Acute infectious mononucleosis
During acute infectious mononucleosis (AIM), large clones of Epstein-Barr virus-specific T lymphocytes are produced. To investigate the dynamics of clonal expansion, we measured cell proliferation during AIM using deuterated glucose to label DNA of dividing cells in vivo, analyzing cells according to CD4, CD8 and CD45 phenotype. The proportion of labeled CD8+CD45R0+ T lymphocytes was dramatically increased in AIM subjects compared to controls (mean 17.5 versus 2.8%/day; p<0.005), indicating very rapid proliferation. Labeling was also increased in CD4+CD45R0+ cells (7.1 versus 2.1%/day; p<0.01), but less so in CD45RA+ cells. Mathematical modeling, accounting for death of labeled cells and changing pool sizes, gave estimated proliferation rates in CD8+CD45R0+ cells of 11–130% of cells proliferating per day (mean 47%/day), equivalent to a doubling time of 1.5 days and an appearance rate in blood of about 5×109 cells/day (versus 7×107 cells/day in controls). Very rapid death rates were also observed amongst labeled cells (range 28–124, mean 57%/day),indicating very short survival times in the circulation. Thus, we have shown direct evidence for massive proliferation of CD8+CD45R0+ T lymphocytes in AIM and demonstrated that rapid cell division continues concurrently with greatly accelerated rates of cell disappearance.
Infection with Epstein-Barr virus (EBV) during adolescence and adulthood causes acute infectious mononucleosis (AIM), an acute systemic febrile illness. The primary cellular target for infection is the B lymphocyte, but powerful T cell responses are elicited as demonstrated by massive increases in circulating activated CD8+ T cells. Such T cell responses are thought to be crucial in suppressing viral replication to low levels such that individuals remain healthy thereafter, despite life-long persistence of virus. A minority suffers from a more chronic viral illness with persisting higher level virus replication. Early events in the response to virus infections may be critical in determining long term outcome 1, 2. AIM may thereforebe a useful model to understand the regulation of clonal proliferation and how the immune system either succeeds or fails to suppress virus production adequately.
CD8 expansion in AIM is characterized by major expansions of a relatively small number of clones that are readily detectable ex vivo, while CD4 clones are not. Other evidence suggeststhat CD4 responses are spread across a broader range of smaller clones 3, 4. Recent studies using EBV-specific tetramers and heteroduplex analysis of T cell receptors have shown that a high proportion of the activated CD8 cells are specific for EBV antigens and that the response is focussed on a small number of dominant epitopes 5. The increase in cell numbers in peripheral blood thus reflects the presence of large numbers of virus-specific T cells that have undergone dramatic expansion from rare precursors in the naive T cell pool.With resolution of the acute infection, cell numbers decrease rapidly, typically returning to resting values, but with preservation of a pool of antigen-specific primed lymphocytes 6.
Such a dynamic response must occur within the context of the normal homeostatic regulation of the T cell pool; in healthy individuals, both the number of peripheral T cells and the proportionsthat are naive and primed remain relatively constant over time 7. In addition, animal model experiments using cell surface markers to distinguish "naive-phenotype" from "memory-phenotype" T cells have shown that cell numbers in these two subpopulations are maintained by different mechanisms 8. Thus naive T cells appear to persist mainly as long-lived cells that divide infrequently, while memory-phenotype T cells undergo cell division more often, presumably balanced by higher rates of cell death 9–11.
Alterations in T cell homeostasis which occur in acute viral infection are the consequence of changes in both production and death rates; thus a clone may become enlarged by increased proliferation or increased longevity. Full understanding of the magnitude of such changes and the relative importance of increased proliferation and altered longevity for acute clonal enlargement and long-term persistence hinge on methods allowing direct measurement of the underlying cellular kinetics. Such methodologies using stable (non-radioactive) isotope labeling of dividing cells applicable to human in vivo studies have recently been developed 12 and applied to study lymphocyte homeostasis during chronic human immunodeficiency virus (HIV) infection 10, 13, 14. We have applied such methodologies to investigate the kinetics of lymphocytes in vivo during AIM in human volunteers.
2.1 Subjects and clinical observations
Five previously healthy subjects with AIM were investigated. All had positive monospot, EBV-specific IgM and fever, pharyngitis and fatigue. Clinical details are shown in Table 1, including symptom severity graded as: none, mild, moderate, or severe for local and systemic features. Investigation was performed 4–14 days from onset of symptoms and 0–7 days from maximal symptoms, as judged by the subject. All made a prompt complete clinical recovery subsequent to investigation. Control subjects (n=8) were healthy young volunteers of similar median age (23 years; 4M:4F); control data have been fully presented elsewhere 11.
|Subject||Age||Gender||Duration of symptomsa)||Time from maximal symptomsa)||Local disease (pharyngitis)||Systemic disease||Lymphocyte count|
2.2 Lymphocyte phenotype
Examination of peripheral blood lymphocyte phenotype revealed an elevated lymphocyte count in four out of five subjects (Table 1) and an increased percentage of CD8+ T cells relative to total lymphocytes in all subjects. Such cells were predominantly CD45R0+, as previously described 15, and most demonstrated an activated phenotype. On average, 48.7% (range 39–89%, n=5) and 64.5% (44–81%, n=4) of CD8+ cells were positive for HLA-DR or CD38, respectively at first measurement in AIM subjects (Fig. 1), whereas only 17 and 19%, respectively of cells were positive in controls. The percentage of HLA-DR+ and CD38+ cells declined during the course of the study in all AIM subjects, commensurate with clinical recovery from AIM.
Subject 1 was unusual, in that he was found to be heterozygous for the CD45 C77G allelic variant known to cause altered splicing of CD45 16. This mutation, found in approximately 1:60 of the UK population, results in an inability to generate low molecular weight isoforms of the CD45 molecule encoded by that allele 17. Hence, subject 1 had no single-positive CD45R0+RA– cells (all CD45R0+ cells were also CD45RA+). It is interesting that this individual exhibited the most severe symptoms in the cohort since an increased frequency of the C77G allele has been found to be associated with susceptibility to other viral infections such as HIV 18.
2.3 CD8+ T cell kinetics in AIM
Cell kinetics were calculated from the quantitative incorporation of deuterium from deuterated glucose into DNA of dividing cells following a 24-h infusion of deuterium-labeled glucose 12. Acquisition and loss of label by peripheral blood T cells are shown in Fig. 2. When CD8+ T cells were analyzed (Fig. 2a), there were three notable findings. First, there was a dramatic increase in peak labeling in AIM subjects compared to controls, consistent with very rapid cellular proliferation in AIM. Although there was a high degree of variability between subjects with AIM, mean peak proportion of labeled CD8+CD45R0+ cells (17.5±7%) was much higher than that in control subjects (2.8 5±2.1%, p<0.01; Table 2). The highest value was seen in subject 1 where nearly 30% of CD8+CD45R0+ cells (which were also CD45RA+) were labeled on day 3. Such dramatically increased labeling occurred in the context of an expanded CD8+ T lymphocyte pool, indicating greatly increased total rates of CD8+ T lymphocyte production, as discussed below.
Secondly, peak labeling appeared to occur earlier in AIM subjects. In control subjects, peak labeling was observed with equal frequency at days 3 and 4 with this protocol, whereas, in AIM subjects, day 3 values always exceeded those on day 4; indeed the peak may have preceded the first sampling on day 3. This suggests that the normal post-mitotic residence time of lymphocytes (presumably in lymph nodes) prior to release into peripheral blood, is reduced in AIM.
Thirdly, labeled cells in the CD8+CD45R0+ pool disappeared very rapidly from the circulation. Thus, in all subjects there was a considerable fall in labeling from day 3 to 4 and by day 10 labeling was about a quarter of peak values (Fig. 2a).
When CD8+CD45RA+ cells were examined, increased labeling was noted in two subjects, while, in two others, values were very close to controls (Fig. 2b). In subject 1, labeling was very high but this was attributed to the splicing variant in this individual (Fig. 2b inset). The mean value, excluding subject 1, was increased compared to control values (Table 2; p<0.05) but the magnitude was considerably less than that seen in CD8+CD45R0+ cells. Thus, a massive alteration in the kinetic behavior of the CD8+CD45R0+ T cell pool was associated with relatively minor changes in the CD45RA+ pool in most subjects, consistent with independent regulation of these T cell sub-populations.
|% labeled cells||% labeled cells||% labeled cells||% labeled cells|
2.4 CD4 lymphocyte proliferation
When CD4+ lymphocytes were analyzed, AIM was found to be associated with a substantial increase in the proportion of labeled CD45R0+ memory-phenotype cells in peripheral blood at day 3 (Fig. 2c), with peak values about three times higher than controls; 7.1% versus 2.1% (Table 2, p<0.01). As for CD8+ cells, there appeared to be more rapid release of labeled CD45R0+ cells into the circulation. Such results reveal a hitherto unrecognized sizeable alteration in CD4+CD45R0+ T cell kinetics during AIM. By contrast, for CD4+CD45RA+ cells, we found similar rates of labeling in AIM subjects and controls, except for subject 1 with the splicing variant (Table 2; Fig. 2c).
2.5 Modeling proliferation and death from cell number, label incorporation, and loss
When lymphocyte pool sizes in peripheral blood were estimated from total lymphocyte numbers and relative proportions in PBMC, a marked elevation in the absolute number of CD8+CD45R0+ cells was initially observed, with values falling over the following 3 weeks in all subjects (Fig. 3). Other cell types appeared relatively stable, although in some subjects there was a concomitant fall in CD8+CD45RA+ and CD4+CD45R0+ cells and, in one subject (no. 3), a rise in CD4+CD45RA+ cell numbers (Fig. 3).
Applying the model described below to fit labeling and lymphocyte number data yielded the three parameters summarized in Table 3: the average proliferation rate of the whole cell population (p), its average death rate (D) and the death rate specifically of labeled cells (d*) 11. Predicted curve fits matched observed data closely (Fig. 4) with a predicted peak in labeling at the end of day 1. For subject 1, only disappearance curve fits and a value for d* were calculated because of missing initial pool size data. In subject 4, peak height was strongly influenced by the rapid fall in labeling between days 3 and 4, yielding high standard deviations for parameter estimates; this rapid fall suggests very high death rates of labeled cells in this subject.
|t1 2 or t2||1.5||0.9||1.0||31.5||2.1||2.4|
|t1 2 or t2||13.6||7.9||13.6||154||5.9||154|
|t1 2 or t2||2.9||1.6||2.6||4.5||1.1||4.8|
|t1 2 or t2||26.2||9.4||26.2||118||9.6||118|
2.6 Modeling CD8+ T cell kinetics
Modeled proliferation rates for CD8+CD45R0+ cells in all AIM subjects were much higher than average control values with a mean value for CD8+CD45R0+ proliferation rate in AIM subjects nearly ten times mean control value (46.8%/day versus 5.1%/day, p<0.02; Table 3) and equivalent to an average doubling time of 1.5 days (versus 14 days in controls, p<0.02) 11.
Interestingly, such increased proliferation rates were found despite the fact that in every subject the CD8+ pool was contracting during the observation phase with clinical resolution of AIM. Absolute appearance rates of CD8+CD45R0+ cells in blood, based on model estimates for proliferation rate, measured cell numbers and predicted blood volumes ranged from 1.7×108 to 1.9×1010 (mean 5×109) cells/day in AIM subjects, compared with the mean value for controls, 7×106 cells/day. Thus, these results directly show massive proliferation of CD8+CD45R0+ T cells, even during resolution of acute EBV infection, with a high proportion of this pool replaced daily from rapidly dividing precursors. Interestingly, CD8+ T cell proliferation rates appeared to correlate with crude clinical assessments of disease severity.
The two measures of cell disappearance, the mean rate for labeled cells (d*), and the overall rate for the whole pool, including both labeled and unlabeled cells (D), were found to be markedly elevated for CD8+CD45R0+ cells in AIM subjects (Table 3). Thus, labeled CD8+CD45R0+ cells disappeared at a rate of over 80%/day in AIM subjects, compared to 8.8%/day in controls, while overall corresponding disappearance rates (D) for CD8+CD45R0+ cells were 71.3%/day and 5.1%/day, respectively (Table 3, p<0.01 for both d* and D). Such values are particularly striking when it is considered that a mean rate of disappearance of 70%/day is equivalent to a half-life of less than 1 day. The average death rate of the whole population (D) and the rate of loss of labeled cells (d*) were closely correlated (r>0.95; p<0.01 by Spearman rank, 2-tailed test), suggesting that the contraction of the population is largely attributable to the death of proliferating cells. The rapid death rate of unlabeled cells probably represents death of proliferating cells which are unlabeled because they divided before or after the 24-h labeling period rather than contraction of non-proliferating pools.
2.7 Modeling CD4+ T cell kinetics
Proliferation rates of CD4+CD45R0+ cells were also markedly elevated compared to control subjects; 23.7%/ day versus 2.7%/day, respectively (Table 3, p<0.02). When increased pool size was taken into account, this constituted an absolute increase in CD4+CD45R0+ T cell production of about fifty times control values (1.1×109 versus 2.2×107 cells per day in blood; p< 0.01). Disappearance rates of both labeled (d*) and unlabeled cells (D) were also both elevated compared to control subject values (Table 3, p<0.02). Thus modeling confirms a dramatic increase in both proliferation and disappearance of CD4+CD45R0+ cells in AIM.
2.8 Modeling kinetics of CD45RA+ T cells
Fitting CD45RA+ data for both CD4+ and CD8+ cells was difficult because of the low enrichment values seen in these slow turnover populations. Values obtained are shown in Table 3 but it is difficult to draw conclusions because of the small number of values obtained, the low levels of enrichment and the need to exclude data from subject 1 because of his splicing variation. However, it can be seen that proliferation rates of CD45RA+ cells fell below those of their CD45R0+ counterparts in all subjects. Where significant numbers of CD45RA+ cells were labeled, death rates of labeled cells did appear to be rapid suggesting that there may be a small subpopulation of CD45RA+ responding to AIM by proliferation and rapid death. For CD4+CD45RA+ cells, d* always exceeded D, suggesting that such subset proliferation may occur within the context of a larger pool of cells which retain their slower proliferation and death rates despite active AIM and a massive ongoing response in the CD45R0+ compartment.
This study has quantified the proliferative response of lymphocytes to acute EBV infection in man. The implications for turnover in the activated CD8+ T cell pool are striking: in the five subjects studied, between 11 and 30% of CD8+CD45R0+ cells were labeled at day 3 after only a single days labeling. Modeling such labeling, accounting for changes in pool size and loss of labeled cells, gave estimates of proliferation of between 11 and 130%/day equivalent to doubling times for the whole activated CD8+ population of about 1–2 days. The absolute appearance rate of new cells in peripheral blood was increased several hundred-fold compared to healthy controls. The extent of such clonal proliferation in AIM refers to proliferation within the entire CD8+CD45R0+ T cell pool. Although it is possible that some "bystander" activation of memory cells may have occurred 19, it is likely that the non-EBV-specific fraction is proliferating much more slowly than EBV-specific cells. The actual value for EBV-specific cells may therefore be even higher than the values above.
In addition we have demonstrated very rapid loss of labeled CD8+CD45R0+ cells from the circulation with a half-life of only 1–2 days. Loss of labeled cells may result either from death or disappearance of activated effectors from the circulation, for example by migration into inflamed tissues. The lack of reappearance of label at later time points suggest that such loss is irreversible and we suggest that it represents death of activated cells, either in the circulation or in situ in tissues, after a very short lifespan. This suggestion is consistent with the finding that pro-apoptotic factors such as CD95 are up-regulated and the anti-apoptotic molecule bcl-2 is down-regulated in CD8 lymphocytes in AIM 20. Furthermore, CD8+ T cell clones may depend upon cytokine stimulation for their survival in vivo and rapid turnover may drive some cells into cytokine starvation. Indeed, there is evidence that EBV-specific clones may be rescued from apoptosis ex vivo by cytokines 20. Death of responding cells might represent a mechanism for immune evasion, although it has been demonstrated that most clones are not exhausted and can be found months after resolution of the initial infection 3.
Thus it appears that the great majority of activated cells are culled within a few days of appearance, at least at or shortly after the peak of the immune response. This raises some interesting questions. First, why does elevated cellular proliferation persist in the context of resolution of AIM as total lymphocyte counts are decreasing? Does this seemingly inefficient and energy consuming process occur during other infections? The only other human studies of in vivo proliferation and death of lymphocytes are those performed in HIV infection; such studies have shown parallel increases in proliferation and death rates during chronic infection 14, 21, findings which are qualitatively similar, albeit quantitatively much smaller, than those described in this study in an acute viral infection. Secondly, how is a small but critically important portion of the activated EBV-specific population of cells retained as a memory population? Our observations of the highly dynamic nature of proliferation and death of CD8+ T cells in AIM suggest that very small changes in the proportion of cells which die will make a very large difference to the size of the surviving clonal pool.
CD8+ T cell clones in acute EBV infection may reach between 108 and 109 cells, according to studies of V-beta utilization 3. The size of such clones before infection is not known but, from numbers of T cell receptors and numbers of lymphocytes in humans, is likely to be of the order of 10422. To reach a clone size of 108 would therefore require a minimum of the order of 15 divisions, assuming all progeny cells survive. However, the data presented here suggest that proliferation is accompanied by substantial cell death, implying that considerably more rounds of division may be required. In view of this extensive proliferation, the up-regulation of telomerase that has been shown to occur during clonal expansion in AIM may be crucial in preventing clonal exhaustion during the acute response 23.
Although the CD8+ response is dominant in AIM, we have shown that there is considerable proliferation in the CD4+CD45R0+ population. This accords with the presence of high IgG titers to viral antigens consistent with active helper T cell involvement in acute EBV infection. Nevertheless, our failure to demonstrate CD4 clones ex vivo in AIM 3 and the relatively modest proliferation detected here, suggest that clonal expansion among CD4+ is much less than among CD8+ cells 24.
Similarly, the absence of substantial proliferation among CD45RA+ cells in AIM would argue that the CD45R0+ and CD45RA+ pools are regulated independently, a view supported by murine data 25. Our data also imply that even under conditions of massive clonal expansion, bystander effects on CD45RA+ (naive) cells are minimal, consistent with murine data showing that administration of cytokine inducers or cytokines to mice induces proliferation of memory/effector phenotype cells with little effect on naive cells 19, 26, 27. The rate of proliferation of CD45RA+ cells might be underestimated if they rapidly converted to a CD4R0+ phenotype on activation 28 but this is not the case for T cells stimulated with certain combinations of cytokines 29. Primed CD8 T cells in the CD45RA+ subset 3, 4, 30 may contribute to the increased proliferation rates seen in this population in some AIM subjects.
Our data are limited by our ability to sample only peripheral blood. Although studies in sheep suggest relatively homogeneous and rapid mixing of different lymphoid compartments 31, this may not be the case for T cells entering non-lymphoid tissues 32. Specifically, it is possible that EBV-specific T cells that were previously trapped in lymph node are released into the circulation as AIM resolves.
In conclusion, we have directly measured very rapid proliferation of lymphocytes in vivo in humans during an acute infection, with appearance rates of CD8+CD45R0+ lymphocytes in blood about 500 times normal. Decay kinetics indicate that cell death is also accelerated implying that such extensive clonal expansion must involve many cell divisions. These data indicate that clonal expansion in AIM is a balance between increased rates of cell division and cell death. Interventions that could affect either of these parameters may have dramatic effects on immune memory and this has implications for the efficacy of vaccination procedures.
4 Materials and methods
4.1 Subjects and measurement of cell turnover
All subjects gave written informed consent. The protocol was approved by the local research ethics committee. Labeling consisted of a primed 24-h intravenous infusion of 1 g/kg 6,6–2H2-glucose (Cambridge Isotopes, MA) during which diet was restricted to small, low-energy meals. Blood for measurement of plasma glucose deuterium (2H) enrichment was taken at 1 h and approximately 4-hourly thereafter. Blood samples for estimation of deuterium enrichment in DNA were taken 3, 4, 10 and 21 days after infusion; in subjects not available on day 21, blood was taken on the closest day possible.
4.2 Cell sorting
PBMC were isolated from 50 ml of heparinized fresh blood samples by Ficoll-Paque (Pharmacia, St Albans, GB) density gradient centrifugation. Cells (1×107/ml in PBSA+0.2% BSA) were stained with CD3-RPE (Serotec Ltd, Oxford, GB) for 30 min on ice, then sorted into CD3+ and CD3– fractions using a MoFlo cytometer (Cytomation, Colorado). CD3+ cells were further stained with biotinylated anti-CD8 (Serotec) antibody + streptavidin-allophycocyanin (Sav-APC, PharMingen, San Diego, CA) together with CD45RA-RPE-CY5 (Serotec) and sorted into CD8+CD45RA+, CD8–CD45RA+, CD8+CD45RA– and CD8–CD45RA– subsets.
4.3 Immunofluorescence staining
PBMC (2×105), or sorted CD3+ cells were stained with CD8-APC (PharMingen), CD4-RPE-CY5 (Serotec) and either HLA-DR-FITC (Caltag, Burlingame, CA) or CD38-FITC (PharMingen) in a single step at 4°C for 30 min. Isotype-matched mAb were used as controls. In total 10,000 events were collected on a FACSCalibur flow cytometer (Becton Dickinson, San Jose, CA) and analyzed using CellQuest software.
4.4 Analysis of deuterium enrichment
Enrichment of deuterium in DNA was assayed essentially as described 12, 33. DNA from sorted subsets was extracted and digested enzymatically to deoxynucleosides. Deoxyadenosine, purified by C-18 SPE column chromatography, was converted to its aldononitrile acetate derivative by reaction with hydroxylamine/pyridine (1% w/v, 100°C, 45 min) and acetic anhydride (room temperature, 30 min) and analyzed by GCMS, monitoring ions m/z 198 and 200 by PCI in SIM (HP-225 column, HP 6890/5973 GCMS; Hewlett Packard, Bracknell, GB) 33.Abundance-matched samples were analyzed in triplicate alongside a standard curve derivatized concurrently. Plasma glucose enrichment was measured using the same derivatization (m/z 328 and 330). Typical precision of reproducibility of the M+2/M+0 ratio was ± 0.02%.
Results were expressed as the fraction of labeled cells (F) present on each day, given by the ratio of the enrichment of label in DNA (E) and the precursor enrichment, b (mean glucose enrichment ×0.65) 12. The magnitude of the peak value for F represents a crude measure of the cellular proliferation rate. Because this estimate does not take into account either cell death between the end of labeling and sampling on day 3 or changes in pool sizes, appearance and disappearance of labeled cells were modeled as described 11, 34, with a modification to allow for changing lymphocyte subpopulation pool sizes.
The observed proportion of labeled cells at a given time will depend, first, upon the rate of proliferation during labeling and subsequent loss of labeled cell loss, and, secondly, upon the expansion and contraction of each lymphocyte pool. The rate of change of deoxyadenosine (A) and labeled deoxyadenosine (A*) is given by:
during the whole experiment
during the labeling period
after the labeling period where p is the average proliferation rate of the whole population of interest, D the average death rate of the whole population, d* the rate of loss of labeled cells and b the precursor enrichment 11. These equations can be solved to find the amount of deoxyadenosine (A) which is proportional to the number of cells and the fraction of labeled cells F=A*/Ab
during the labeling period
after the labeling period
For a population of constant size the average proliferation rate of the whole population equals the average loss rate of the whole population (p=D) and the equations above reduce to the form in Asquith et al. 34. These equations were fit simultaneously to the lymphocyte count (scaled to avoid a disproportionate effect) and labeling data to estimate the three parametersp, d* and D.
An alternative, related model was also considered in which the assumption that the change in lymphocyte count could be described by simple exponential growth was removed. Labeling data were normalized using lymphocyte counts to allow for (1) the change in population that can potentially be labeled during the labeling phase and (2) the change in the population size post-labeling (which has the effect of concentrating or diluting the label depending on whether the population is contracting or expanding). In theory, the alternative model based on normalization is better because it does not assume simple exponential growth. When results from the two models were compared, very similar parameter estimates of p and d* were found. We therefore present results derived from the model assuming exponential decay, as we feel it is intuitively more obvious.
Where proliferation is expressed as doubling time or disappearance as half-life, these were calculated as ln2/p and ln2/d, respectively. Data are expressed as mean ± 1 SD; standard deviation of the parameter estimates was approximated using the asymptotic covariance matrix method (shown to be more efficient than the bootstrap method for this model in a large number of numerical simulations). Comparisons between groups were made by Mann-Whitney U-test (2-tailed).
We wish to thank Nadine Clifford and Elka Giemza for technical and nursing assistance in performing the studies and thank the subjects who willingly took part. This work was supported by the Edward Jenner Institute for Vaccine Research (publication number 54). DM was supported by a Fellowship from Serono International S.A., and a Medical Research Council Glaxo-Wellcome Clinician Scientist Fellowship. BA was supported by the Wellcome Trust (ref: 054451).