Can peripheral blood γδ T cells predict osteonecrosis of the jaw? An immunological perspective on the adverse drug effects of aminobisphosphonate therapy


  • For a Commentary on this article, please see Weitzmann (J Bone Miner Res. 2013;28:723–727. DOI: 10.1002/jbmr.1886).


Nitrogen-bisphosphonates (n-BP), often referred to as aminobisphosphonates, are the most commonly prescribed drugs for the treatment of disorders of bone fragility. However, long-term continuous treatment predisposes certain individuals to serious rare side effects, such as bisphosphonate-associated osteonecrosis of the jaw (BAONJ). n-BP use is known to unintentionally activate a subset of innate T cells called Vγ9Vδ2 T cells, but the consequence of this chronic immune stimulation has remained unexplored. The primary objectives of this study were to 1) determine the fate of Vγ9Vδ2 T cells in osteoporotic patients on n-BP therapy as a function of time and type of therapy; 2) evaluate the proportion of Vγ9Vδ2 T cells in patients who had recently experienced n-BP–associated ONJ. We found there is a notable loss of Vγ9Vδ2 T cells over time in osteoporotic patients on n-BP therapy, particularly those on intravenous (iv) therapy (Spearman r = −0.55, p < 0.0001 iv; r = −0.3, p < 0.03 oral) (n = 68); no difference was observed in total T cells, monocytes, or granulocytes. Importantly, the observed negative effect on Vγ9Vδ2 T cells coincides with the reported route of administration and timing of the rare occurrence of BAONJ. Patients (n = 6) who had experienced BAONJ were all found to be significantly deficient in Vγ9Vδ2 T cells (median = 0.07%) in comparison to age- and sex-matched treatment-naïve controls (N = 11; median = 2.40%), U = 0, p = 0.001; this was the only consistent difference in the leukocytes assessed. All BAONJ cases had an underlying condition that further contributed to impaired immunity. We propose Vγ9Vδ2 T cells show a strong potential to serve as harbingers of possible adverse immune effects of n-BP therapy, particularly in those patients already having a compromised immune system as they may be most vulnerable to the development of conditions such as BAONJ. © 2013 American Society for Bone and Mineral Research.


Aminobisphosphonates (n-BP) are the most widely prescribed class of drugs for the treatment of pathological disorders of bone fragility.1 n-BP impair osteoclastic bone resorption via inhibition of farnesyl pyrophosphate synthase (FPPS), a key enzyme in the mevalonate pathway for isoprenoid synthesis.2 The natural substrate of FPPS, isopentenylpyrophosphate (IPP), is an endogenous antigen recognized by an important subset of innate human T cells that bear the Vγ9Vδ2 T cell receptor.3 These cells typically comprise 1% to 10 % of circulating T cells and respond to stress-induced molecules that are nonpeptidic in nature and preclude the requirement for processing or presentation by classical major histocompatibility complex (MHC) molecules.3

A significant side effect of n-BP blocking FPPS function is the accumulation of cellular IPP, which leads to the unintentional chronic stimulation of Vγ9Vδ2 T cells. It was first observed that the acute phase response experienced by some initiating intravenous (iv) n-BP therapy was directly linked to the activation of Vγ9Vδ2 T cells and the release of pyrogenic cytokines.4

Bisphosphonate-associated osteonecrosis of the jaw (BAONJ) is a serious rare adverse drug effect characterized as a lesion of exposed necrotic bone in the maxillofacial region that takes >8 weeks to heal.5 The exact mechanism leading to BAONJ is unknown, and no biomarkers exist to identify those at risk. The most prevalent theory is that it is induced by direct drug toxicity on bone and soft tissue.6 Proposed risk factors include dental treatment, periodontal disease, smoking, cancer, chemotherapy, glucocorticoid use, and diabetes,7 and it is frequently associated with microbial biofilms.8

We hypothesized that BAONJ is a consequence of drug-induced immune dysfunction and will be associated with a loss of Vγ9Vδ2 T cells via activation-induced cell death (AICD). The primary objectives were to 1) determine the fate of Vγ9Vδ2 T cells in osteoporotic patients on n-BP therapy as a function of time and type of therapy; and 2) evaluate the proportion of Vγ9Vδ2 T cells in patients who had recently experienced BAONJ.

Materials and Methods

Study subjects

Ethics to conduct research on human subjects was obtained through the Clinical Ethics Board of the University Clinic of Schleswig-Holstein (UK-SH), Kiel Campus, and adhered to the tenets held by the World Medical Association Declaration of Helsinki–Ethical Principles for Medical Research Involving Human Subjects. Potential study participants with osteoporosis who were either n-BP–treatment naive or who were currently on n-BP therapy (either oral or intravenous) were identified and approached to participate in the study through a network of community physicians in Kiel, Germany. Exclusion criteria included use of strong immuno-modulatory drugs such as systemic corticosteroids or other immunosuppressive agents, and the presence of any malignancy. Enrollment was limited to postmenopausal women or men with age-related osteoporosis to reduce variability in immune parameters owing to both age and the action of ovarian hormones on lymphocyte distribution. The study design was cross-sectional except for those who were just starting treatment or were changing treatment regimens from oral to iv administration and who agreed to come in again for sampling after starting therapy to determine the immediate consequence of n-BP therapy on peripheral blood immune cells over the short course. The study was strictly observational and the investigators had no role in regard to the treatment regimens of the patients enrolled in the study, which were solely based on the judgments of the primary physician and the patients themselves. Patients who had recently experienced BAONJ were contacted through the Department of Oral and Maxillofacial Surgery at UK-SH. For the BAONJ cohort, no exclusion criteria for malignancy or drug use were in place because we wanted a representative of those who develop the condition including their associated risk factors. Additional control subjects age-matched for the ONJ cohort were recruited through physician offices and the Department of Immunology (UK-SH). All study subjects were given an explanation of the objective of the study and gave informed consent. A medical questionnaire and blood sample collected in EDTA tubes was obtained from each participant. Blood samples were processed for flow cytometric analysis within 8 hours of collection.

Flow cytometric analysis for peripheral blood Vγ9Vδ2T cell and leukocyte distribution

Whole blood was incubated in V-bottom microtitre plates (Nerbe Plus GmbH, Winsen/Luhe, Germany) with the following fluorescently conjugated antibodies: mouse anti-human Vδ2-fluorescein isothiocyanate (FITC) (clone IMMU389, Beckman Coulter GmbH, Krefeld, Germany), mouse anti-human CD3-phycoerythrin (PE) (clone SK7); mouse anti-human CD14-FITC (clone MoP9), mouse anti-human CD16-PE (clone B73.1); and mouse IgG1-FITC (clone X40) in conjunction with mouse IgG2a-PE (clone X39) served as isotype controls (all from BD Pharmingen, Heidelberg, Germany). After incubation with the fluorescent-conjugated antibodies, red blood cells (RBC) were lysed using BD FACS Lysing Solution (BD Biosciences, Heidelberg, Germany) according to manufacturer's protocol and subsequently washed once in phosphate-buffered saline (PBS) buffer containing 1% bovine serum albumin (BSA) before another brief RBC lysis step using 50 µL/well (<5 minutes incubation at room temperature). Cells were subsequently washed two more times and resuspended in 1% paraformaldehyde (PFA) buffer and analyzed by flow cytometry on a FACSCalibur (BD Biosciences) machine equipped with CellQuest Software (CellQuest, Tampa, FL, USA).

Statistical analysis

Because this is the first study to evaluate changes in Vγ9Vδ2 T cell distribution as a consequence of n-BP therapy for osteoporosis, we aimed to enroll a minimum of 60 participants with osteoporosis equally distributed into the following cohorts: 1) patients with osteoporosis who had not yet initiated n-BP therapy or were on alternative nonhormonal treatment; 2) patients with osteoporosis on oral n-BP therapy; and 3) patients with osteoporosis who were on intravenous (iv) n-BP therapy. We anticipated this would give us 80% power to see a 60% difference in peripheral Vγ9Vδ2 T-cell numbers between groups. Based on our own experience regarding the high interindividual variance of Vγ9Vδ2 T cells in circulation (which typically make up 1% to 10% of the peripheral T-cell pool), a real significant clinical effect of the drug on Vγ9Vδ2 T-cell numbers and function would likely have to be observed at this higher threshold. The proportion of Vγ9Vδ2 T cells present was calculated as a fraction of the peripheral T-cell pool present in the white blood cell population of whole blood to account for the high interindividual variability in leukocyte distribution and density in whole blood. Nonparametric tests (Kruskal-Wallis and Mann-Whitney) were used to compare the distribution of immune cells between groups, and Spearman ranked correlation was used to determine associations between immune parameters measured and length of time on therapy. All tests were two-tailed with α set at 0.05.


Osteoporosis cohort included in the study

A total of 70 samples were obtained from patients with osteoporosis, and 68 were included in the analysis. Samples from two patients who were in the treatment naive group were subsequently excluded because one was later found to have been on oral bisphosphonate for 14 months and had stopped treatment 2 years before the study and one was on methotrexate, a strong immunosuppressant, for rheumatoid arthritis. The vast majority of the participants were women (97%). The age and length of time on therapy for each group are shown in Table 1. There was no difference between groups in age F(2, 65) = 0.29, p = n.s. (one-way ANOVA). The most frequent oral drug used was alendronate (70 mg/week), which was taken by 88%, and oral ibandronate (150 mg/month) was taken by the remaining 12%. All those on iv therapy for osteoporosis were on ibandronate (3 mg/3 months). Of these, 9 (36%) had switched from oral medication at one point with four of these changes occurring within the course of the study, so we had the opportunity to get a follow-up sample to determine if any immediate changes occurred in leukocyte distribution.

Table 1. Age, Route of Administration, and Length of Time of Aminobisphosphonate Therapy of Osteoporotic Patients Included in the Study
Osteoporosis patientsAge (years)Length of time on therapy (days)
TreatmentNMean (±SD)(95% CI)Mean (±SD)Range
  • CI = confidence interval; n/a = not available.

  • a

    Patients diagnosed with osteoporosis who had not yet initiated n-BP or were on alternative treatment.

  • b

    Nine patients had originally started on oral therapy and had been switched to iv therapy; four of these switches occurred during the time of the study that allowed assessment of immediate changes in the leukocyte repertoire in response to treatment over a short period of time (Supplemental Fig. S1).

Controlsa1872.9 (4.7)(70.6, 75.3)n/a
Oral n-BP2572.9 (5.5)(73.1, 75.2)456 (567)(5 to 1769)
Iv n-BPb2573.9 (5.4)(72.2, 75.2)526 (517)(7 to 1989)

Phenotypic characterization of immune cells as a function of time and type of n-BP therapy

There was a significant difference in the proportion of γδ T cells present between the three cohorts (Kruskal-Wallis test statistic = 10.92, p = 0.004); with the treatment naive control group having a median (Mdn) of 1.92% (range 0.12% to 6.95%), iv group Mdn = 0.35% (range 0% to 8.26%), and the oral group Mdn = 0.82% (range 0% to 7.17%). Dunn's multiple post-test comparison revealed that the treatment naive and the iv-treated groups were significantly different in their γδ T-cell population distribution (p < 0.01) with both populations overlapping with the oral treatment group. As shown in Fig. 1, there is a notable loss of peripheral blood Vγ9Vδ2 T cells over time in osteoporotic patients on n-BP therapy, and this is most strikingly observed in the group on iv therapy who were found to be relatively deficient after 1 year on continuous therapy (r = −0.55, p < 0.0001) (Fig. 1A). Those on oral n-BP maintain fairly high Vγ9Vδ2 T-cell numbers for some time, and the loss is seen as more gradual (r = −0.32, p ≤ 0.03) (Fig. 1B). There was no observable effect of either iv or oral n-BP on the proportion of total T cells, monocytes, or granulocytes making up the peripheral leukocyte pool (Fig. 1C, D).

Figure 1.

Proportion of Vγ9Vδ2 T cells, total T cells, monocytes, and granulocytes present in patients with osteoporosis in relation to the length of time on nitrogen-bisphosphonate (n-BP) therapy. (A) In patients on intravenous (iv) treatment, a significant loss of peripheral blood Vδ2 T cells (r = −0.55, p < 0.0001) is evident within 18 months of continuous therapy. (B) In patients on oral therapy, a decline in Vδ2 T cells is observed (r = −0.32, p ≤ 0.03), but it is much more gradual, taking seemingly almost 4 years to materialize. (C) There was no observable difference in any of the other white blood cells assessed in relation to the length of time on intravenous (iv) treatment or (D) oral treatment. T cells are shown as black circles, monocytes as squares, and granulocytes as gray diamonds.

Monocytes are the primary immune cells that have been identified to take up n-BP and activate Vγ9Vδ2 T cells,9–11 and while conducting the flow cytometry analysis, it was observed that some patients had a subpopulation of monocytes with a high forward scatter (FSC) (Fig. 2A), which is a measure of cell size. We, therefore, looked to see if there was an association between the number of Vγ9Vδ2 T cells present in patients on n-BP and having large circulating monocytes. Interestingly, a negative correlation (r = −0.30, p < 0.05) could be observed between the number of Vγ9Vδ2 T cells present and the tendency to have large circulating monocytes for those on active n-BP therapy (Fig. 2B).

Figure 2.

An assessment of peripheral blood monocyte forward scatter (FCS), a measure of cell size, in patients with osteoporosis on nitrogen-bisphosphonate (n-BP) therapy. (A) Left panel depicts a representative flow cytometry plot of a monocyte population (CD14+ cells, y-axis) having a uniform population of cells with relatively low mean FCS (x-axis) in an individual with a relatively high proportion of Vγ9Vδ2 T cells (7.9%). On the right panel is a plot showing a patient who had a very low proportion of Vγ9Vδ2 T cells (0.3%) having a subset of CD14+ cells with a fairly high FCS. (B) A significant negative correlation (r = −0.30, p < 0.05) was found overall between the median FCS of monocytes of patients on n-BP therapy and the proportion of Vγ9Vδ2 T cells present.

We followed 9 of the study participants who were initiating or changing therapy over the short course to observe immediate changes (Supplemental Fig. S1). A slight peak in Vγ9Vδ2 T cells was observed in a number of patients immediately starting treatment (within 7 to 8 days), mostly in those with fairly robust baseline level of these T cells. The majority of those on oral treatment generally returned to around baseline levels in the time frame measured; however, a few patients had a drop after their first treatment. Those with initial low baseline levels of Vγ9Vδ2 T cells (ie, <2% of the T-cell pool) did not exhibit notable changes in peripheral blood γδ T cells during the short course. Of particular interest were 2 patients (participant ID 23 and 13 in Supplemental Fig. S1) switching from oral to iv administration because of intolerable gastrointestinal effects on the oral drug who had the highest proportion of Vγ9Vδ2 T cells in our study cohort with 7.87% and 8.26%, respectively. No acute phase reaction was reported.

Distribution of Vγ9Vδ2T cells in patients who recently experienced BAONJ

A total of 5 patients who had experienced ONJ within the last 24 months and 1 who was actively seeking treatment during the course of the study were enrolled in the BAONJ cohort to determine whether there was any association between their levels of Vγ9Vδ2 T cells and the status of their adverse drug reaction. The age at the time of the study, the type of n-BP therapy, the length of time on treatment before the development of BAONJ, and the medical history of these patients are listed in Table 2. The youngest patient in the ONJ group (age 48 years), who had been started on iv zoledronate after treatment for breast cancer, reported experiencing an acute phase reaction upon initial treatment.

Table 2. Characterization of Patients With BAONJ Included in the Study
 AgeRouteType of n-BPLength of time on drug before ONJ diagnosis (days)Time between ONJ and study (days)Continuation with n-BP after ONJ?Medical history
ONJ 185ivZoledronate 5 mg/year10411n/aTreated for blood malignancy before starting n-BP for osteoporosis; on prednisone, blood thinners, and cholesterol-lowering drugs
ONJ 255oralAlendronate (70 mg/week)1779422NoOn systemic corticosteroids and mycophenolate mofetil (strong immunosuppressant)
ONJ 360oralAledronate (70 mg/week), switched to ibandronate (150 mg/month)1256606NoInterferon therapy for hepatitis B and C
ONJ 460ivZoledronate 4 mg/4 weeks590329YesDiagnosed and treated for breast cancer before being put on zoledronate, on selective estrogen receptor antagonist
ONJ 560ivZoledronate 4 mg/4 weeks501425YesDiagnosed and treated for breast cancer before being put on zoledronate, on selective estrogen receptor antagonist
ONJ 648ivZoledronate 4 mg/4 weeks10171663-month pause before continuingDiagnosed and treated for breast cancer before being put on zoledronate, on selective estrogen receptor antagonist; experienced acute-phase response when starting n-BP therapy

The ONJ cohort were significantly younger than the n-BP treatment-naive osteoporosis controls (p < 0.01, t test), thus additional age and sex-matched controls (n = 11, mean age 58 ± 9 years) were recruited for a more appropriate comparative group. The levels of Vγ9Vδ2 T cells in the BAONJ patients (Mdn = 0.07%; range 0% to 0.50%) were significantly deficient in comparison to age-matched controls (Mdn = 2.40%; range 0.98% to 5.57%), U = 0, p = 0.001, as well as to the older osteoporosis controls (Mdn = 1.92%; range 0.12% to 6.95%), U = 3, p = 0.0008, and this was the only consistent difference that was observed. The flow cytometry plots showing the clear observable distinction between the BAONJ cases and controls in regard to Vγ9Vδ2 T cells are shown in Supplemental Fig. 2A, B, respectively.


γδ T cells are innate lymphocytes that have a nonredundant role in regulating immune homeostasis. They are known to both initiate a vigorous inflammatory response upon the first sign of danger12–14 as well as control excessive inflammation by killing activated macrophages and promoting resolution from trauma.15–19 Despite the knowledge that n-BP treatment leads to the stimulation, intended or not, of this influential subset of peripheral T lymphocytes, there has been a lack of information regarding the consequence of this chronic immune activation over time with continued n-BP treatment in patients with osteoporosis. We observed that there is a significant depletion of Vγ9Vδ2 T cells in osteoporotic patients who have no other underlying malignancy, which is directly related to the potency of the systemic dose and the length of time on therapy. This is of particular clinical relevance because much hope has been placed on the potential use of the serendipitous immunomodulatory effects of n-BP for the adjuvant treatment of cancer based on their demonstrated ability in vitro to induce the expansion of γδ T cells in the presence of the T-cell growth factor, IL-2, and enhance their ability to effectively target and kill cancer cells.20, 21 However, clinical trials have thus far proved disappointing because of the difficulty in sustaining γδ T-cell numbers in vivo in the majority of patients treated with n-BP even in the presence of IL-2.20, 22, 23 This failure has largely been attributed to the immunosuppressive environment present in the cancer patient as opposed to the potential overstimulatory effects of the drug.20 We observed that the loss of these T cells is seen most starkly in those on iv therapy who appeared to be relatively deficient within 18 to 24 months of being on treatment. In contrast, those on oral therapy showed a much more gradual decline that appears to take much longer to materialize. This difference is speculated to be owing in part to the poor absorption and bioavailability of the drug when taken orally (estimated to be less than 0.7%) in comparison to the intravenous route, which bypasses the variable and low absorption of the gut.24 Furthermore, there is also the potential contribution of a problem with compliance for those on oral treatment that is not present in patients receiving intravenous treatment.24 Of particular note, the negative relationship we found between the type and length of time of n-BP and the loss in peripheral Vγ9Vδ2 T-cell numbers almost perfectly parallel the reported timing of the development of the rare condition of BAONJ observed in susceptible patients.1, 25 BAONJ is far more commonly associated with iv n-BP, and for the most potent n-BP currently on the market, zoledronate, the average weighted time to the development of BAONJ was calculated to be 1.8 years and a minimum of 10 months.25 In contrast, less than 5% of BAONJ cases are associated with oral administration, and the mean time to its development was found to be 4.6 years, with a minimum of 3 years.25

Unexpectedly, we also observed that the two patients who experienced adverse gastrointestinal effects to the oral drug severe enough to warrant a switch to iv administration had the highest proportion of Vγ9Vδ2 T cells in our cohort. This may be coincidental, or it could imply that some of the more severe gastrointestinal side effects observed in patients on oral therapy may be attributed to the proportion of responsive Vγ9Vδ2 T cells present that have the potential to cause local inflammation in the gastrointestinal tract at the site of drug uptake. This is in line with the recent finding that the acute phase response experienced by patients with osteoporosis initially put on zoledronate is more common in younger individuals as a consequence of them having more Vγ9Vδ2 T cells.26 Together, these findings support the notion that monitoring the number and activation of Vγ9Vδ2 T cells both before and during treatment could be of important predictive value as this cell population may constitute a biomarker for n-BP side effect severity.

Monocytes are the primary immune cells identified to take up n-BP in the periphery and to activate γδ T cells.9–11 Our observation that there tended to be more large circulating monocytes in patients on treatment with low or deficient Vγ9Vδ2 T cells suggests that these lymphocytes may in part function to remove dysfunctional monocytes that have taken up n-BP before they are able to migrate to tissues and contribute to pathology. This may be an important mechanism of n-BP–induced adverse effects because it has been observed that patients on n-BP therapy have a distinct subset of giant osteoclasts, which are cells derived from the same lineage as monocytes, with unusual protracted apoptosis.27

Unlike αβ T cells, γδ T cells express growth factors important for tissue regeneration, such as fibroblast growth factor28 and connective tissue growth factor,29 that are critical for wound and skeletal fracture healing. Therefore, their deficiency may have a more fundamental impact on tissue homeostasis beyond their established role in regulating the activity of monocytes. The exact mechanism of the observed loss has yet to be fully delineated; however, γδ T cells are highly susceptible to AICD by FAS-mediated apoptosis, which occurs after repeated stimulation of their T-cell receptor.30 If prolonged/potent activation via treatment with n-BP leads to the depletion of these innate T cells, their loss of function may be among the predisposing factors culminating in the multifactorial phenomenon of BAONJ. Indeed, the patients who had developed BAONJ were notably deficient in Vγ9Vδ2 T cells, and this was the most consistent association out of the immune parameters assessed. Although a possibility, it is doubtful the observed deficiency of peripheral blood Vγ9Vδ2 T cells is because of their migration to the jaw or tissues in response to the local production of IPP because two of the BAONJ subjects had stopped n-BP therapy when first diagnosed with ONJ, which was more than a year before their participation in this study. This suggests that for those who have compromised immunity, like our BAONJ cohort, spontaneous recovery of Vγ9Vδ2 T cells after stopping treatment appears unlikely.

It should be noted that Vγ9Vδ2 T cells are a primate inheritance, which limits the potential of rodent animal models to fully assess the effects of n-BP therapy on immune function in humans. Similarly, human immunodeficiency virus (HIV) infection has shown to result in a loss of peripheral blood Vγ9Vδ2 T cells also supposedly via AICD,31 and it is a matter of potential interest to determine whether the increased susceptibility to osteonecrosis observed in these patients32 has any association with the loss of these innate T cells in a backdrop of a compromised immune system.

Vγ9Vδ2 T cells show a strong potential to serve as harbingers of possible adverse immune effects of n-BP therapy. We found these unique innate T cells are lost in osteoporotic patients on n-BP treatment, and this loss is related to the potency of the systemic dose and the length of time on therapy. Importantly, the observed negative effect on Vγ9Vδ2 T cells coincides with the reported timing of the development of the rare condition of BAONJ. We propose Vγ9Vδ2 T cells may be among the best biomarkers to identify those whose immune system have been affected as a result of n-BP therapy—particularly important in patients already having impaired immunity because they may be most vulnerable to the development of conditions such as BAONJ. There is, however, currently no evidence of a direct causative relationship between the number or function of circulating Vγ9Vδ2 T cells and BAONJ, and further studies are now needed to explore this novel potential link.


All authors state that they have no conflicts of interest.


SK is supported by a Fellowship from the Alexander von Humboldt Foundation of Germany. DK is supported by the “Inflammation-at-Interfaces” Cluster of Excellence. The successful completion of this work was greatly assisted by the support of the network of physicians in Kiel, especially Dr. Martin Mrugalla, who were invaluable in their assistance with the sample collection and recruitment of patients with osteoporosis. Nina Hedemann was instrumental in helping us establish key networking collaborations. We also thank Dr. Millan Patel for the critical review of the manuscript.

Authors' roles: SK conceptualized, designed, and carried out the analysis described in the study and wrote the first draft of the manuscript. ESQ assisted in the implementation of the study and ensured its successful completion. JW assisted with identifying and collecting data from patients with ONJ. HM assisted with the collection of the data from the osteoporosis cohort. DK provided the infrastructure and expertise for the flow cytometric analysis and assisted in the revision of the final manuscript. All authors approved the final version and verify the integrity of the data presented.