explore the phenotype of PI5P4K B-null cells exposed to this stress and find that they can sustain a greater degree of nutrient deprivation than their wild-type counterparts possibly as a result of up-regulation of autophagy.
The phosphatidylinositol 5-phosphate 4-kinases (PI5P4Ks) are a family of three (PI5P4Ks α, β and γ), whose cellular functions are still poorly understood (for reviews see [1–3]). Their principal activity is believed to be to remove, and thus regulate the levels of, their substrate phosphatidylinositol 5-phosphate (PI5P), which is present in cells at slightly lower levels than phosphatidylinositol 3-phosphate (PI3P) and much lower levels than phosphatidylinositol 4-phosphate (PI4P) [1,4]. PI5P has been reported to be present in the plasma membrane [4,5], intracellular membranes  and the nucleus , and several studies suggest effectors and functions for this lipid in the nucleus [2,7–9] and the cytoplasm .
The localization of the PI5P4Ks is clearly central to understanding their functions and those of PI5P, yet the literature has been confusing on this critical issue. Using immunolocalization of endogenous proteins, Boronenkov et al.  reported that both PI5P4Kα and PI5P4Kβ were present in the nucleus, whereas transfection studies suggested that PI5P4Kα is primarily cytoplasmic and PI5P4Kβ is mostly nuclear with its localization being dictated by an unusual nuclear localization sequence, an α-helix  numbered α-7 in the structure of PI5P4Kβ . However, using highly isoform-specific antibodies, Bultsma et al.  reported that both PI5P4Kα and PI5P4Kβ were in the nucleus and cytoplasm, whereas, using genomic tagging of PI5P4Ks in DT40 cells, Richardson et al.  and Wang et al.  reported that PI5P4Kβ was almost entirely nuclear, whereas PI5P4Kα was 40% nuclear and 60% cytoplasmic.
A complicating factor in understanding these issues has been the discovery that PI5P4Kα and PI5P4Kβ heterodimerize [14,16] as might be expected given that their dimerization sequences  are near identical between the isoforms. Indeed Bultsma et al.  showed that PI5P4Kα and PI5P4Kβ are likely to exist as a mixture of obligate homo- and hetero-dimers in vivo. DT40 cells, being of avian origin, lack a PI5P4Kγ isoform [16,17], but we have shown that PI5P4Ks α and γ can also heterodimerize at least in vitro , so there is no reason to suppose that PI5P4Kβ and PI5P4Kγ will not heterodimerize too. So two key questions concerning PI5P4K and PI5P functions in the nucleus are: is PI5P4Kβ cytoplasmic  because it is pulled out of the nucleus by one or both of the other isoforms, and is a primary role of PI5P4Kβ to target the much more enzymatically active PI5P4Kα [14,16] to the nucleus ?
To resolve these issues we have further employed the remarkable genetic power of DT40 cells to tag both the endogenous PI5P4Ks with enhanced monomeric green fluorescent protein (EmGFP), to delete both alleles of PI5P4Kβ and to tag both alleles of PI5P4Kβ with the auxin degron  so that we can remove it from the cells within minutes. In all cases we can follow the localization of the endogenous proteins by cell fractionation or confocal imaging, and using these strategies we have gained significant insight into the dynamic relationship between these two proteins.
MATERIALS AND METHODS
DT40 tissue culture
Basic culture medium for DT40 cells was formulated according to the protocol of . To 500 ml of RPMI 1640 medium was added 50 ml of FBS, 5 ml of 200 mM L-glutamine, 5 ml of chicken serum and 120 μl of 50 mM 2-mercaptoethanol. Cells were cultured at 41°C under 5% CO2.
Generation of targeting constructs for targeted transfection of DT40 cells
Targeting constructs were built into an appropriate plasmid vector [typically pBluescript SK+(Stratagene)]. For gene deletions sequences of genomic DNA flanking the region to be deleted (targeting arms) were PCR-amplified from DT40 genomic DNA and ligated into the plasmid vector separated by an antibiotic-resistance cassette to allow selection of drug-resistant clones. The targeting arms were typically 2 kb in length. Following successful transfection and integration into the germline by homologous recombination the antibiotic-resistance cassette replaced the region of the gene targeted for deletion. For integration of tags into the germline a similar strategy was employed, but in this case the targeting arms were manipulated such that successful recombination led to introduction of the new sequence of choice. See the Supplementary Methods for further details.
Targeted transfection of DT40 cells
Targeting constructs were linearized using an appropriate restriction enzyme, recovered by lithium chloride precipitation and transfected into DT40 cells in the exponential phase of growth by electroporation (600 V, 25 μF, exponential pulse, Bio-Rad Gene Pulser XCell). The cells were then immediately seeded out in 96 -well plates and 24 h later the appropriate selection antibiotic(s) was added. Colonies resulting from single cells were picked and subjected to sequencing in order to verify correct construct integration. All results were verified in multiple independently derived clones.
Fractionation of nuclear and cytoplasmic proteins in DT40 cells
This was performed essentially as set out in . In total, 107 cells were harvested by centrifugation at 300 g for 5 min and washed twice in PBS before being resuspended in 1 ml of suspension buffer (0.32 M sucrose and protease inhibitor cocktail in PBS). Ten microlitres of 10% (v/v) IGEPAL CA-630 in PBS was added to the cell suspension with mixing by gentle pipetting to give a final detergent concentration of 0.1%. The suspension was immediately centrifuged at 2000 g for 30 s at 4°C in a microcentrifuge to pellet the nuclei. The supernatant was aspirated and kept as the cytoplasmic protein fraction. The nuclei were resuspended in 1 ml of suspension buffer to wash them. Prior to centrifugation at 150 g for 5 min at 4°C, an aliquot was examined by light microscopy to check that the nuclei were clean. The nuclear pellet was next resuspended in 1 ml of nuclear extraction buffer (20 mM HEPES, pH 7.7, 1.5 mM magnesium chloride, 0.42 M sodium chloride and protease inhibitor cocktail in water) and incubated on ice for 20 min. This suspension was then centrifuged at 6000 g for 15 min at 4°C in a microcentrifuge and the supernatant was aspirated as the nuclear protein fraction. In all fractionation experiments the protein species being examined was immunoprecipitated from the fractions before being blotted. We always used an excess of immunoprecipitating antibody and so all the protein we could recover from the fractions is loaded on to the gel.
Immunoprecipitation of proteins from DT40 cells
The antibody for immunoprecipitation was conjugated to Protein G–Sepharose beads which were then used to immunoprecipitate the target protein from cell-free extract. An excess of antibody over antigen was used as determined empirically. Antigen was eluted from the beads by boiling in SDS loading buffer in preparation for denaturing PAGE.
Following denaturing SDS/PAGE, proteins were transferred on to an appropriate blotting membrane, which was probed with antibodies as detailed in the Supplementary Methods and detected using chemiluminescent reagent of an appropriate sensitivity.
Images were acquired with a Leica SP5 confocal microscope attached to a Leica DMI6000 inverted microscope stand. For GFP imaging, excitation was with the 488 nm line of the argon laser and emission was detected between 500 and 550 nm. For all imaging the confocal pinhole was set to 1 Airy unit, and a ×40 oil-immersion 1.25 numerical aperture (NA) objective lens was used. All imaging was performed on live cells, which were embedded in Matrigel to prevent cell movement and transferred to uncoated glass-bottomed MatTek dishes. Just prior to imaging, full growth medium was removed and replaced with RPMI 1640 medium without Phenol Red. Imaging was performed at 41°C under 5% CO2.
Untagged recombinant PI5P4Kγ was allowed to reach equilibrium at a concentration where dimers were the predominant form  and then diluted to a concentration where we would expect dissociation into monomers . Samples of this solution were then taken every 5–15 min to monitor the kinetics of dimer dissociation. These samples were cross-linked within 1 s (not shown) with a 50-fold molar excess of BS3 [bis(sulfosuccinimidyl)suberate] and the reaction was quenched with 50 mM Tris/HCl, pH 7.4. The mixture of dimers and monomers present was assayed by denaturing SDS/PAGE followed by Western blotting with an anti-PI5P4Kγ antibody (see the Supplementary Methods).
Cell cycle analysis by flow cytometry
In total, 107 cells were fixed in ethanol then stained with 25 μg/ml propidium iodide containing RNase A at 250 μg/ml. Cells were analysed using a BD FACScan instrument, exciting with the 488 nm line of the argon laser and collecting through the 650 nm long-pass filter. Dead cells were excluded as were cellular doublets and higher order clumps. At least 1×105 events were collected for each sample.
RESULTS AND DISCUSSION
PI5P4Kα is almost exclusively cytoplasmic upon genetic deletion of PIP4K2B
If our speculation based upon previous data  is correct that PI5P4Kα requires targeting by PI5P4Kβ to enter the nucleus then we reasoned that PI5P4Kα should be exclusively cytoplasmic in PIP4K2B-null cells. We therefore created a cell line in which both alleles of PIP4K2B were deleted and one allele of PIP4K2A was endogenously tagged at the C-terminus with the coding sequence for EmGFP (PIP4K2AEmGFP/wt/wt PIP4K2B−/− cells). Note that karyotype analysis shows DT40 cells exhibit trisomy for chromosome 2, the chromosome on which the PIP4K2A gene is located (not shown). PIP4K2AEmGFP/wt/wt cells [with wild-type (WT) PIP4K2B] acted as a control. As anticipated , PI5P4Kα is distributed approximately equally between cytoplasm and nucleus in a WT PIP4K2B background, whereas deletion of PIP4K2B results in a marked shift of PI5P4Kα out of the nucleus (Figures 1A and 1B). It is impossible to know whether the small amount of PI5P4Kα that persists in the nuclear fraction in PIP4K2B-null cells is truly nuclear or whether this represents an artefact of an inevitably less than perfect separation of nucleus from cytoplasm. Certainly no nuclear PI5P4Kα–EmGFP is visible in the nuclei of these cells upon confocal microscopy (results not shown) although this is a less sensitive technique than immunoprecipitation and Western blotting and so does not really help to answer the question. Whatever the truth of the matter, it is certainly accurate to say that genomic deletion of PIP4K2B results in PI5P4Kα distributing away from the nucleus largely, possibly entirely. It is, of course, important to know that endogenous tagging does not affect protein function, and we address this further in the Supplementary Results and Discussion.
Effect of deletion of PIP4K2B on the cellular localization of PI5P4Kα
Acute depletion of PI5P4Kβ does not result in any redistribution of PI5P4Kα
In the light of these findings we thought it would be interesting to study the kinetics of redistribution of PI5P4Kα away from the nucleus upon acute removal of PI5P4Kβ. We therefore created a cell line in which PI5P4Kβ was endogenously tagged (at both alleles) with the auxin degron system  in order to enable the localization of PI5P4Kα endogenously tagged with EmGFP to be studied upon rapid depletion of PI5P4Kβ by addition of auxin. Note that we placed a His6–FLAG tag in frame with the degron tag on the PIP4K2B alleles to allow for easy immunoprecipitation and blotting of the fusion protein. The resultant cell line is PIP4K2AEmGFP/wt/wtPIP4K2Bdegron/degron. To our surprise removal of PI5P4Kβ, which, as far as we can tell from immunoprecipitation and Western blotting is (virtually) complete (Figure 2A), has no effect on the localization of PI5P4Kα which remains approximately equally distributed between the cytoplasm and nucleus (Figure 2B). This persists even after 16 h of PI5P4Kβ depletion by which time these rapidly proliferating cells have gone through two or three rounds of division with breakdown and re-formation of the nuclear envelope. This makes unlikely the explanation that PI5P4Kα has simply been ‘ferried’ to the nucleus by PI5P4Kβ where it becomes trapped by the nuclear envelope, but, coupled with the requirement PI5P4Kα has for PI5P4Kβ to enter the nucleus in the long-term (Figures 1A and 1B), it is possible that nuclear PI5P4Kα is chromatin-associated. Unfortunately problems with auxin toxicity after approximately 24 h mean that we are unable to continue PI5P4Kβ depletion for long enough to observe the ‘acute’ phenotype transform into the ‘chronic’ phenotype.
Effect of acute removal of PI5P4Kβ on the cellular localization of PI5P4Kα
In these cells we also took the opportunity to study the distribution of degron-tagged PI5P4Kβ between cytoplasm and nucleus. To our surprise PI5P4Kβ was predominantly (approximately 90%) cytoplasmic (Figure 2B). This is in complete contradiction to our previous data where PI5P4Kβ endogenously tagged with His6–FLAG was found to be 90% nuclear . In order to confirm this finding and convince ourselves that everything we found with the degron system was not artefactual, we decided to re-explore the entire issue of the localization of PI5P4Kα and PI5P4Kβ using endogenous EmGFP tags in an otherwise WT background, particularly as we have data elsewhere (Bulley, S.J., Droubi, A., Clarke, J.H., Anderson, K.A., Stephens, L.R., Hawkins, P.T. and Irvine, R.F., unpublished work) demonstrating that endogenous EmGFP tags do not affect the in vivo function, and therefore presumably localization, of the PI5P4Ks. Note that we have already presented data from PIP4K2AEmGFP/wt/wt cells (Figure 1A), but we decided to go on to tag a second PIP4K2A allele with EmGFP in order to enhance the signal for confocal microscopy.
Localization of EmGFP-tagged PI5P4Kα and PI5P4Kβ
Once again we used the highly homologous recombination frequency of the DT40 line to knock EmGFP tags into two PIP4K2A alleles and both PIP4K2B alleles in separate cell lines to create PIP4K2AEmGFP/EmGFP/wt and PIP4K2BEmGFP/EmGFP cells respectively. We looked at the localization of the tagged proteins by both confocal microscopy and cell fractionation. As expected PI5P4Kα has an approximately equal cytoplasmic/nuclear localization (Figures 3A and 3B) [cytoplasm to nuclear ratio 1.2, 95% CI (confidence interval) 1.1–1.3, n=3], which is consistent with what we have seen throughout the present study, and consistent with what we have reported before in these cells  and what others have reported for the endogenous protein in mammalian cells [11,14].
Cellular localization of PI5P4Kα and PI5P4Kβ with genomic EmGFP tags in DT40 cells
PI5P4Kβ, in contrast, is almost entirely cytoplasmic (Figures 3A and 3B) (cytoplasm to nuclear ratio 11.5, 95% CI 8.2–14.7, n=3). Although this is consistent with what we saw with degron-tagged PI5P4Kβ (above) it is completely inconsistent with what we previously reported in DT40s [15,16] where we found PI5P4Kβ to be almost exclusively nuclear upon both overexpression and endogenous tagging, and indeed inconsistent with most reports on other cells [11,14]. We wondered whether this inconsistency was a result of the different growth state of the cells, as in the present experiments we were careful to assay the cells in the exponential phase of growth, whereas in our earlier experiments  we had grown the cells into or close to the stationary phase in order to obtain the maximum yield of protein for downstream applications. We therefore let the cells grow on into the stationary phase and an almost complete shift in PI5P4Kβ towards the nuclear compartment was revealed (Figure 4A), making the present results consistent with our earlier study  as long as the cells are sampled under the correct experimental conditions. Live-cell confocal microscopy reveals that translocation of PI5P4Kβ to the nucleus upon growth into the stationary phase occurs in some cells before others (Figure 4C, left panel). We confirmed this result in the cells we have already described (Figure 2) in which PI5P4Kβ is tagged at both alleles with the auxin degron and His6–FLAG (Figure 4B). Once again acute removal of PI5P4Kβ by addition of auxin resulted in no change in the subcellular localization of PI5P4Kα (Figure 4B). To investigate what the major factor might be in causing this change we transferred PIP4K2BEmGFP/EmGFP cells in the exponential phase of growth to a medium with no soluble nutrients (amino acids etc.), or to one with no added serum. The former caused no change (results not shown), but in cells deprived of serum, a translocation of PI5P4Kβ to the nucleus occurred within 1 h (Figure 4D). Finally, we repeated these experiments with PIP4K2AEmGFP/EmGFP/wt cells, and saw no significant change in PI5P4Kα localization, either growing the cells into the stationary phase (results not shown) or depriving them of serum (Figure 4D).
Nuclear translocation of PI5P4Kβ upon cell growth into the stationary phase or serum starvation
Kinetics of dimer dissociation
In the context of seemingly independent behaviour of PI5P4Kα and PI5P4Kβ under conditions of serum deprivation and acute depletion of PI5P4Kβ, it is relevant to remember that the two PI5P4K isoforms are extensively heterodimerizing in DT40 cells . So the independent movement of PI5P4Kβ into the nucleus, whereas PI5P4Kα remains apparently unmoved (Figure 4D), plus the unchanged location of PI5P4Kα, when PI5P4Kβ is removed, both suggest that the two PI5P4Ks are in quite a rapid equilibrium; these observations would be inconsistent with a stable heterodimerization where, once bound to each other, the two remained so for a long time. However, we currently have no idea at all what the likely timescale of their binding to each other is. But we can address this question indirectly by exploring the dissociation of PI5P4Kγ homodimers.
We used this surrogate strategy for three reasons: first, because it is essential that we should use untagged protein (tags might enhance or hinder the interaction) and we do not have the necessary isoform-specific antibodies for PI5P4Kα and PI5P4Kβ, whereas we do have an anti-PI5P4Kγ antibody ; secondly, creating and isolating a pure PI5P4Kα–PI5P4Kβ heterodimer population to investigate dissociation would be extremely difficult; and, thirdly, the hydrogen-bonding between PI5P4Kγ homodimers is almost exactly the same as PI5P4Kα–PI5P4Kβ heterodimers , so PI5P4Kγ homodimers are a suitable model to investigate this question.
Figure 5 shows a typical experiment where we incubated recombinant PI5P4Kγ at a concentration at which it is dimeric , and then diluted rapidly to a concentration at which it is entirely monomeric , taking samples every 5–15 min and adding a cross-linker at a concentration that stabilized the dimers within 1 s (results not shown). Because our antibody does not immunoprecipitate we could not easily concentrate the diluted enzyme before gel electrophoresis and Western blotting, so detection and quantification of the extremely dilute protein preparations pushes the sensitivity of Western blotting to the extreme. But it is nevertheless clear even from this semi-quantitative experiment (Figure 5) that the dissociation of the PI5P4Kγ dimers is detectable within a few minutes with an approximate half-life of 10–15 min. This shorttime frame is consistent with the independent behaviour of PI5P4Ks α and β in the contexts discussed above.
Kinetics of dissociation of PI5P4Kγ dimers
PIP4K2B−/− cells are able to sustain high levels of nutrient deprivation
Given the nuclear translocation of PI5P4Kβ upon proliferation of cells into the stationary phase, we wondered whether the growth characteristics of PIP4K2B-knockout cells would differ from those of their WT counterparts. As a control we generated PIP4K2A-null cells (PIP4K2A−/−/−). Consistent with the subcellular localization data, the growth characteristics of PIP4K2A−/−/− cells are unchanged compared with WT (results not shown) but PIP4K2B−/− cells remain in the exponential phase of growth for longer than their WT controls. This results in PIP4K2B−/− cells attaining a higher cell density in the stationary phase than WT DT40 cells (Figure 6A). This difference in growth characteristics is interesting, particularly since the type 1 and type 2 phosphatidylinositol phosphate kinases can interact , and the type 1 phosphatidylinositol phosphate kinases in turn interact with the G1/S checkpoint protein retinoblastoma . We therefore wondered whether loss of PIP4K2B resulted in dysregulation of the G1/S checkpoint. Cell cycle analysis by flow cytometry shows that PIP4K2B−/− cells continue to proliferate for longer than WT cells (which we knew anyway from the growth curves), but following this extended period of growth they do enforce the G1/S checkpoint appropriately (Figure 6B).
Growth characteristics of PIP4K2B−/− cells
Autophagy is up-regulated in PIP4K2B−/− cells
Given that the outgrowth of PIP4K2B−/− cells does not seem to be due to breakdown of the G1/S checkpoint we wondered whether autophagy, a mechanism for promoting cellular survival during periods of starvation , might be affected. We therefore assayed autophagy by blotting against the autophagy marker light chain 3 II (LC3-II), and found that autophagy is enhanced in PIP4K2B−/− cells (Figure 7A) but not in PIP4K2A−/−/− cells (results not shown). The best characterized regulator of autophagy is mammalian target of rapamycin complex 1 (mTORC1), which acts as a negative regulator . We therefore hypothesized that mTORC1 would be less active in PIP4K2B−/− cells (but not in PIP4K2A−/−/− cells) and indeed find this to be the case by blotting for phosphorylation of the mTORC1 target p70 S6K (S6 kinase) (Figure 7B). We have data elsewhere (Bulley, S.J., Droubi, A., Clarke, J.H., Anderson, K.A., Stephens, L.R., Hawkins, P.T. and Irvine, R.F., unpublished work) showing that loss of PI5P4Kβ does not affect Akt phosphorylation and so the influence of PI5P4Kβ does not appear to extend to mTORC2 targets. It may therefore be that PI5P4Kβ regulates autophagy via a direct or indirect impact on mTORC1 activity. It was recently suggested that knockdown of any of the three mammalian isoforms of PI5P4K can induce autophagy due to build up of their substrate PI5P which in turn recruits autophagy effectors normally held to be dependent upon PI3P . Our data are not entirely consistent with this but investigation of the mechanism linking PI5P4Kβ to autophagy may be an interesting and fruitful avenue for future work.
Autophagy and mTORC1 activity in PIP4K2B−/− cells
Alaa Droubi, Simon Bulley, Jonathan Clarke and Robin Irvine designed and performed the experiments. Simon Bulley and Robin Irvine wrote the paper.
We thank Ashok Venkitaraman and Gerard Evan for the gifts of reagents.
This work was supported by Sidney Sussex College, the Cambridge Overseas Trusts, and the Säid Foundation (to A.D.); the Medical Research Council [grant number RG64071 (to R.F.I. and J.H.C.)]; and the British Pharmacological Society A.J. Clarke Studentship (to S.J.B.).
These authors contributed equally to this work.
Present address: Faculty of Life Sciences, University of Manchester, The Michael Smith Building, Oxford Road, Manchester M13 9PT, U.K.
Present address: Department of Haematology, University of Cambridge School of Clinical Medicine, Cambridge Institute for Medical Research Wellcome Trust/MRC Building, Cambridge Biomedical Campus Box 139, Hills Road, Cambridge CB2 0XY, U.K.