Introduction

The advent of the proteasome inhibitor bortezomib (Velcade; Millennium Pharmaceuticals, Cambridge, MA, USA; formerly PS-341) has yielded improvements in the overall survival of patients with multiple myeloma (MM).1 Bortezomib-based therapies are currently standards of care in patients with newly diagnosed2, 3 and relapsed/refractory4, 5, 6 MM.

This drug is generally well tolerated, although one of its most frequent adverse events is sensory peripheral neuropathy (PN). Bortezomib-induced peripheral neuropathy (BiPN) has previously been characterized in the setting of single-agent bortezomib±dexamethasone for relapsed or refractory patients.7, 8 In previously untreated patients, BiPN has been reported in up to 44% of patients, including grade 3 or higher sensory BiPN in up to 13% of cases, in phase III trials of bortezomib-based combinations.2, 3, 6, 9, 10, 11, 12 Similar rates have been observed in the relapsed setting,6, 7, 8, 13, 14 in which BiPN has been shown to be dose-related and cumulative.7, 8 Thus, BiPN often (17–30%) requires dose modification or discontinuation of bortezomib,2, 15, 16, 17, 18 which compromises the treatment outcome and quality of life of MM patients.

MM itself is also implicated in PN, and bortezomib and thalidomide may cause or exacerbate an existing neuropathy. These PN-inducible drugs have become the mainstay of therapeutic combinations with or without conventional chemotherapy.2, 3, 6, 19, 20, 21, 22, 23, 24 Furthermore, the introduction of bortezomib and thalidomide into induction regimens prior to high-dose therapy followed by autologous stem cell transplantation9, 12, 19, 20, 22, 24 leads to an increased risk of PN early in the disease course. These unwanted effects may hamper further treatment with these agents for consolidative or maintenance use and for the treatment of relapsed or progressive disease. Therefore, minimizing PN induced by these drugs with the use of predictive biomarkers will have significant impacts for clinicians in terms of treatment for as well as prevention of BiPN.

Lower doses (1.3 mg/m2 to 1.0 mg/m2)7, 16, 20, 24 or a weekly dose schedule21, 23, 25, 26 of bortezomib instead of the standard twice-weekly regimen is used to lower the incidence of PN. However, these alternative regimens reduce the efficacy of treatment.20, 21, 24, 25 In particular, weekly administration fails to overcome the negative impact on prognosis for a high-risk cytogenetic abnormality of t(4;14),21, 25 although this can be overcome with the original twice-weekly schedule.22, 27, 28

Little is known about the mechanism leading to BiPN, and it is likely that several factors contribute to BiPN, including dysregulation of mitochondrial calcium homeostasis,29 increased levels of polymerized and stabilized tubulin,30 autoimmune factors and inflammation,31, 32 and intracellular misfolded proteins.33 Furthermore, knowledge of predisposing factors may allow for the identification of patients at risk of BiPN, including age,34 baseline PN grade,7, 10, 17, 35, 36 and diabetes mellitus;17 however, conflicting results have also been reported.7, 8, 10, 17, 34, 36 Only one study using pretreatment plasma cells obtained from the bone marrow of MM patients has characterized the potential role of protein-related factors in BiPN.18 In addition, a small number of other reports investigating single nucleotide polymorphisms have strongly suggested a role for the patient’s genetic background in the development of BiPN.11, 37

Because PN limits patients’ treatment options, the prediction of BiPN is highly anticipated clinically to achieve personalized treatment with bortezomib. We therefore analyzed gene expression profiles in peripheral blood (PB). Unlike conventional gene expression profiling, where mRNA expression is characterized in isolated mononuclear cells, we quantified mRNA expression after stimulation of PB with appropriate agents ex vivo. This unique assay was designed to characterize diverse leukocyte functions in clinical samples.38, 39

Patients and methods

Patients

Eligible patients with symptomatic MM, who were previously treated or untreated, received twice-weekly or weekly bortezomib doses of 1.3 mg/m2 intravenously on days 1, 4, 8 and 11, with or without 20 mg of dexamethasone mainly on days 1, 2, 4, 5, 8, 9, 11 and 12. Otherwise, these patients received 1.3 mg/m2 of bortezomib on days 1, 8, 15 and 22 with or without dexamethasone. To be included in the study, all patients had to have measurable M-protein levels. This study was approved by the institutional review board or the independent ethics committees at all participating institutions and was conducted according to the principles of the Declaration of Helsinki and the International Conference on Harmonisation guidelines on Good Clinical Practice. We also obtained written informed consent from the patients for sample procurement. Severity of PN was graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events version 4.0. Dose modification in cases of BiPN was managed as previously described.7

Ex vivo blood treatment

In eight-well strips, 1.2 μl each of 50 × concentrations of phytohemagglutinin (PHA) (2 mg/ml), heat-aggregated immunoglobulin G (HAG) (10 mg/ml), lipopolysaccharide (LPS) (0.5 mg/ml), zymosan A (ZA) (75 mg/ml) and solvent phosphate-buffered saline were dispensed (three wells were blank) and then delivered to each institution in a dry ice package. These strips were kept frozen at −80 °C. Two milliliters of heparinized whole blood was obtained from each patient before (Day 0), 2–3 days (Days 2–3) and 1–3 weeks (Weeks 1–3) after the first dose of intravenous bortezomib. Blood was immediately delivered to the designated laboratory, and 0.06 ml of whole PB was added into each well of three strips (triplicate). Three wells were removed and immediately stored at −80 °C. The remaining five wells, containing PHA, heat-aggregated immunoglobulin G, LPS, ZA and phosphate-buffered saline, were incubated for 4 h at 37 °C. The total blood volume required was as low as 1.44 ml (0.06 ml/well × 8 wells/strip × 3 strips). After incubation, samples were stored at −80 °C.

mRNA measurement

The levels of the target mRNAs (interleukin (IL)2, interferon (IFN)G, granulocyte-macrophage colony-stimulating factor, tumor necrosis factor ligand superfamily member 2 (TNFSF2), CCL4, IL6, IL10, CXCL10, CTL-associated antigen 4 (CTLA4) and control action (ACT)B) in each aliquot were quantified using a previously described method.38, 39 The melting curve was always analyzed to confirm that PCR signals were derived from a single PCR product. The cycle threshold (Ct) was determined using analytical software (SDS, ABI, Foster City, CA, USA). The Ct values of drug-treated triplicate samples were subtracted individually from the mean Ct values of control samples to calculate ΔCt, and the fold increase was then calculated as 2^(−ΔCt), as described previously.39 The collection of clinical data (TW and MC) and the analysis of mRNA (MM) expression were performed separately at different centers.

Control experiments

Whole PB samples were obtained from three healthy volunteers at APEX Research (Tustin, CA, USA) and used as controls for the ex vivo study. In brief, heparinized PB was pretreated with 0.1, 1, 10, 100, or 1000 ng/ml of bortezomib overnight at 37 °C prior to 4-h stimulation with PHA, heat-aggregated immunoglobulin G, LPS, ZA and phosphate-buffered saline, respectively. The same mRNAs used for MM patients were also quantified. The study was approved by the institutional review board, and written informed consent was obtained from each participant. The purpose of control samples was to screen biomarkers and to design assay condition, not to compare the data between control and MM patients. Thus, the number of control samples is small (n=3) without being age- and gender-matched.

Statistical analysis

Parametric (Student’s t-test) and non-parametric (unpaired Mann–Whitney U-test and Pearson’s χ2 test, paired Wilcoxon test) were used for continuous variables and categorical variables, respectively, to compare mRNA levels between the two groups. P-values <0.05 were considered significant. The statistical analyses were performed using Excel (Microsoft) and Prism 6 (GraphPad Software, La Jolla, CA, USA).

Results

Patient characteristics

Between March 2010 and March 2012, a total of 64 patients (33 male and 31 female patients) were enrolled from six centers. The median age of the patients was 62 years (Table 1). The original protocol stipulated that only relapsed or refractory patients were eligible, but in October 2011, the protocol was amended to include newly diagnosed patients owing to the changes in the Japanese National Health Insurance policy. Consequently, 42 patients were previously treated and 22 patients were untreated (Table 1).

Table 1 Patient demographics and baseline characteristics
Full size table

Assessment of PN before and after bortezomib treatment

After excluding one patient each for loss of consciousness, early death due to progressive disease, or additional treatment with chemotherapy, 61 patients were assessable for toxicity analysis with a minimum follow-up of 1 month (Table 2). The median number of bortezomib doses before the onset of PN was 4.5 (range, 1–14). The median time to onset of BiPN was 30.5 days (range, 3–83). Unexpectedly, only two (3.2%) and one (1.6%) patients reported grade 1 and 2 PN at baseline, respectively (Table 2).

Table 2 PN before and after treatment with bortezomib
Full size table

Of the 58 (95%) patients without PN before enrollment, 18, 9 and 4 patients developed grade 1, grade 2 and grade 3 BiPN after treatment, respectively (Table 2). In addition, 1, 8 and 3 patients who experienced grade 1, 2 and 3 BiPN with no baseline neuropathy, respectively, experienced pain (Table 2). After treatment with bortezomib, the symptoms of both of the two patients with grade 1 PN prior to treatment worsened to grade 2. No change was observed in the one patient with grade 2 PN at baseline (Table 2). None of the patients with grade 1 or 2 PN at baseline or grade 2 PN after treatment complained of pain. The median follow-up time of assessable patients was 162 (range, 31–666) days.

Blood sample preparation

A total of 2444 (64 patients × 5 stimulants × 3 time points × 3 (triplicate) with some sampling failures subtracted) preparations for both mRNA and cDNA synthesis were carried out. The number of PCRs performed was 19 552 (2444 cDNA × 8 primers). In total, the perfect sets for 53 patient blood samples were tolerable for qualification for mRNA analysis.

Effect of bortezomib on ex vivo mRNA induction in healthy volunteers

As shown in Supplementary Table 1, ex vivo mRNA induction was confirmed using various combinations of stimulants and the corresponding changes in mRNA expression, as previously described.39 Pretreatment with high concentrations of bortezomib (300 and 1000 ng/ml) inhibited these ex vivo mRNA inductions, with the exception of ZA-induced IFNG, PHA-induced CLL4 and PHA-induced CTLA4 (Supplementary Table 1). Supplementary Figure 1 shows a graphical representation of some of the key results corresponding to subsequent follow-up studies. As shown in Supplementary Figure 1, PHA-induced IL2 (a), IFNG (b) and TNFSF2 (c), and LPS-induced IL6 (e) were inhibited by pretreatment with bortezomib at doses greater than 300 ng/ml in a dose-dependent manner, whereas expression of the control housekeeping gene ACTB was unchanged by PHA (d) or LPS (f) stimulation.

Association between the incidence of BiPN and ex vivo mRNA induction at Day 0

The data set consisting of four stimulants × eight different mRNAs obtained prior to bortezomib treatment showed that BiPN grade was significantly associated with PHA-induced IL2, IFNG and TNFSF2, and LPS-induced IL6 levels (Supplementary Table 1). Graphical representations of these positive results are shown in Figures 1a–d. The fold increase in PHA-induced IL2 in the patients who never experienced BiPN was significantly higher than those who developed grade 1 (P=0.041 with Student’s t-test only) or grade 1–3 combined BiPN (P=0.007 with Student’s t-test and P=0.049 with Mann–Whitney U-test) (Figure 1a). Similarly, PHA-induced IFNG (Figure 1b) and TNFSF2 (Figure 1c) and LPS-induced IL6 (Figure 1d) levels were significantly higher in patients with PN grade 0 than those with grade 1–3 combined BiPN (P-values were shown in Figure 1). Because the number of patients with PN grade 3 was very small (n=3), it was difficult to assess these changes statistically; however, the fold increase in PHA-induced TNFSF2 and LPS-induced IL6 levels in patients with PN grade 2 were significantly lower than those in patients with PN grade 0 (P-values were shown in Figure 1), whereas no difference was found between patients with grade 0 and 1 PN (Figures 1c–d).

Figure 1
figure 1

Association between the incidence of BiPN and ex vivo mRNA induction at Day 0. The fold change of PHA-induced (a) IL2, (b) IFNG and (c) TNFSF2, and LPS-induced IL6 (d) mRNA in whole blood obtained prior to bortezomib administration was compared with the grade of BiPN. The grade was determined according to the criteria of the National Cancer Institute Common Terminology Criteria for Adverse Events version 4.0. All patients who did not show any PN prior to bortezomib treatment were included in the analysis. P-values were calculated using both parametric (Student’s t-test) (t) and non-parametric (unpaired Mann–Whitney) (M) test.

Full size image

Prediction of BiPN by PHA-induced IL2 mRNA at Day 0

The results of PHA-induced IL2 (Figure 1a) were also statistically significant by χ2 test using a threefold increase as the cutoff value (Table 3). More importantly, of the 19 patients with a >3-fold increase in PHA-induced IL2 at Day 0, 14 did not suffer from BiPN (73.7% prediction for non-BiPN patients); in contrast, of the 34 patients with a <3-fold increase in PHA-induced IL2 at Day 0, 23 experienced BiPN (67.6% prediction for BiPN patients) with the sensitivity (56%) and the specificity (82%), respectively (Table 3). Although PHA-induced IFNG and TNFSF2 and LPS-induced IL6 levels at Day 0 were significantly different between the patients with PN grade 0 and 1–3 combined BiPN (Supplementary Table 1 and Figures 1b–d), no clear cutoff value was identified. The mRNA data from blood samples obtained after bortezomib treatment (Days 2–3, Weeks 1–3) did not show any substantial correlation to BiPN (data not shown).

Table 3 Prediction of BiPN according to pretreatment PHA-induced IL2 mRNA
Full size table

Changes in PHA-induced IL2 mRNA after bortezomib treatment

In patients without PN even after bortezomib treatment, PHA-induced IL2 was significantly reduced at Days 2–3 after bortezomib treatment (P=0.046 with paired Student’s t-test, Figure 2a), and such inhibition was transient and had been neutralized by Weeks 1–3 (Figure 2a). Owing to the large variation, no significant results were obtained in the patients with grade 1 PN (Figure 2b). Interestingly, in the patients with grade 2 or 3 PN, bortezomib treatment showed a significant slow and sustained inhibition of PHA-induced IL2 (P=0.044 with paired Student’s t-test, P=0.027 with paired Wilcoxon) at Weeks 1–3 with much weaker inhibition at Days 2–3 (P=0.020 with paired Wilcoxon) (Figure 2c).

Figure 2
figure 2

Changes in PHA-induced IL2 mRNA before and after bortezomib treatment. The fold changes in PHA-induced IL2 mRNA expression at Day 0 (D0), Days 2–3 (D2–3) and Weeks 1–3 (W1–3) were compared among the patients with BiPN grade 0 (a), grade 1 (b) and grade 2–3 (c). P-values were calculated using both parametric (paired Student’s t-test) (t) and non-parametric (paired Wilcoxon (W)) tests.

Full size image

Discussion

The results from this study demonstrated several new findings. First and most importantly, we found that the amplitude of the PHA-induced increase in IL2 mRNA expression at Day 0 was capable of distinguishing patients who experience BiPN from those who do not with cutoff values of threefold (Table 3). As this assay was done at Day 0, before administering bortezomib, the effect of bortezomib was excluded. Thus, poor responses of PHA-induced IL2 mRNA found in BiPN may suggest a couple of possibilities, such as the decrease in the number of PHA-responsive cells, decrease in PHA signaling cascades, presence of inhibition against PHA signaling, or increase in the baseline IL2 mRNA levels. Because the original plan of this study was to evaluate gene expression changes before and after bortezomib treatment, the discovery of predictable biomarker was a big surprise. Thus, we were not prepared for the characterization of IL2 mRNA responses. Such detailed analysis will be conducted in subsequent studies.

Second, the results of PHA-, LPS- and ZA-induced mRNA expression were carried out in PB, although the primary action of bortezomib is against myeloma cells in the bone marrow. Because myeloma cells are difficult to obtain as a routine diagnostic test, PB cells are used as a practical surrogate marker for the prediction of BiPN.

Third, the blood manipulation and laboratory procedure were robust, high throughput and easily applicable to a clinical setting. The blood volume needed for the simple and unique assay system used in this study was as small as 1.44 ml, and there was no need for mononuclear cell isolation. The 96-well format facilitated a smooth transition from mRNA purification to cDNA synthesis to PCR, which allowed us to analyze many samples simultaneously and quantitatively.

Fourth, our results supported the hypothesis that BiPN may be linked to the alteration of inflammatory processes.11, 31, 32 This was not only supported by the observed PHA-induced IL2 mRNA expression (Figure 1, Table 3 and Supplementary Table 1) but also the PHA-induced IFNG and TNFSF2 and LPS-induced IL6 mRNA levels (Supplementary Table 1). Moreover, patients who developed grade 2–3 BiPN demonstrated sustained inhibition of PHA-induced IL2 mRNA even after 1–3 weeks of bortezomib treatment, whereas patients who did not experience BiPN only showed transient inhibition at Days 2–3 (Figure 2). Moreover, ex vivo bortezomib treatment in healthy adult blood samples significantly inhibited PHA-, LPS- and ZA-induced mRNA induction (Supplementary Table 1 and Supplementary Figure 1). All of these mRNAs are known to encode inflammatory cytokines, and the precise mechanisms responsible for their effects should be characterized in subsequent studies.

The combination of bortezomib and thalidomide does not result in additive neurologic toxic effects. It is worth noting that in a phase III trial comparing VTD (bortezomib with thalidomide plus dexamethasone) and thalidomide plus dexamethasone as induction therapy before and consolidation therapy after double autologous stem cell transplantation in untreated MM patients,22 the rate of grade⩾3 PN during VTD induction therapy was similar to that reported by other groups after a similar number of cycles of bortezomib and dexamethasone.9 In addition, the retrospective analysis assessing bortezomib in combination with thalidomide showed that the incidence of grade 3 to 4 PN was relatively low.17 These results appear paradoxical because both drugs are known to cause PN. However, it is likely that thalidomide has a protective effect with respect to BiPN through its anti-inflammatory action, specifically through inhibition of TNFα. Furthermore, unexpectedly, BiPN was significantly ameliorated in some patients who received a salvage treatment with lenalidomide,17 and this improvement may also be attributable to the putative anti-inflammatory actions of lenalidomide as well as dexamethasone. Moreover, the low rates of severe PN observed in most RVD (lenalidomide with bortezomib plus dexamethasone) studies are also likely explained by the anti-inflammatory effects of lenalidomide and the reduced dose of bortezomib,40, 41 although dexamethasone may also contribute to the attenuation of BiPN through its anti-inflammatory effect.

The results of this study are reminiscent of earlier findings in patients with an bortezomib-induced erythematous maculopapular rash, which skin biopsies revealed perivascular leukocytoclatic vasculitis42 and disappeared following the corticosteroid treatment.43 In addition, our previous study demonstrated that the bone marrow microenvironment cells likely have a role in the periodic fevers after each bortezomib dose.44

If BiPN can be predicted, we may be able to choose an alternative administration route14 of bortezomib by weighing the risks against the benefits of bortezomib. Alternatively, physicians may select to treat with carfilzomib, which is another proteasome inhibitor with fewer complications of PN, although frequent visits to the clinic are necessary for two consecutive days per week in three consecutive weeks in a 28-day cycle.45, 46

The use of whole blood in this experiment has pros and cons. The advantage is that the assay is technically applicable to clinical settings. The disadvantage is that we do not know which cells are responding to PHA. When peripheral blood mononuclear cells were isolated and stimulated with PHA, the induced IL2 mRNA was substantially smaller than that of whole blood (data not shown). If we further isolate subsets of PMBC, more artifacts will be introduced, and required large volume of blood. As cellular experiments are very complex, such experiments should be done as an independent study.

In conclusion, our study found that PHA-induced IL2 mRNA levels in PB obtained prior to bortezomib treatment may serve as a biomarker for BiPN in MM patients. Although PB samples must be treated with PHA in test tubes ex vivo to perform this analysis, this is clinically feasible. Together, our findings indicate that PHA-induced IL2 mRNA levels may help clinicians identify patients at greater risk of BiPN and personalize treatment regimens for each patient. Owing to the significant clinical importance of these findings, the results warrant validation in a larger patient cohort.