Do mammals make all their own inositol hexakisphosphate?

Abstract

A highly specific and sensitive mass assay for inositol hexakisphosphate (InsP₆) was characterized. This centres around phosphorylating InsP₆ with [³²P]ATP using a recombinant InsP₆ kinase from Giardia lambia, followed by HPLC of the ³²P-labelled products with an internal [³H]InsP₇ standard. This assay was used to quantify InsP₆ levels in a variety of biological samples. Concentrations of InsP₆ in rat tissues varied from 10–20 μM (assuming 64% of wet weight of tissue is cytosol water), whereas using the same assumption axenic Dictyostelium discoideum cells contained 352±11 μM InsP₆. HeLa cells were seeded at low density and grown to confluence, at which point they contained InsP₆ levels per mg of protein similar to rat tissues. This amounted to 1.952±0.117 nmol InsP₆ per culture dish, despite the cells being grown in serum shown to contain no detectable (less than 20 pmol per dish) InsP₆. These results demonstrate that mammalian cells synthesize all their own InsP₆. Human blood was analysed, and although the white cell fraction contained InsP₆ at a concentration comparable with other tissues, in serum and platelet-free plasma no InsP₆ was detected (<1 nM InsP₆). Human urine was also examined, and also contained no detectable (<5 nM) InsP₆. These results suggest that dietary studies purporting to measure InsP₆ at micromolar concentrations in human plasma or urine may not have been quantifying this inositol phosphate. Therefore claims that administrating InsP₆ in the diet or applying it topically can produce health benefits by increasing extracellular InsP₆ levels may require reassessment.

INTRODUCTION

InsP₆ was originally discovered as the major plant phosphate storage compound [1,2], hence its alternative name, phytic acid. It was generally assumed to be plant specific until the rekindled interest in inositol phosphates that followed the discovery of Ins(1,4,5)P₃ as a second messenger [3] led to the realization that InsP₆ is also found in animal cells (e.g. [4,5] and see [6,7] for reviews).

In the intervening years, the route by which animals synthesize InsP₆ was controversial [6], but further studies have led to a likely consensus. Unlike slime moulds [8] and plants [9], which can synthesize InsP₆ by stepwise phosphorylation directly from inositol, animals apparently incorporate the first three phosphates while the inositol is in its lipid form. That is, Ins(1,4,5)P₃ is formed by phospholipase C action on PtdIns(4,5)P₂, and Ins(1,4,5)P₃ is then phosphorylated up to InsP₆. The synthesis of Ins(1,3,4,5,6)P₅ from Ins(1,4,5)P₃ is most likely to be mediated by a two-step conversion catalysed by a single enzyme, Ipk2 (inositol polyphosphate multikinase) [10–12]; this enzyme is essential for mammalian embryonic survival [13], though it is not certain yet whether that is because of its activity in converting Ins(1,4,5)P₃ into Ins(1,3,4,5,6)P₅. The alternative route from Ins(1,4,5)P₃ to Ins(1,3,4,5,6)P₅ via Ins(1,3,4,5)P₄, Ins(1,3,4)P₃ and Ins-(1,3,4,6)P₄ is probably quantitatively minor, at least in fibroblasts, as suggested by the results of Leyman et al. [14], who showed that mouse embryonic fibroblasts devoid of any Ins(1,4,5)P₃ 3-kinase isoforms synthesized InsP₆ at a rate indistinguishable from wild-type fibroblasts; in contrast, fibroblasts derived from Ipk2 knockout mice show a greatly compromised InsP₆ synthesis [13]. The final step of InsP₆ synthesis (as in plants and slime moulds [8,9]) is the 2-phosphorylation of Ins(1,3,4,5,6)P₅ by InsP₅ 2-kinase [11,15], an enzyme which is also essential to viable mammalian development [16].

However, although we now know that mammalian cells can synthesize InsP₆, how they probably do so, and that it is essential that they do synthesize InsP₆, the contribution of exogenous InsP₆ to mammalian InsP₆ homoeostasis is not so clear. A number of studies have reported increases in InsP₆ levels in tissue and body fluids resulting from the supply of exogenous InsP₆ [17–20], and other work has suggested that InsP₆ added to cell cultures or tissues can impinge on their physiology or pathology [21–23]. As a result of these and other studies there are many claims of therapeutic or other beneficial effects of InsP₆ (e.g. [23,24] and many of these are predicated on the idea that exogenous InsP₆ can directly and significantly influence both extra- and intracellular InsP₆ levels in the body.

In the present study, we have designed and characterized an unambiguous mass assay for InsP₆. This assay is more time-consuming and complex than others (e.g. [18,25]), but its particular strength is a very high sensitivity and applicability to most sample types. Our results provide some new insights into InsP₆ levels in a number of tissues and address whether mammalian cells can make all their own InsP₆. Our results also cast doubts on the idea that InsP₆ is directly taken up into human tissues and cells.

EXPERIMENTAL

Reagents

All chemicals and biochemicals used were of AnalaR grade unless specified otherwise. [³H]InsP₆ was from PerkinElmer. A [³H]InsP₇ standard (probably 5-[PP]InsP₅, see the Results section) was prepared by incubating [³H]InsP₆ with Giardia lambia InsP₆ kinase under the normal assay conditions (see below for details) such that approx. 50% of the InsP₆ was phosphorylated. The correct nomenclature for the product of InsP₆ phosphorylation by InsP₆ kinase is [PP]InsP₅, and we have used this nomenclature when necessary for unambiguity (e.g. when specifying isomers). However, for most of the present paper we have used InsP₇ as a simpler alternative. Sources of InsP₅ isomers and scyllo-inositol hexakisphosphate were as described previously [26].

InsP₆ kinase

This is a recombinant enzyme from G. lambia. The sequence of Giardia InsP₆ kinase (Genbank® Nucleotide Sequence Database accession number AY227443) was amplified from Giardia genomic DNA (the gene contains no introns), and then cloned into the hexahistidine bacterial vector pET43A, which fuses the solubility-enhancing bacterial protein NusA to the N-terminus of the fusion protein. The enzyme was purifed by talon-bead absorption and imidazole elution as described previously [27]. As discussed previously [27], there is no obvious reason why a recombinant mammalian InsP₆ kinase should not be substituted for the Giardia enzyme in this assay, although the high affinity and specificity of the Giardia InsP₆ kinase (see the Results section) does make it particularly suitable in this context.

Extraction protocol

For solid tissue samples, 20–50 mg of tissue was weighed before homogenizing with a ground glass pestle in 180 μl of water. This was rinsed with 70 μl of water and then transferred to a 1.5 ml Eppendorf tube, and 50% (w/v) TCA (trichloroacetic acid) added to give a final concentration of 8.3% (v/v). After vortexing and standing on ice for 30 min, samples were centrifuged at 13000 g for 15 min in a Microfuge, the supernatant was removed and the pellet was kept for protein determination (see below). The supernatant containing inositol phosphates was washed ten times with water-saturated diethyl ether to remove TCA [28] and finally neutralized with 15 μl of 1:10 (v/v) saturated ammonia. Tubes were spun in a centrifugal vacuum dryer for 5 min at 35 °C and then for 30 min at 45 °C to remove the ether, before being adjusted to 300 μl with water and the pH checked to ensure it was approx. pH 6.0. For most assays, we took as an aliquot the equivalent of 1 mg of wet weight of starting tissue for the InsP₆ assay.

InsP₆ mass assay

Samples of tissue extract were incubated in a final volume of 300 μl containing the following: 30 μl of 10× assay buffer [500 mM Hepes/NaOH (pH 7.5), 1 M KCl, 50 mM MgCl₂, 10 mM EGTA and 5 mg/ml BSA], 6 μl of 600 μM 2-mercaptoethanol and 10 μl of [γ-³²P]ATP (10 μCi; PerkinElmer). After 60 min incubation at 37 °C, 45 μl of 0.1 M glucose and 6 μl of hexokinase (20 units, Sigma) were added, and the incubation was continued at 37 °C for another 90 min to convert all the remaining radioactivity into [³²P]glucose 6-phosphate.

The reaction was stopped by the addition of 0.5 ml of 0.4 M ammonium formate/0.1 M formic acid, and the samples were loaded on to 0.5 ml of Dowex Formate in a BioRad Econo Column (with a disc of Whatman Filter paper number 1 on top of the Dowex resin) [29]. Samples were rinsed on to the column with 4 ml of 0.4 M ammonium formate/0.1 M formic acid, and the column was washed with 10 ml of 0.8 M ammonium formate/0.1 M formic acid to remove any remaining [³²P]glucose 6-phosphate. The higher inositol phosphates were eluted with 2×1 ml of 2.5 M ammonium formate/0.1 M formic acid. They were spiked with 600–2000 d.p.m. (disintegrations per min) of [³H]InsP₇ and loaded on to a Whatman Patisil 10 SAX WCS analytical HPLC column (4.6×250 mm, catalogue number 4226-001) or a similar-sized column packed with the same material by Phenomenex.

HPLC

The column was eluted at 1 ml/min flow rate with the following programme: solution A, water and solution B, 1.2 M (NH₄)₂HPO₄ (pH 3.8) adjusted with phosphoric acid. The profile gives the percentage of solution B (v/v) added: 0 min, 0%; 5 min, 0%; 30 min, 78%; 75 min, 78%; 76 min, 100%; 81 min, 100%; 82 min, 0%; 102 min, 0%.

Fractions (1 ml) were usually collected between 40 and 75 min (InsP₇ elutes at approx. 60 min) and were counted in a scintillation counter after adding 10 ml of Ultima-Flo AP ammonium phosphate-tolerant scintillation fluid (PerkinElmer). The [³H] and [³²P] windows were calibrated using standards to ensure each isotope was counted independently (spill over between channels was <1%).

Protein assay

Pellets from the TCA tissue extracts were dissolved in 4 ml of 8 M urea by sonication (in a Decon ultrasonic bath) and vortexing. The majority of material dissolved, but in samples with a high fat content some fat could remain in suspension and may interfere with the assay (this did not happen in any sample analysed in the present study). Samples were adjusted to a final concentration of 6 M urea, and assayed for protein using the Bio-Rad protein assay, based on a method published previously by Bradford [30], using BSA in 6 M urea as a protein standard.

Blood fractionation

Blood was taken from unfasted human volunteers who denied having taken aspirin in the preceding 48 h, mixed with acid citrate dextrose (97 mM sodium citrate, 78 mM citric acid and 110 mM D-glucose) and centrifuged at 200 g for 20 min (all centrifugations were performed at 4 °C). The supernatant from this centrifugation (platelet-rich plasma) was further centrifuged at 2500 g for 20 min, and the resulting supernatant (platelet-free plasma) was removed and frozen. The pellet from the first centrifugation was resuspended in 25 ml of resuspension buffer [10 mM Hepes/NaOH (pH 7.4), 10 mM glucose, 1 mM MgSO₄, 145 mM NaCl and 5 mM KCl], and then centrifuged (2500 g for 20 min). The ‘buffy coat’ (mostly white cells) on top of the pellet was removed, along with the supernatant. The pellet (mostly erythrocytes) was resuspended in 25 ml of resuspension buffer and re-centrifuged (2500 g for 20 min). The small amount of buffy coat on this pellet was removed, along with the supernatant, and the remaining pellet (erythrocytes) was then resuspended in 10 ml of resuspension buffer and frozen. The combined supernatants from the two preceding centrifugations were re-centrifuged and resuspended several times as above (2500 g for 20 min) until no erythrocytes were evident at the bottom of the pellet, at which point the final pellet of mostly white cells was resuspended and frozen.

For preparation of serum, fresh blood was left to stand for 4 h, then centrifuged at 3000 g for 20 min, and the supernatant (serum) was decanted and frozen.

Human experimentation

All human fluid samples were taken with informed consent of the donors in accordance with the Declaration of Helsinki (2000) of the World Medical Association and with the relevant ethical guidelines of the University of Cambridge. The urine sample was taken from an adult male at 10:30 h, 3 h after a meal consisting of a bowl of breakfast cereal and a cup of coffee, with the subject subsequently having free access to water before the sample was taken.

RESULTS

Characterization of the enzyme and the assay

Using [³H]InsP₆, we established that the Giardia kinase has a K_m of approx. 60 nM (results not shown). We used this enzyme to make our own [³H]InsP₇ standard for HPLC. Because the Giardia gene is closely related to InsP₆ kinases from other eukaryotes, we assumed that its product is 5-[PP]InsP₅ (see [31] for discussion), though we have not established its structure. Draskovic et al. [32] have shown that the mammalian InsP₆ kinases are remarkably specific for the 5-position such that even on prolonged incubation they prefer to add a third phosphate there rather than use another position on the inositol ring. We exposed bona fide 5-[PP]InsP₅ synthesized from InsP₆ by a mammalian InsP₆ kinase [a gift from Dr S.B. Shears, Inositol Signaling Group, NIEHS (National Institute of Environmental Health Sciences), Research Triangle Park, NC, U.S.A.] to the Giardia enzyme for several hours, but detected no formation of any InsP₈, which suggests that Giardia InsP₆ kinase is a 5-kinase too, and that extra phosphorylation of 5-[PP]InsP₅ is not a complicating factor in the InsP₆ assay.

In the context of this assay, the crucial point is that we always included a [³H]InsP₇ standard made by the Giardia enzyme in every HPLC run, and collected and counted 1 min samples individually for both isotopes. This leads to a precise co-chromatography of the ³H-labelled standard and the ³²P-labelled product formed from InsP₆ in the cell extract (e.g. Figure 1); any sample in which this co-chromatography was not absolutely precise was discarded (except for some blanks, where low levels of radioactivity meant that the correlation was less evident: see below). Note that in the Figures, the left-hand axis (♦) is always the ³H-labelled InsP₇ standard, and the right-hand axis (■), whose scale varies greatly between Figures (which serves to illustrate the precise co-chromatography of the two isotopes), is the ³²P counts for the same fractions.

Figure 1. HPLC profile of a liver InsP₆ assay.

Three typical profiles from three assay tubes are illustrated, with the radioactivity in the InsP₇ region (indicated by the peak in [³H]InsP₇ elution) shown. Left-hand axis (◆), [³H]; right-hand axis (■), [³²P]. Note the differing scales on the [³²P] axes [dpm (d.p.m.)]. (A) Blank incubation (control), (B) 5 μl of liver extract, and (C) 5 μl of liver extract with 10 pmol InsP₆ added just before the assay was performed.

We have discussed elsewhere previously [27], and discounted, the possibility of InsP₅ in extracts interfering significantly with the assay. Briefly, InsP₅s are poor substrates for the Giardia enzyme (see below), and any PPInsP₄ products formed elute well before InsP₇ under our HPLC protocol. Any tri-phosphorylated InsP₅ [32] would be formed in minute amounts during our 60 min incubations and also would not precisely co-elute with 5-[PP]InsP₅. High background levels in the region of the HPLC profile where PPInsP₄ elutes prevented us from extensive quantitative analysis of InsP₅s as substrates for the Giardia InsP₆ kinase. However, in one set of experiments with a fresh batch of [γ-³²P]ATP (and con-sequently lower background) we were able to gain some results on the approximate efficacy of the six InsP₅ isomers (four of which are presented as racemic pairs) as substrates. Using 300 nM substrate, the rate of phosphorylation of InsP₅ isomers relative to InsP₆ (InsP₆=100%) was 2-OH-InsP₅, 0.5%; 1/3-OH-InsP₅, 3.2%; 4/6-OH-InsP₅, 3.0%, and 5-OH-InsP₅ showed no detectable phosphorylation over background levels. In parallel experiments, scyllo-inositol hexakisphosphate was phosphorylated at 4.7% of the rate of InsP₆.

We explored whether other compounds present in rat tissue extracts might interfere with the assay by ‘spiking’ samples with known quantities of InsP₆. We found significant levels of interference caused by factors present in the extracts, which varied considerably between tissues (from a 2–10-fold underestimate of the InsP₆ added), and we attribute this to endogenous ATP diluting our [γ-³²P]ATP, plus other unknown compounds inhibiting the enzyme. We tried a number of ‘purification’ protocols to separate InsP₆ from these interfering factors, but this only led to losses of InsP₆, so, in the end, we decided to adopt a protocol where we performed six assays for any single tissue sample. Three are aliquots of the tissue extract, and three are identical aliquots spiked just before the assay with a known amount (usually 10 pmol) of InsP₆. From the difference between these latter spiked samples and the unspiked samples, we can calculate the d.p.m. of InsP₇ formed from the known InsP₆, and comparing this with a 10 pmol standard assayed at the same time quantifies the ‘quenching’ present in each individual tissue sample. This can then be used to correct the InsP₇ formed in the unquenched samples. This is a rather laborious process (as we also included blank incubated controls in all experiments), but it does allow us to be confident that each sample is quantitatively assayed.

This approach assumes that the quenching is linear with InsP₆ concentration (doing a complete standard curve within each sample would be the only rigorous way of ensuring that). Although we believe such linearity is indeed likely, some non-linearity would not seriously compromise the conclusions drawn below other than to lead to an underestimate of InsP₆ in some tissues listed in Table 1. Whenever possible, we used quantities of cell extract that contained approximately the same amount of InsP₆ as the added spike, which self-evidently helps to ensure linearity and also simplifies the quantification of the degree of quenching. We tested our extraction protocol by spiking several samples of rat tissues with [³H]InsP₆ before extraction, and found that we recovered 80–85% up to the point of adding the enzyme to the assay. We have corrected all our results for tissue extracts by assuming 80% recovery.

Table 1. InsP₆ content of tissues and fluids.

Tissues were extracted and assayed as described in the Experimental section. Each assay was performed in triplicate, and results are means±S.E.M. For all tissues, at least one independent assay was performed on another sample, with similar results (results not shown). Note that the InsP₆ concentration for all samples except serum, plasma and urine, is an estimated intracellular concentration.

Tissue	InsP6 (pmol/mg of protein)	InsP6 concentration
Rat liver	39.8±0.43	11±0.12 μM
Rat brain	194.7±25.5	14.5±1.9 μM
Rat kidney	55.5±7.1	10.9±1.4 μM
Rat lung	69.8±12.4	15.8±2.8 μM
HeLa cells	175±16.6	37.6±3.5 μM
Slime mould	3645±114	352±11 μM
Calf serum	−	<8 nM
Human erythrocytes	0.037±0.0057	26±4.0 nM
Human white cells	43.5±6.1	−
Human serum	−	<0.5 nM
Human plasma	−	<0.5 nM
Human urine	−	<5 nM

InsP₆ content of rat tissues

Figure 1 shows a typical set of single HPLC profiles from an assay performed on an extract from rat liver. The combination of the triplicate samples leads to a level of 4.75 pmol in the sample, which represents 0.84 mg of wet weight of tissue (0.149 mg of protein). If we assume an 80% recovery and that there is 0.64 g water/g of tissue (see [33,34] for derivation of this latter assumption), then the complete assay leads to an estimated intracellular level of 11±0.12 μM InsP₆ in the liver.

In a further set of experiments, we then assayed InsP₆ levels in a selection of rat tissues, and the results are shown in Table 1. We observe no obvious pointers to tissue-specific functions in these results, as the concentrations of InsP₆ were broadly similar. Perhaps the most relevant observation in that regard is that comparing the two internal tissues in the body exposed to the highest and lowest O₂ levels (lung and kidney respectively), there is no large difference. If a principal function of InsP₆ in vivo was to act as an antioxidant [26,35], then one might expect there to be much more InsP₆ in the lung. In fact, the likelihood of InsP₆ interacting with iron ions (the reason for its antioxidant action) inside cells has been questioned in the light of the discovery that InsP₆ probably exists mostly as a penta-Mg²⁺ salt in vivo [36].

Slime moulds

Our calculated concentrations of InsP₆ in mammalian tissues are overall consistent with the indirect calculations derived from equilibrium labelling of cell cultures with [³H]inositol (e.g. [6,37]), but nevertheless we sought an independent validation of the assay. One organism for which there is already an accurate mass estimation of InsP₆ is D. discoideum, which has an estimated intracellular concentration of InsP₆ assayed by inorganic phosphorous determination as 594 μM [8]. This value is significantly higher than the estimated maximum solubility of the penta-Mg²⁺ salt of InsP₆ proposed to exist in vivo (approx. 50 μM [38]), and may be explained by some form of compartmentalization in acidic vesicles (see [38] for discussion).

We therefore assayed pellets of axenal cultures of slime moulds for InsP₆. The pellets were weighed before extraction, and the mass levels of InsP₆ measured gave an estimated concentration in the cells of 352±11 μM (Table 1). The figure of 594 μM [8] used an assumption of 0.4 g of water/g of pelleted cells, which corrects to 371 μM if we use the assumption of 0.64 g of water as determined above. The close agreement between the two sets of data provides an independent validation of the present assay, and confirms the really remarkably high InsP₆ levels in these organisms.

HeLa cells

We next assayed InsP₆ in confluent cultures of HeLa cells (Table 1). These have apparently higher levels of InsP₆ than rat tissues; however, the difficulty of accurately measuring cellular wet weight, plus the absence of connective tissue in a cell culture, limit the interpretability of such comparisons. Nevertheless, these experiments provided us with an opportunity to investigate the extent to which mammalian cell lines make all their own InsP₆. InsP₆ is reported to be present in human and rat serum or plasma [17–19], so it is possible that cultured cells derive some or even all of their InsP₆ from the calf serum in the medium in which they are grown. In this respect, cell cultures could be argued to be a paradigm for whole animal requirements, as they are with inositol (see [39] and the Discussion section). So we tried to measure InsP₆ in the calf serum in which the cells were cultured, but could detect none within the limits of the assay. The arguments for these limits are detailed below when we consider human serum, and, using this logic, we estimate that there could be no more than 16 pmol of InsP₆ in the medium of each 10-cm-diameter dish of cells. Yet a dish of HeLa cells grown to confluence in that medium contained 1.952±0.117 nmol of InsP₆. The same medium has been used in our laboratory to grow numerous other cell lines [e.g. HEK-293 cells (human embryonic kidney cells), COS cells and CHO cells (Chinese-hamster ovary cells)] for many passages, and so from these results we conclude that cultured mammalian cells make all their own InsP₆.

Human blood

The absence of detectable InsP₆ in calf serum raised the question of whether this is unique to cows, given that human and rat serum are reported to have levels of InsP₆ from 30–460 nM (e.g. [19,25] and see also the Discussion section). So we tested human blood. We fractionated the blood in two ways. As serum would be a direct comparison with the calf serum above, for some samples we let the blood clot and took the supernatant (serum) remaining. For comparison with measurements of plasma InsP₆, and also to gain some information about InsP₆ in some of the cells in blood, we also collected blood into citrate, and fractionated it into erythrocytes, white (buffy coat) cells and platelet-free plasma (we discarded the platelet fraction).

The results in Table 1 show that we could detect virtually no InsP₆ in erythrocytes (the small amount present is probably a result of white-cell contamination), which is in itself an interesting comparison with nucleated erythrocytes of amphibians and reptiles, which make high levels of Ins(1,3,4,5,6)P₅ [6,40]. For the white cell (buffy coat) fraction, we only used cellular protein as a standard, and the results suggest that, as one might expect from nucleated mammalian cells, human white cells contain InsP₆ at a concentration similar to the rat tissues assayed above (Table 1).

However, in both serum and plasma, we again found no significant quantities of InsP₆. To illustrate this point, Figure 2 shows typical HPLC profiles from assays on human serum samples which were (Figure 2C) or were not (Figure 2B) ‘spiked’ with 2.5 pmol of InsP₆ before extraction; this extra part of the protocol was introduced to ensure that there was no extraction problem with this particular type of preparation (which has no cells, but a high protein content). Similar experiments were performed on blood taken from three individuals and all gave similar results. From these profiles (Figure 2), it is difficult to be accurate as to how much InsP₆ we would have been able to detect. The background in the ³²P channel where [³H]InsP₇ elutes varied between experiments, in general this was lower if the [γ-³²P]ATP was fresh [27], as this ³²P background in the HPLC eluate is the ultimate limitation to the sensitivity of the assay. A further complication is that, in that background, there is frequently an apparent minor peak of ³²P that approximately co-chromatographs with [³H]InsP₇, which is present in both blank and human serum samples (Figures 2A and 2B). Nevertheless, if we suggest that 100 d.p.m. above backgound in triplicate samples would have shown up as a distinct and detectable ³²P peak co-incident with [³H]InsP₇ in Figure 2(B), we can calculate that this would represent no more than 64 fmol of InsP₆, which leads us to suggest that the plasma sample typified by Figure 2 probably contains less than 0.5 nM InsP₆. Moreover, during plasma preparation, some damage may have occurred to white cells (or platelets) to release traces of InsP₆, and, with this in mind, we suggest that in vivo blood probably contains no extracellular InsP₆ at all.

Figure 2. HPLC profile of a human plasma InsP₆ assay.

Three typical HPLC profiles from three assay tubes are illustrated, with the radioactivity in the InsP₇ region (indicated by the peak in [³H]InsP₇ elution) shown. Left-hand axis (♦), [³H]; right-hand axis (■), [³²P]. Note the differing scales on the [³²P] axes [dpm (d.p.m.)]. (A) Blank incubation (control), (B) extract of 100 μl of human plasma, and (C) extract of 100 μl of human plasma with 2.5 pmol (25 nM final concentration) InsP₆ added before extraction.

Human urine

The other human fluid that has been analysed for InsP₆ is urine, which is reported to contain 1–3 μM InsP₆ (e.g. [20] and see the Discussion section). We therefore measured InsP₆ in a urine sample from one of our blood donors, again spiking some samples before extraction as well as afterwards to confirm the recovery of InsP₆. The results revealed that there is no detectable InsP₆ present. There is more ‘interference’ with the assay from unknown factors present in urine than in plasma or serum, which made the assay less sensitive. But the assays included 50 μl of samples of urine spiked with 2.5 pmol InsP₆ (final concentration of 50 nM) before extraction, and these showed a clear peak of InsP₇ product; from a similar semi-quantitative logic to that stated above, we can set a lower limit of InsP₆ detection in the samples analysed of approx. 250 fmol InsP₆, the equivalent of 5 nM InsP₆ in human urine.

DISCUSSION

The levels of InsP₆ that we have measured in mammalian tissues are of a similar order to those deduced from equilibrium labelling of cultured cells with [³H]inositol (e.g. [37], and see [6] for discussion) and we can conclude that mammalian cells contain tens of μM InsP₆. If anything, these may be underestimates compromised by a potential non-linearity of the assay when performed on tissue extracts. Moreover, these calculations assume a uniform cytosolic localization, but obviously local concentrations may be higher. This may be most relevant in the context of well defined nuclear functions for this inositol phosphate and its more highly phosphorylated derivatives [11,41–43]. Slime moulds, which synthesize very high levels of InsP₆ very quickly [8], remain an enigma only deepened by our results.

Crucially, we have shown, as far as we know, for the first time, that HeLa cells in culture (and by implication, all mammalian cell lines in perpetuity) synthesize all their own InsP₆. This depends, of course, on the cells having sufficient inositol to make PtdIns(4,5)P₂ ([39] and also see the Introduction and Discussion sections), and we suggest that they probably serve as a paradigm for humans. The key to this observation is our discovery that cultured cells have no significant external source of InsP₆ because there is none detectable in the calf serum in which we grow them. This was the most unexpected finding of our study, and because of its implications for human diet and health, we went on to demonstrate, as thoroughly as the assay allows, that there is also no detectable InsP₆ in human serum, plasma (Figure 2) or urine.

These results have important implications for the apparent uptake of exogenous InsP₆ into cells or tissues, and for the proposed extracellular actions of InsP₆ in health and disease. To consider cells first, when radiolabelled InsP₆ is added at micromolar concentrations to mammalian cells in culture it has been reported to be taken up, but the radiolabel is then found as lower inositol phosphates or inositol [23,44]. InsP₆ binds strongly to Ca²⁺, and its solubility limit in the presence of millimolar levels of Ca²⁺ is less than 1 μM [36,38]. However, the (insoluble) Ca²⁺ salt of InsP₆ has been found to exist naturally, for example, as a component of the extracellular coat around the parasitic cestode Echinococcus granulosus [45] to which it can be transported as nanometre diameter granules [46]. From these observations, it seems that the most plausible mechanism for cellular InsP₆ uptake is endocytotic absorption (possibly as a Ca²⁺–InsP₆ precipitate), accompanied by dephosphorylation within the cell.

Consistent with this argument are results from the classic studies of Eagle et al. [39], in which mammalian cell lines in culture were shown to require approx. 0.3 μM inositol for growth, and this requirement could be fully substituted for by inositol monophosphates, or, with a much lower efficacy, by 1 μM InsP₆. In this context, we should note that adding InsP₆ at much higher (up to millimolar) concentrations to cell cultures will reduce extracellular Ca²⁺ to 1 μM or lower [36], so any subsequent cell death caused by such high concentrations (e.g. [21,22], and see [23] for a review) would be simply explained by the very well established requirement for extracellular Ca²⁺ in order to maintain mammalian cell function [47]; in short, the cells die because InsP₆ removes all the multivalent cations from the culture medium.

Our results have an impact most directly on the issue of the uptake of InsP₆ from the human diet. In a number of studies, Grases et al. [17–19] have reported that adding InsP₆ to the diet, or the topical application of InsP₆-containing creams [20], increases the levels of InsP₆ in plasma, urine and tissues. We have not explored different dietary regimes here, but we would argue that any such difference is not likely to be crucial. Typical levels of InsP₆ in urine from human subjects kept on an ‘InsP₆-free diet’ are 2–3 μM [20], and in plasma approx. 0.1 μM [19], whereas reverting to an ‘InsP₆-normal diet’ leads to approx. 4-fold increases in these parameters. Yet our estimates of maximal levels of InsP₆ in human subjects consuming a normal, carnivorous, cereal-containing diet are respectively (at least) 200 times (plasma) and 500 times (urine) lower than those reported for human subjects free of exogenous InsP₆ [17–19]. These maxima are only estimates, and perhaps a more direct demonstration of the inconsistency between results published previously [17–20 and ours lies the results for human serum in Figure 2(C) [and the corresponding results for urine (results not shown)], where spiking serum or urine with InsP₆ concentrations that are still respectively 4-fold and 40-fold lower than those claimed to be found in these fluids in InsP₆-deprived individuals [17–19] gives robust signals in our assay that are very much higher than those from the unspiked samples.

For these reasons, we suggest it is our assay that differs from other studies, not the diet of the donors. Experiments feeding radioactive InsP₆ to rats showed that it is dephosphorylated within the gut, and that the radiolabel is subsequently found in plasma and urine only as inositol and InsP [48], and it has been shown that microbes in the gut contribute the phytase activity responsible for this hydrolysis [49]. As the measurement of InsP₆ has so far been largely performed by (Dowex) ion-exchange chromatography (e.g. [18]) or HPLC [25] of acid extracts, followed by dephosphorylation of the fractions with phytase and then determination of myo-inositol (with a scyllo-inositol standard) by MS, it is conceivable that, as also suggested by Shears [7], the presence of high concentrations of anions (including multivalent anions such as ATP) in the tissue extracts might compromise inositol phosphate separation protocols characterized in their absence.

Thus although our assay is not designed for the quantification of lower inositol phosphates, we can combine our results with the literature published previously [48,49] to suggest that InsP₆ is not absorbed directly from the diet, but rather is dephosphorylated before it is taken up; note that the results of Wise and Gilburt [49] confirm that in the absence of any hydrolysis by gut flora, there is no significant InsP₆ absorption. The inositol and InsP generated by dephosphorylation will of course not be without effect: inositol is synthesized de novo by animals (from glucose 6-phosphate via Ins3P, see [6]), but most mammalian tissues do not produce enough of it, and so inositol is a vitamin (vitamin Bh) for which a significant dietary source is InsP₆ from cereals. Free levels of inositol are high in many mammalian cells (e.g. [50]), but the K_m for PtdIns synthetase is also high [51], so if raising levels of dietary InsP₆ leads to increases in the availability of inositol and InsP to tissues, a mass-action effect might be expected on PtdIns synthetase, perhaps leading to increases in steady-state inositol lipid and phosphate levels.

So, if the conclusion of the present study is that we (humans) do indeed make all our InsP₆ (by the route described in the Introduction), and that we derive none directly from our diet, what effect(s) will exogenous InsP₆, something claimed to be of significant medical benefit [23,24], cause? From the arguments in the preceding paragraph, increases may occur in cellular inositide levels (including InsP₆), but this effect would be indirect, and high dietary inositol or InsP would be expected to do the same thing. Any specific and direct actions of exogenous InsP₆ will be confined to its place of application. Such direct actions would stem from its ability to bind all multivalent cations with high affinity [36,38]. Binding Fe³⁺ gives InsP₆ unique and potent antioxidant properties [26,35], so it is likely that dietary InsP₆ will alter oxidative processes within the gut. This, and its general chelation of cations (e.g. Fe³⁺, Zn²⁺ and Ca²⁺) to restrict their uptake, probably underlie the reported effects of InsP₆ in vivo (e.g. [23,24]). These ion-binding actions of InsP₆ can have undesirable consequences, either environmental, such as depleting Fe³⁺ or Zn²⁺ from some soils [52], or physiological, by InsP₆ removing these same essential cations from the human diet to a degree that can be deleterious, depending on the nutritional status of the individual [53–56].

However, a major thrust of many of the arguments for the benefits of InsP₆ to humans [23,24] stems from a combination of two factors: (a) the killing of tumour-cell cultures by applying to them high extracellular levels of InsP₆, which, we argue above, is of questionable physiological relevance; and (b) an assumption that exogenous InsP₆ is taken up directly to form a significant extracellular InsP₆ pool, which bathes some or all our cells and which, because of its unique chemistry, can in turn influence their physiology or pathology. It is this latter assumption that the results in the present study suggest requires reconsideration.

FUNDING

This work was supported by a Programme Grant from the Wellcome Trust, and by the Royal Society.