Genomic tagging reveals a random association of endogenous PtdIns5P 4-kinases IIα and IIβ and a partial nuclear localization of the IIα isoform

Among the many known intracellular roles for inositol lipids, the best understood are those in the cytoplasm. However, there are also a number of less well studied functions for these lipids within the nucleus (for reviews, see [1–4]). There is clear evidence in the nucleus for the minor route of PtdIns(4,5)P₂ synthesis [5], the 4-phosphorylation of PtdIns5P by type II PtdIns5P 4-kinases. The probable major route of PtdIns5P synthesis in animals is by 4-dephosphorylation of PtdIns(4,5)P₂ [6,7], with the re-phosphorylation of PtdIns5P by type II PtdIns5P 4-kinases serving to remove it. Thus the PtdIns(4,5)P₂ 4-phosphatases and the type II PtdIns5P 4-kinases can form a reversible pathway regulating the levels of PtdIns5P (see Lecompte et al. [8] for evolutionary arguments for this pathway). PtdIns5P is present in the nucleus [9] and evidence has been presented that it regulates the activity of the transcription factor ING-2 (inhibitor of growth family 2) [10]. Nuclear levels of PtdIns5P increase during the cell cycle [9] or when cells are stressed [11], and it is generally accepted that a key regulator of PtdIns5P levels in the nucleus is PtdIns5P 4-kinase IIβ whose activity in stressed cells is decreased as a result of it being phosphorylated, and thus inhibited, by the p38 MAPK (mitogen-activated protein kinase) [11]. This, together with an increase in nuclear PtdIns(4,5)P₂ 4-phosphatase [7], causes the increase in nuclear PtdIns5P. Nuclear PtdIns5P 4-kinase IIβ has also been reported to associate with and regulate the activity of Cul3 (cullin 3)–SPOP (speckle-type POZ protein) ubiquitin ligase [12].

The nuclear localization of PtdIns5P 4-kinase IIβ is mediated by a novel nuclear localization sequence [13] consisting of a 17-amino-acid acidic α-helix, numbered α-helix 7 in the PtdIns5P 4-kinase IIβ structure described by Rao et al. [14]. We have shown previously that any disruption of this α-helix in PtdIns5P 4-kinase IIβ leads to a cytosolic localization [13,15]. On the other hand, PtdIns5P 4-kinase IIα, which is cytosolic when transfected, lacks this α-helix 7 altogether, and merely introducing that helix into PtdIns5P 4-kinase IIα is sufficient to target the enzyme to the nucleus [13]. However, there is evidence, though not yet definitive, that some endogenous PtdIns5P 4-kinase IIα may be nuclear [13,16,17].

We established previously that endogenous PtdIns5P 4-kinase IIβ is indeed mostly nuclear by genomic tagging of the enzyme in DT40 cells [18]. We also showed that when PtdIns5P 4-kinase IIα is acutely transfected into DT40 cells it is, as in other cells, cytosolic [18]. In the present study we have used the specificity of genomic tagging together with quantitative MS with internal peptide standards to show that endogenous PtdIns5P 4-kinases IIα and IIβ associate in vivo, probably by random heterodimerization, and that approximately 40% of the endogenous IIα isoform is nuclear.

The culturing of DT40 cells, extraction of genomic DNA, Western blotting, protein purification and immunoprecipitation of tagged proteins were all performed as described in Richardson et al. [18].

This was accomplished using the same strategy as for the tagging of PtdIns5P 4-kinase IIβ, described previously [18,19]. Primers were used for the PCR amplification of genomic DNA upstream and downstream of the stop codon of the gene encoding PtdIns5P 4-kinase IIα, PIP5K2A (5′ construct, forward primer 5′-GCATGGTACCTCAGAGTCATGTTAGCTGTG-3′ and reverse primer 5′-GCATTCTAGACGTCAAGATGTTGGCAATAAAG-3′; 3′ construct; forward primer 5′-CATGGATCCTCCCCTCATGTCACACCGGACAG-3′ and reverse primer 5′-GCATGCGGCCGCGTCTGTCAGTGCTACAGAAGTG-3′). The subsequent insertion of the puromycin selection cassette, plasmid linearization, transfection and selection of DT40 lines that had incorporated the insert, and PCR confirmation of the correct incorporation, were all as described in Richardson et al. [18].

DT40 cells were swollen and disrupted by syringing as described previously [18], but the subsequent fractionation was modified as follows. Samples were layered on to 3 ml of 1.8 M sucrose solution and spun at 35000 rev./min for 1 h on a SW55Ti rotor using a Beckman Optima L-100 XP ultracentrifuge (Supplementary Figure S1 at http://www.BiochemJ.org/bj/430/bj4300215add.htm). After centrifugation, two separate volumes were taken from above the buoyant membranes (fraction C, the top 0.4 ml, and fraction P, the remaining volume). The cytoplasmic membranes (fraction M) were added to 0.5 ml of lysis buffer P1 (0.1% Triton X-100 and 0.5% protease inhibitor cocktail in PBS buffer). The nuclei were collected, and their purity determined by brightfield microscopic inspection after staining in Trypan Blue solution or by fluorescent microscopic inspection using a wheat germ agglutinin (conjugated to Alexa Fluor® 555) plasma membrane stain in combination with DAPI (4′,6-diamidino-2-phenylindole) nuclear staining. The nuclei were lysed in 0.8 ml of lysis buffer (50 mM sodium phosphate buffer, pH 7.4, containing 300 mM NaCl, 0.1% Triton X-100 and 0.5% protease inhibitor cocktail) on ice for 20 min, and the supernatant (fraction N) was collected after centrifugation. Identical proportions of each fraction were taken for SDS/PAGE and immunoblotting with anti-histone or anti-α-actin antibodies, and the remainder was used for affinity purification of the FLAG–(His₆)₂-tagged proteins using TALON beads, followed by Western blotting with anti-FLAG antibodies [18].

The anti-FLAG M2 monoclonal antibody (Stratagene) was coupled to Protein G–Sepharose beads using dimethyl pimelimidate dihyrochloride, and these were incubated for 2 h at 4 °C with 400 μl of DT40 cell lysate [18]. Beads were pelleted by centrifugation and washed twice in PBS buffer, and bound protein eluted with 3× FLAG peptide buffer [100 μg/ml 3× FLAG peptide (Sigma–Aldrich) and 0.5% protease inhibitor cocktail in TBS (Tris-buffered saline)].

PtdIns5P 4-kinase assays were carried out as described previously [20] with slight adaptations. Substrate (300 pmol of PtdIns5P) lipid was dried under vacuum and micelles made by sonication in kinase buffer (50 mM Tris/HCl, pH 7.4, containing 10 mM MgCl₂, 80 mM KCl and 2 mM EGTA). Recombinant human type II PtdIns5P 4-kinase isoforms, cloned and purified as described in [21], were added to the reaction mixture with 10 μCi of [γ-³²P]ATP for 90 min at 30 °C. Lipids were extracted and separated by silica-gel TLC [20] and detected by autoradiography. Radioactivity was quantified by scintillation counting of labelled PtdIns(4,5)P₂.

Total mRNA was extracted from approx. 3×10⁷ DT40 cells from three different cultures of each clone tested, with the SV Total RNA Isolation System (Promega) and cDNA libraries were constructed from each preparation with the Sprint RT (reverse transcriptase) complete kit (Clontech). Primers for singleplex qRT-PCR were designed using Primer3 [22]. Owing to the high sequence similarity between PIP5K2A and PIP5K2B, primer pairs were designed spanning exon 10 and the 3′-UTR (untranslated region) (PIP5K2A, forward 5′-TCAAAGCGCTTCTTGGACTT-3′ and reverse 5′-CTACCCTCGTGGTCACTGCT-3′; PIP5K2B forward 5′-GCAGTACTCCAAACGCTTCA-3′ and reverse 5′-GGCAAGTAGCCCCTCTTCTC-3′). Primer pairs specific to the tagged genes were generated from exon 10 and the inserted tag sequence (PIP5K2A-tag forward 5′-CGCAGAAATTTCAACCGTTA-3′; PIP5K2B-tag forward 5′-AGAGATCTCCACCGTGAACC-3′; common reverse 5′-CTTGTCATCGTCGTCCTTGT-3′). A primer set was also designed against the chicken β-actin housekeeping gene, spanning exons 1 and 2 (forward 5′-TGGCAATGAGAGGTTCAGGT-3′ and reverse 5′-CGGATATCCACATCACACTTCA-3′). All primers were obtained from Sigma–Aldrich and produced single PCR products in conventional reactions. PCR amplification was performed over 40 cycles (95 °C for 15 s, 60 °C for 30 s) in a 96-well plate using SYBR Green I master mix (Applied Biosystems, Warrington, U.K.) and C_t (threshold cycle) values were collected using a CFX96 qRT-PCR detection system (Bio-Rad Laboratories). Dissociation curves for each primer set indicated a single product and no-template controls were negative after 40 cycles. Control reactions using RNA from each extraction as a template were negative, showing that the preparations were free from genomic DNA contamination. Primer and template concentrations were optimized to give C_t values of 20–27. Validation experiments showed equivalent relative amplification efficiencies between products from PIP5K2 genes and the chicken β-actin reference gene (Supplementary Figure S2 at http://www.BiochemJ.org/bj/430/bj4300215add.htm). Normalized expression ratios were calculated by the 2^−ΔΔCt method [23].

Absolute protein quantification of PtdIns5P 4-kinase II isoforms was achieved by isotope dilution and SRM (selective reaction monitoring). Briefly, proteotypic peptides unique to each isoform were selected as internal standards giving consideration for mass spectra, size and residue composition (see below and Table 1). Internal standards were synthesized (Anaspec) incorporating a single ¹³C₆,¹⁵N₄-labelled arginine residue at the C-terminus and each was independently quantified by amino acid analysis [24].

For quantifying total protein in unfractionated cells, 7 mg of lysate digest, with 30 pmol of internal standards was separated by strong cation exchange (0–150 mM KCl over 90 min on a 200 mm NEST column with a 2.1 mm internal diameter). Subsequent fractions were each dried, resuspended in 0.1% formic acid and analysed by UPLC (ultra-performance liquid chromatography)–SRM. To determine the ratio of the PtdIns5P 4-kinase IIα and IIβ isoforms in pull-downs, protein precipitates were digested by incubation overnight with 40 ng of trypsin, with 100 or 200 fmol of internal standards, and analysed by nano-UPLC–SRM as appropriate.

Nano-flow UPLC (300 nl/min) was used for separation of pull-down digests and normal-flow UPLC (400 μl/min) used for the second dimension of separation for lysate digests. In both cases peptides were eluted from preformed 1.7-mm-diameter BEH packed columns [Waters; nanoUPLC (150 mm column with 75 mm internal diameter) and UPLC (100 mm column with 2.1 mm internal diameter)] using gradients of linearly increasing acetonitrile concentration from 0 to 45% over 40 min (nano-flow) or 10 min (normal-flow).

SRM was conducted using a Quattro Premier XE (Waters). For each peptide monitored by SRM, up to four transitions were configured to monitor the four most intense fragment ions (Supplementary Table S1 at http://www.BiochemJ.org/bj/430/bj4300215add.htm). Transition dwell times were 5 ms for two-dimensional UPLC–SRM and 50 ms for nanoUPLC–SRM. Signals derived from internal standards and analytes of interest were mean smoothed (window of 2 with 1 iteration) and integrated using QuantLynx (Waters). Ratios of standards and analyte peak integrals were calculated and analyte concentration determined.

We planned to use the DT40 cell line JPR3, in which PtdIns5P 4-kinase IIβ is genomically tagged with a FLAG–(His₆)₂ tag [18], to identify protein partners that associate with this enzyme. As a proof of principle that we could specifically pull-down and detect endogenous PtdIns5P 4-kinase IIβ by MS, we conducted a series of pull-downs from 3 l of cells using anti-FLAG antibody beads, followed by three washes of the beads and elution with 3× FLAG peptide. We performed this on triplicate samples from WT (wild-type) and JPR3 cells in parallel, and MS analysis revealed that PtdIns5P 4-kinase IIβ is the major protein pulled down, and that it was detectable only in pull-downs from JPR3 and not from WT cells (results not shown). With one exception (see below), the next 20 most abundant proteins identified (mostly contaminants) were present in much lower amounts and were common to WT and JPR3 pull-downs, establishing that this stringent protocol is highly specific for the tagged PtdIns5P 4-kinase IIβ.

However, unexpectedly we also detected PtdIns5P 4-kinase IIα, which was identified as the other major protein present, and was again found only in pull-downs from JPR3 cells and not WT cells; quantitative results (Table 2) confirm that this co-precipitation of PtdIns5P 4-kinase IIα with the IIβ form is highly reproducible. Given that only the IIβ form is FLAG-tagged in JPR3 cells, these results unambiguously establish that the two PtdIns5P 4-kinase II isoforms associate with each other and the absence of any detectable third protein suggests that their association is direct. As both PtdIns5P 4-kinase IIβ [14] and IIα [25] are homodimers in solution and the proposed dimerization interface is conserved between the two isoforms [14], it is likely that this association reflects a heterodimerization (Supplementary Figure S3 at http://www.BiochemJ.org/bj/430/bj4300215add.htm), but as yet we have no results directly addressing the structure of the interaction.

There are many examples of different isoforms of the same enzymic activity heterodimerizing to regulate localization and function, such as the myotubularins [26]. However, these phenomena are usually based on associations of transfected proteins and the actual extent of association of endogenous species, and how this may be regulated, have not been quantified. Our genomic tagging approach gave us a unique opportunity to address these issues, and prompts three questions: (i) Is this PtdIns5P 4-kinase II association random and complete, i.e. dictated only by the relative mass levels of the two isoforms? (ii) Given that PtdIns5P 4-kinase IIβ is mostly nuclear [18], are there therefore significant levels of endogenous PtdIns5P 4-kinase IIα in the nucleus? (iii) What is the physiological relevance of these phenomena?

To answer the first question, we needed to know both the mass ratio of the enzymes in the JPR3 DT40 line, and their relative levels in the pull-downs from JPR3 cells. Both PtdIns5P 4-kinase II isoforms are expressed at very low levels in DT40s, and were at the limit of accurate quantification by MS, with PtdIns5P 4-kinase IIα being the predominant isoform (see below). Additionally we quantified the mRNA levels by qRT-PCR of WT and JPR3 DT40 cells, and the results shown in Figure 1 and Supplementary Figure S2 show that the ratio of PtdIns5P 4-kinase IIβ to IIα mRNA is approx. 0.23 and is similar in WT and JPR3 cells. Moreover, the results also suggest that the tagged and untagged PtdIns5P 4-kinase IIβ alleles are expressed in approximately equal amounts. If these proportions are reflected in the protein level, and an association between the isoforms were random, then we can define the cellular PtdIns5P 4-kinase IIβ/IIα ratio as x, and the proportion of PtdIns5P 4-kinase IIβ that is tagged as t, so the ratio of PtdIns5P 4-kinase IIβ to IIα in a pull-down from PtdIns5P 4-kinase IIβ-tagged cells will be (2−t)x−1 (see the Supplementary Experimental section at http://www.BiochemJ.org/bj/430/bj4300215add.htm). Thus if we take x as 0.23 and t as 0.5, random association would give a predicted PtdIns5P 4-kinase IIβ/IIα ratio in the pull-downs from JPR3 cells of 1.36:1.

MS was applied to quantify PtdIns5P 4-kinase II isoforms at the protein level using an AQUA (absolute quantification) [27] isotope dilution strategy. Given the high sequence identity, suitable proteotypic candidate peptides were limited. Unique candidates were ranked according to their perceived fitness for quantification by considering the quality of empirical data and residue composition. The two best proteotypic peptides for each isoform were selected and AQUA peptides for use as internal standards synthesized (Table 1).

Initially the cellular PtdIns5P 4-kinases IIα and IIβ in whole cell lysates were analysed; 7 mg of lysate digest from 2×10¹⁰ cells, with 30 pmol of AQUA peptides, were separated by two-dimensional LC and peptides were quantified by selective reaction monitoring using a triple quadrupole mass spectrometer. Preliminary experiments indicated which of the first-dimension fractions the AQUA (and analyte) peptides would elute in. Two transitions were monitored for each peptide across the selected first-dimension fractions by UPLC–SRM and quantifiable signals with acceptable signal-to-noise response were integrated (Figure 2). Signals were within the linear response range of the instrument (results not shown) enabling direct calculation of the analyte from the known amount of internal standards. Both diagnostic peptides for PtdIns5P 4-kinase IIα were detected and the protein was quantified at 1.5 pmol/mg of total protein, but no quantifiable signals were detected for PtdIns5P 4-kinase IIβ.

We then turned to analysis of the anti-FLAG immunoprecipitations from JPR3 cells, and good quantification of both isoforms was achieved. Pre-elution wash steps did not affect the PtdIns5P 4-kinase IIβ/IIα ratio observed, but provided a better signal-to-noise response allowing more reliable quantification; the data from four independent experiments using two different sets of internal AQUA peptides (Table 2) yielded a mean±S.E.M PtdIns5P 4-kinase IIβ/IIα ratio in the immunoprecipitates of 1.39±0.10. This is close enough to the ratio of 1.36 predicted above to suggest strongly that the association between PtdIns5P 4-kinases IIα and IIβ is random and complete.

To gain independent data of this association, and to answer directly the second question posed above (the likelihood of some endogenous PtdIns5P 4-kinase IIα being nuclear), we knocked the same FLAG-(His₆)₂ tag into the 3′-end of one allele of the PIP5K2A gene by a strategy identical with that which we used previously for the PIP5K2B gene [18]. This proved more difficult to do than for the PIP5K2B gene, and once we had achieved it, expression of tagged protein in several DT40 lines, including the one we studied most (MW2), declined during the first eight passages (assessed by Western blotting), stabilizing at a level per cell approx. 5-fold lower than the tagged PtdIns5P 4-kinase IIβ expression in JPR3 cells (results not shown). Moreover, PCR analysis of the MW2 cells revealed unequal transcription of the tagged and untagged alleles (Figure 1), and an increase in overall PtdIns5P 4-kinase IIα transcription, presumably to compensate for the low expression of tagged protein. Clearly there are problems with tagging this isoform that do not apply for PtdIns5P 4-kinase IIβ in JPR3 cells, for which expression of tagged protein was always stable, and where no distortion of transcription is evident compared with WT cells (Figure 1). This distortion of both transcription and translation is probably due to a gene-specific effect of the tag, particularly of the puromycin selection cassette in the 3′-UTR. Overall these factors limit the quantification that can be achieved, but the MW2 line nevertheless yielded quantitative results to complement data from JPR3 cells.

In two independent immunoprecipitation experiments where quantification was clear the PtdIns5P 4-kinase IIα/IIβ ratios were 11:1 and 8.8:1. As discussed above, our analysis by Western blotting suggested that the level of total tagged PtdIns5P 4-kinase IIα protein in MW2 cells was approx. 20% that of IIβ in JPR3 cells. So if we assume that the untagged PtdIns5P 4-kinase IIα allele was translated into protein with the same efficiency in MW2 cells as in WT and JPR3 cells, then from Figure 1 we can to a first degree of approximation estimate that x=0.2 and t=0.05 in MW2 cells. These parameters lead to a predicted PtdIns5P 4-kinase IIα/IIβ ratio in pull-downs from MW2 cells of 10.3:1, which is consistent with the mean of 9.9:1 from the above experiments.

Immunocytochemistry in these tagged DT40 cell lines is not possible with such low expression levels [18], and so we previously used cell fractionation to show that most of the genomically tagged PtdIns5P 4-kinase IIβ is nuclear [18]. In the present study we quantified the cytoplasm-to-nuclear ratio of the type II PtdIns5P 4-kinase isoforms by another fractionation protocol (Supplementary Figure S1 and see the Materials and methods section) that is more quantitative than that used previously [18] in that we quantified all fractions to give a complete balance sheet of the distribution of the enzymes and markers. Figure 3(A) gives representative images of nuclei obtained from this protocol. The results (Figure 3B) confirm our previous observation [18] that >95% of PtdIns5P 4-kinase IIβ is nuclear. Note that the fractionation protocol has caused significant leakage of histones (as exemplified by histone H1) from the nuclei (Figures 3B–3D), indicating some damage to the nuclear membrane. The retention of most of the PtdIns5P 4-kinase IIβ in the nucleus (Figure 3B), in contrast with this histone leakage, suggests that the enzyme is not freely diffusible within the nucleus but is interacting with nuclear structures.

Examination of the nuclear fraction with DAPI staining revealed <1% nuclei with any detectable cytoplasmic debris attached (Figure 3A). We also probed for remaining plasma membrane using wheat germ agglutinin, and again found <1% of nuclei had any detectable signal (Figure 3A). Additionally we quantified cytoplasmic contamination of the nuclei by examining for actin by Western blot, and found it to be low (Figure 3D); indeed, the levels of actin (10% of total) that are detected may be mostly accounted for by bona fide intranuclear actin [28] and/or F-actin closely associated with the nucleus [29]. This protocol is therefore apparently effective enough to remove >90% of the cytoplasm from the nuclei, yet fractionation of MW2 cells showed that 40% of the tagged PtdIns5P 4-kinase IIα is nuclear and 60% cytoplasmic (Figures 3C and 3D). This suggests that in contrast with transfected PtdIns5P 4-kinase IIα [13,18], the endogenous enzyme is significantly nuclear. It also is important to emphasize another difference; we found that endogenous cytoplasmic PtdIns5P 4-kinase IIα is in the membranous fraction, not in the cytosol (Figures 3C and 3D). We are not sure to which membrane PtdIns5P 4-kinase IIα is attached, but from our current knowledge we suggest that it is most likely to be the plasma membrane, to which it could be targeted by its interaction with type I PtdIns4P 5-kinases [30].

The simplest interpretation of our results overall is that the PtdIns5P 4-kinases IIα and IIβ associate randomly, and one consequence may be that PtdIns5P 4-kinase IIβ therefore targets the IIα to the nucleus. The third question posed above then remains: what is the physiological significance of these events? One possibility is that the two isoforms could be differentially regulated within the nucleus, e.g. we have shown that PtdIns5P 4-kinase IIα is phosphorylated by casein kinase II on a residue that is unique to that isoform [31,32]. However, another (not mutually exclusive) possibility is that the two isoforms have very different enzymic activities. This idea is made more likely by our previous demonstration of an association between PtdIns5P 4-kinases IIα and IIγ; the latter had negligible activity when we assayed it in comparison with the former, and IIγ may target IIα to the vesicular compartment in which it (IIγ) is localized [21,33]. To address the issue directly we assayed the PtdIns5P 4-kinase activity of recombinant human PtdIns5P 4-kinases IIα and IIβ in parallel (Supplementary Figure S4 at http://www.BiochemJ.org/bj/430/bj4300215add.htm and Supplementary Table S2). The specific activity in μmol/min/mg of protein of PtdIns5P 4-kinase IIα is 3.25×10⁻² (S.E.M.±1.9 × 10⁻⁴, n=4), and that of IIβ is 1.57×10⁻⁵ (S.E.M.±7.9 × 10⁻⁷, n=4). Thus strikingly PtdIns5P 4-kinase IIα is 2000-fold more active than the IIβ isoform, which means that in a PtdIns5P 4-kinase IIα/IIβ heterodimer the activity of the PtdIns5P 4-kinase IIα is going to be by far the most significant.

We should note that our suggestion that PtdIns5P 4-kinase IIβ may target IIα to the nucleus could throw a new light on experiments where PtdIns5P 4-kinase IIβ has been increased by transfection [34] or decreased in a knockout mouse [35], both of which cause alterations in (cytoplasmic) insulin sensitivity, an effect primarily evident in muscle and liver. One consequence of manipulating PtdIns5P 4-kinase IIβ might be to alter significantly the amount of the IIα isoform present in the cytoplasm, and this would be especially evident in muscle and liver; of the tissues we analysed these have the highest PtdIns5P 4-kinase IIβ/IIα ratios [21].

More importantly, the apparently complete and random association (probably heterodimerization) of endogenous proteins that we have revealed and quantitatively examined, plus the much greater enzymic activity of PtdIns5P 4-kinase IIα compared with IIβ, make it reasonable to pose the question: is a (or even the) major function of PtdIns5P 4-kinase IIβ simply to act as a IIα nuclear-targeting protein? This may not be the whole story, e.g. the requirement for PtdIns5P 4-kinase IIβ to be catalytically active to fulfil at least some of its nuclear functions is suggested by a striking dominant-negative effect of a kinase-dead PtdIns5P 4-kinase IIβ construct on Cul3–SPOP ubiqutin ligase [12]. As discussed previously, there are precedents for inactive or nearly inactive isoforms of proteins targeting active isoforms to cellular locations, an example taken from phosphoinositide metabolism being the myotubularins [26]. Much of the evidence for this idea is based on association of transfected proteins, and the degree, regulation and extent of association and targeting in vivo is not always clear. In the present study we have combined the powers of DT40 genomic tagging with MS to place this type of functional relationship, at least for PtdIns5P 4-kinases IIα and IIβ, on a new level of quantitative clarity.

While the present work was in progress we learned that another group independently discovered the association between these two PtdIns5P 4-kinase II isoforms [36].

AQUA

absolute quantification

Cul3

cullin 3

DAPI

4′,6-diamidino-2-phenylindole

qRT-PCR

quantitative real-time PCR

SPOP

speckle-type POZ protein

SRM

selective reaction monitoring

UPLC

ultra-performance liquid chromatography

UTR

untranslated region

WT

wild-type

Minchuan Wang performed the pull-downs, made the IIα-tagged cells, did the fractionation experiments and co-wrote the initial draft of the paper. Nicholas Bond co-designed and then performed and co-interpreted all the MS experiments, and co-wrote the inital draft. Andrew Letcher contributed to the pull-downs and fractionation experiments. Jonathan Richardson made the IIβ-tagged cells. Kathryn Lilley co-designed and co-interpreted all the MS experiments, and co-wrote the first draft. Robin Irvine conceived the project, co-designed all the experiments other than the MS and co-wrote all drafts of the paper. Jonathan Clarke performed the qRT-PCR experiments, made the recombinant enzymes and assayed their activity, contributed to the fractionation experiments, co-designed all the experiments other than the MS and co-wrote all drafts of the paper.

We thank Frederic Langevin, K.J. Patel and members of the Irvine laboratory for helpful discussions.

This work was supported by a Wellcome Trust programme grant [grant number 079194/Z/06 (to R.F.I.)]. N.J.B. is supported by a Ph.D. studentship from the Engineering and Physical Sciences Research Council [grant number EP/D013615/1].