Ex Vivo Assays of Dendritic Cell Activation and Cytokine Profiles as Predictors of In Vivo Effects in an Anti-Human CD40 Monoclonal Antibody ChiLob 7/4 Phase I Trial

Immunotherapy has advanced rapidly over recent years with several monoclonal antibodies entering the clinic for treatment of autoimmunity and cancer. In particular, immunostimulatory antibodies are showing promise in enabling an individual's own immune repertoire to mount a successful immune response specific against cancer; however, their use comes with a high risk of serious cytotoxicity as was notably encountered with the TeGenero TGN1412, anti-CD28 super-agonist antibody in which a near fatal cytokine storm was experienced by all participants (1). The dilemma that researchers face is the ability to design and predetermine whether an immunostimulatory antibody is able to activate the cells they are targeting without triggering a cytotoxic storm. We have therefore investigated whether ex vivo assays are predictive of the ability of an agonistic anti-CD40 monoclonal antibody, ChiLob 7/4, to activate dendritic cells and monitored for the simultaneous expression of cytokines.

CD40 is a 48-kDa, type I membrane protein belonging to the TNF receptor (TNFR) superfamily. It is expressed primarily on antigen-presenting cells (APC) such as dendritic cells, B lymphocytes, and monocytes but it has also been found on endothelial and epithelial cells (2, 3). The natural ligand for CD40, CD154 (CD40L), is a member of the TNF family and exists in both soluble and membrane-bound forms where it appears to form a trimeric protein structure (4). CD154 is expressed predominantly on activated CD4⁺ T lymphocytes.

An adaptive immune response is triggered when resting CD4⁺ T cells are activated following the recognition of MHC–antigenic peptide complex presented by dendritic cells in concert with costimulatory molecules. Once activated, CD4⁺ T cells proliferate and further induce immune and inflammatory responses via the secretion of cytokines and the expression of a variety of cell surface molecules that include CD154. It is through the CD154–CD40 interaction that CD4⁺ T helper (T_H) cells signal back to the dendritic cell leading to their maturation and licensing to activate CD8⁺ cytotoxic T cells (5–8). Preclinical in vivo work in syngeneic mouse models of malignancy suggest that one of the most potent therapeutic effects of agonistic anti-CD40 antibodies relates to their ability to effectively license or condition dendritic cells to completely bypass the need for specific CD4⁺ T-cell help to directly activate CD8⁺ cytotoxic T-cell precursors (9, 10).

The important functional role of CD154–CD40 interaction in vivo has made CD40 an attractive target for cancer immunotherapy, and several clinical trials have been conducted with different agonistic anti-CD40 antibodies (11–13). More recently, the potential of anti-CD40 antibody therapy to engage non–T-cell–dependent immune effector responses has been highlighted through studies on pancreatic cancer (14). Comprehensive human and murine studies support the role of antibody-activated myeloid cells within the tissue stroma as being crucial to the successful outcome of the treatment (15).

As part of a phase I clinical trial (NCT01561911), we have monitored the effects of ChiLob 7/4 on the activation of human myeloid dendritic cells (mDC1) and plasmacytoid dendritic cells (pDC) and monitored the associated plasma cytokine profiles (16). We have shown that the effects of ChiLob 7/4 on dendritic cell activation and the cytokine release profiles using ex vivo whole-blood stimulation assays are reflective of changes seen in the participants on the ChiLob 7/4 phase I study. Furthermore, we observed a mild cytokine release syndrome (CRS) at the first observable biologic effect level using in vivo endpoint assessments that were consistent with our preclinical evaluations. The activation of dendritic cells as measured by CD83 and CCR7 upregulation and the release of MIP1β and IL-12 suggest the initiation of the desired T_H1-type immune environment without the presence of an uncontrolled cytokine storm.

Patients with CD40-positive malignancies who were refractory to conventional anticancer treatments were enrolled in the phase I dose-escalation study to evaluate the chimeric anti-CD40 monoclonal antibody (mAb) ChiLob 7/4. The study protocol and patient consents were approved by the Cancer Research UK (CRUK) Central Institutional Review Board (CIRB) and the local research and ethics committee (LREC). The primary objective of the study was to establish the safety and tolerability of the ChiLob 7/4 mAb. All patients had tumor assessments by computerized tomographic (CT) scan within 4 weeks before treatment. The dosing schedule evaluated cohorts of 3 patients at the dose levels of 0.5 mg (cohort 1), 1.6 mg (cohort 2), 5 mg (cohort 3), 16 mg (cohort 4), 50 mg (cohort 5), and 160 mg (cohort 6) given by weekly i.v. infusion over a 4-week period. Twenty-one patients were recruited into the study, and the demographics of these patients are shown in Supplementary Table S1.

Patients received 4 weekly i.v. infusions of the chimeric IgG1 ChiLob 7/4 mAb at study days 1, 8, 15, and 22 with follow-up at day 56. Total infusion times for the 6 escalating doses were 30, 30, 30, 71, 139, and 273 minutes, respectively. Premedication with oral paracetamol and i.v. chlorpheniramine (10 mg) was administered as standard. Intravenous hydrocortisone was administered if signs of a CRS were noted.

Exceptions to this schedule were noted within cohort 4 at the 16-mg dose level, in which adverse reactions potentially attributable to a CRS were noted (Supplementary Table S2). The initial patient (subject 10) of this cohort developed a probable grade 2 nonserious CRS reaction on the first dose, requiring i.v. steroid treatment and a 50% reduction in infusion rate through to completion. Due to the presence of an adverse event within a cohort, the trial protocol required an additional patient to be recruited to this cohort before escalating the dose for cohort 5. This individual proceeded through the second infusion with no CRS but had milder episodes during infusions 3 and 4 that were treated in a similar manner to the initial adverse reactions. The second patient (subject 11) in this cohort had a probable grade 2 nonserious CRS reaction on the first dose but following steroids and 50% rate reduction completed the infusion. No further CRS events occurred during the 3 additional infusions. The third patient (subject 12) of this cohort did not experience any probable CRS events during the treatment period. The last patient (subject 13) in this cohort experienced a probable CRS event on the first infusion that was treated with i.v. steroids and a 50% rate reduction through to completion. No further CRS events occurred during the 3 additional infusions. Following on from this experience, the premedication was adjusted to include i.v. hydrocortisone before the first infusion for cohort 5 (50 mg) and cohort 6 (160 mg). No further CRS events were recorded from these 5 individuals.

Anti-coagulated EDTA whole-blood samples were collected following informed consent from patients enrolled on the phase I study, ethically approved by the National Research Ethics Service, UK, and conducted in accordance with the Declaration of Helsinki. Samples were taken at preinfusion and at various time points postinfusion and analyzed immediately by flow cytometry. Samples for cytokine analysis were centrifuged at 300 × g for 15 minutes and the plasma removed and stored at −70°C. Whole-blood samples from fully anonymous, consented healthy subjects were used for the ex vivo assays.

Ex vivo whole-blood stimulation assays were set up within 1 hour of blood collection from healthy subjects. One milliliter of the anticoagulated whole blood was stimulated with ChiLob 7/4 (± cross-linking), with either no additional ligand or the addition of Toll-like receptor (TLR) ligands and incubated at 37°C, 5% CO₂. ChiLob 7/4 was cross-linked using a 1:1 ratio of antibody to goat anti-human IgG Fc fragment–specific antibody (Jackson ImmunoResearch Europe Ltd.) and incubated for 30 minutes at room temperature in PBS. The TLR ligands used as established dendritic cell activators were unmethylated cytosine and guanine oligodeoxynucleotides 5′-gggggACGATCGTCgggggg-3′ (CpG2216) (MWG Biotech) and lipopolysaccharide (LPS; Sigma-Aldrich Co. Ltd). At the end of 4-hour incubation, 300-μL samples were removed for flow cytometric analysis. A further 500 μL of each sample was removed and centrifuged at 300 × g for 15 minutes. The plasma was removed and stored at −70°C.

The following directly conjugated mAbs were used: anti-CD1c-(BDCA-1)-PE, anti-CD303-(BDCA-2)-FITC, and anti-IL-12 APC from Miltenyi Biotec; anti-CD14-PE-Cy5 and anti-CD19-PE-Cy5 from Beckman Coulter; anti-CD45-(APC-Cy7), anti-CD83-APC, and anti-MIP-1β-APC from BD Biosciences. FACS Lysing solution and Fix Perm kit from BD Bioscience and Fixation and Dead Cell Discriminator Kit from Miltenyi Biotec were used according to the manufacturer's instructions.

Three hundred microliters of anticoagulated blood samples were labeled with single fluorochrome-conjugated antibodies recognizing cell surface markers and dead cell discriminator and incubated on ice for 10 minutes under a 60 W bulb. Antibodies conjugated to tandem dyes were then added and the tubes incubated for a further 10 minutes on ice in the dark. Red blood cell lysis was conducted using BD FACS lysing solution, followed by 3 washes in wash buffer (PBS/BSA/azide) with the final pellet resuspended in 300 μL wash buffer, 150 μL fix buffer, and 5 μL discriminator stop reagent. Samples were acquired using the BD FACS Canto II and analyzed using the FACS Diva software. The gating strategy used was as reported (17). Briefly, leukocytes were gated using a side scatter versus CD45 gate. This population was assessed for CD19/CD14-expressing cells and high side scatter to gate out the unwanted populations of B cells, monocytes, and granulocytes respectively. The remaining cells were gated (P1 on Fig. 1A), and the BDCA-1 or BDCA-2 positive cells were identified as mDC-1 and pDC, respectively. Changes in their level of CD83 and CCR7 expression were monitored to determine activation status.

All cytokine analyses were conducted on frozen plasma samples. Luminex human custom 10-plex kits (for IL-2, IL-4, IL-6, IL-8, IL-10, IL-12p70, IFNγ, TNFα, MIP1β, and MIP1α) were purchased from Invitrogen Life Sciences. The validation of this 10-plex custom kit has been reported previously (18). All reagents were provided with the kit and prepared according to the manufacturers' protocol booklet. The assays were conducted in 96-well filter bottom plates and analyzed on the Luminex xMAP 100 system (Luminex Corp.), using StarStation software version 2.0 (Applied Cytometry Systems). Each sample was run in duplicate, and the mean concentration of each analyte was calculated using a 4- or 5-parameter logistic fit curve generated from the standards.

For intracellular cytokine measurements, stimulation assays were set up as described above. For intracellular dendritic cell cytokine analysis, Brefeldin A was added after 1-hour stimulation and the tubes incubated for a further 4 hours. Cells were labeled for cell surface markers as described, followed by red blood cell lysis and washing. The final cell pellet was resuspended in 100 μL BD fix/perm buffer and incubated on ice for 20 minutes in the dark. After washing twice with perm wash buffer, the antibodies against the intracellular cytokines were added and the tubes incubated in the dark at room temperature for 20 minutes and then washed once in perm wash followed by one wash with fluorescence-activated cell sorting (FACS) buffer. The samples were analyzed on a BD FACS Canto II flow cytometer.

Differences between groups were assessed for statistical significance using the Student t test. Differences were considered significant when P < 0.05. Pearson correlation coefficient (r) was used to determine the relationship between 2 properties with r > 0.5 considered as high positive correlation. Error bars show SEM.

Unmanipulated anticoagulated whole-blood samples were used for all the ex vivo assays as this was considered to be most physiologically reflective of systemic changes that would occur in peripheral blood. To determine whether ChiLob 7/4 was able to activate circulating mDC1s and pDCs, whole-blood assays were conducted using a titrated concentration of ChiLob 7/4, either soluble or cross-linked with a goat anti-human IgG Fcγ-specific antibody. Using flow cytometry, changes in CD83 surface expression were measured as an early indicator of dendritic cell activation (Fig. 1A). Soluble ChiLob 7/4 was unable to upregulate CD83 on pDCs and mDC1s; however, upon cross-linking, there was a significant upregulation on both populations (Fig. 1B). The level of upregulation had a positive correlation with increasing doses of ChiLob 7/4 (r > 0.6). The level of CD83 upregulation on pDC did not reach the equivalent level seen on mDC1s. No change in CD83 expression was found on mDC1s or pDCs when ChiLob 7/4 was air-dried on plates before addition of whole blood (data not shown).

Plasma from the same whole-blood activation assays was removed and analyzed for changes in secreted cytokines using the Luminex 100 platform. Cytokine profiles were generated for a panel of 10 cytokines using a custom 10-plex kit (Fig. 1C). Little change was measured in samples activated with soluble ChiLob 7/4, but when cross-linked, increases were seen in IL-12, MIP1α, MIP1β, IL-8, IL-6, and TNFα showing strong correlation (r > 0.9) with increasing concentration of cross-linked ChiLob 7/4. The pattern of cytokine secretion for ChiLob 7/4 was distinctive, with MIP1β expressed most abundantly and earliest, accompanied by the other cytokines at the 50 μg concentration. Low levels of IL-4 and IL-10 were measurable at the highest cross-linked concentration of antibody (Fig. 1D). No cytokines were detected when ChiLob 7/4 was air-dried on plates, emphasizing the particular importance of cross-linking with a secondary antibody to observe the activating nature of this antibody.

Although both the flow cytometry and Luminex assays were able to measure responses when the antibody was cross-linked, the flow-based assay was able to detect mDC1 activation with the soluble form of the antibody before any cytokine response was detected. This suggests that the antibody has the potential to activate its target cells without the onset of excessive cytokine release. The pDCs were not activated unless the antibody was cross-linked, suggesting a higher activation threshold, as more antibody or aggregated CD40-CD154 interaction was needed to trigger a response. Both assays provide an effective approach for screening the response to ChiLob 7/4 antibody, flow cytometry to monitor target cell activation, and Luminex to monitor CRS.

We wanted to determine how ChiLob 7/4 compared with other dendritic cell–activating ligands. Therefore, the activation capacity of ChiLob 7/4 was compared with the level of activation triggered by known dendritic cell activating TLR ligands, specifically oligodeoxynucleotide (ODN) containing cytosine–phosphate–guanine (CpG) and LPS (Fig. 2A). Upregulation of CD83 was measured as an early indicator of dendritic cell activation and upregulation of CCR7 as an indicator of dendritic cell maturation and migration. As anticipated, the level of activation varied between individuals; however, the changes induced by each ligand were consistent within each cohort and correlated with those expected by the ligand. CpG2216, which activates via TLR-9, was found to be a weak dendritic cell activator in terms of CD83 and CCR7 upregulation, whereas LPS that activates via TLR-4 was a potent activator of mDC1s but also showed reasonable pDC activation. In comparison, soluble ChiLob 7/4 only minimally activated mDC1s and pDCs, and although nonsignificant was at the equivalent level of the weaker TLR ligand CpG2216. When cross-linked, ChiLob 7/4 was at least as effective as the more potent TLR ligand LPS in activating the mDC1s and pDCs, with CD83 upregulation reaching the equivalent level induced by LPS (r > 0.9), although CD83 was still consistently lower with the cross-linked ChiLob 7/4 in comparison to LPS (Fig. 2B). Our previous studies have shown CCR7 upregulation to occur 24 hours after LPS stimulation (16), and the results of this study would suggest the kinetics of mDC1 CCR7 upregulation with cross-linked ChiLob 7/4 occur more rapidly, detectable after 4 hours. It is clear from these results that in an ex vivo setting ChiLob 7/4 is most effective when cross-linked, and its ability to activate is similar to that of LPS with both dendritic cell subsets becoming activated, mDC1s more so than pDCs, but having the ability to trigger chemotactic markers on dendritic cells more dynamically than LPS.

To determine whether the ChiLob 7/4 activity could synergistically or otherwise enhance TLR ligand responses, whole-blood assays were conducted combining soluble or cross-linked ChiLob 7/4 with LPS or CpG. Although cross-linked ChiLob 7/4 and LPS were each highly effective on their own at activating dendritic cells, in combination, there was no significant additive effect on dendritic cell activation. In combination with cross-linked ChiLob 7/4, CpG2216, which triggered low CD83/CCR7 activation on its own, did not enhance the levels induced by cross-linked ChiLob 7/4 alone. Similarly, there was no additional upregulation with the addition of cross-linked ChiLob 7/4 to LPS. These results suggest that the addition of ChiLob 7/4 (even in the highly activating cross-linked form) would not exacerbate any preexisting activation of dendritic cells.

The plasma removed from the whole-blood assays was used to generate 10-plex cytokine profiles (Fig. 3). We wanted to assess whether during the process of dendritic cell activation the cytokine profile generated by ChiLob 7/4 was similar to that generated by LPS or CpG. The amount of each cytokine produced was variable between individuals; however, the same group of cytokines was induced/secreted by the same ligand. Soluble ChiLob 7/4 produced a similar cytokine expression profile to that of the weaker TLR ligand CpG2216, with only MIP1α, MIP1β, IL-8, and TNFα being minimally increased above the unstimulated levels, with MIP1β already having a relatively high concentration in the unstimulated control sample. Cross-linked ChiLob 7/4 generated a profile similar to that of the more potent TLR ligand, LPS, producing high levels of MIP1α (>5,000 pg/mL), MIP1β (>10,000 pg/mL), TNFα (>5,000 pg/mL), IL-6 (>5,000 pg/mL), IL-8 (>1,000 pg/mL), and IL-12 (>250 pg/mL); notably, the level of IL-6 expression was about 8-fold less than that induced by LPS (50,000 pg/mL), suggesting an overall less pro-inflammatory response.

Combinations of soluble and cross-linked ChiLob 7/4 with either CpG or LPS showed a similar pattern as the changes seen in the CD83 and CCR7 expression. That is, CpG or LPS in combination with cross-linked ChiLob 7/4 resulted in cytokines being produced at the same level as cross-linked ChiLob 7/4 alone and LPS alone, respectively, with no indication of a synergistic effect (data not shown).

Peripheral blood samples from patients treated with ChiLob 7/4 were analyzed for CD83 and CCR7 expression on mDC1s and pDCs, and cytokine profiles were generated from plasma samples. Baseline samples were taken at preinfusion (day 1) and at day 4 postinfusion for each weekly cycle of treatment over 4 weeks to analyze the changes in dendritic cell activation. Changes above the baseline level of expression of dendritic cell activation markers were measurable at a minimal dose of 16 mg (∼3.2 μg/mL; Fig. 4A). Changes in CD83 and CCR7 were evident in mDC1s, whereas pDCs only showed changes in CCR7 (Fig. 4B). The 16-mg dose (cohort 4) was the first point in the study at which a CRS was recorded. Three of the 4 patients in this cohort experienced a first-dose infusion reaction. Two of these patients progressed through further infusions with no additional adverse reactions, whereas the third patient (subject 10) experienced milder reactions following the third and fourth infusions at 16 mg.

Cytokine responses were evaluated at earlier time points than the flow analysis, sampling 3, 6, and 24 hours after completion of the ChiLob 7/4 infusions. The cytokine responses of the comparable cohorts at 1.6 mg (cohort 2), 16 mg (cohort 4), and 160 mg (cohort 6) also show a significant positive response at the 16-mg dose when the dendritic cell activation and adverse reactions first occur (Fig. 4C). The cytokine signature of the in vivo responses displayed similarities to that of the previous ex vivo observations with significant increases in MIP1β and IL-12 and smaller increases above baseline for MIP1α and IL-8. The kinetics of the responses were also shared across the cohort with an early increase in MIP1β at 3 hours, reaching a maximum at 6 hours and the IL-12 and IL-8 responses appearing at 6 hours. These responses were replicated at the 160-mg dose, and although the amount of cytokine measured varied between individuals, the group of measurable cytokines produced showed a clear pattern with MIP1β showing the most rapid and largest spike at 3 hours and dropping at 24 hours, and IL-12 continuing to increase even at the 24-hour time point. The cytokine profiles were remarkably consistent between patients at the 16- and 160-mg doses despite the introduction of premedication steroids for the latter group. The proinflammatory cytokines associated with a cytokine storm, notably TNFα and IL-6, although measurable in the ex vivo assays with cross-linked ChiLob 7/4, were undetectable in patient samples at both the 16- and 160-mg doses. This result may either relate to a true absence of these cytokines in the CRS of this antibody or be secondary to the introduction of steroids and increased infusion time for those experiencing a first dose reaction at 16 mg and the subsequent introduction of premedication steroids for the cohort at 160 mg. However, the persistence of a reproducible cytokine signature across these 2 cohorts may also suggest that the cytokines being released are not in line with a classical CRS, typically T cell focused with variable FcγR-positive innate cell involvement and are secondary to the targeting of a different cellular subset with a milder adverse reaction outcome.

Although there was no consistent pattern in the kinetics of dendritic cell activation by phenotyping, the individual results from the 4 patients on the highest dose (160 mg) showed comparable fluctuating changes above baseline in both CD83 and CCR7 expression for mDC1s and only CCR7 for pDCs (Fig. 5A). This pattern across a heterogeneous patient group is consistent with our ex vivo results where pDC changes in CCR7 expression were more pronounced than those observed for CD83, which only showed minimal levels of upregulation.

Our results from the multiplex cytokine analysis identified MIP1β and IL-12 as the main cytokines being produced in both the ex vivo activation assays and the patient samples. Both of these cytokines can be produced by activated dendritic cells and we therefore conducted intracellular flow cytometric analysis on samples from healthy subjects to determine whether the activated dendritic cells could produce these cytokines. Following stimulation with LPS, both MIP1β and IL-12 could be identified from a significant number of mDCs (Fig. 5B) and a smaller population of pDCs (Fig. 5C). Activation with cross-linked ChiLob 7/4 induced a smaller response for MIP1β and IL-12 production that was only present in the mDC population and not observed in the pDC population. This finding supports the association of dendritic cell activation, principally of mDCs, with the production of MIP1β and IL-12 that may contribute to the cytokine release profile of ChiLob 7/4. This distinctive cytokine profile may contribute to the proposed immunomodulatory effects of CD40 targeting as well as modulating the severity of a CRS.

Agonistic anti-CD40 antibody has an important role in triggering the activation of adaptive immune cells that can be directed toward an antitumor response. Although other agonistic anti-CD40 antibodies entering the clinic have been shown to activate dendritic cells with the use of model systems (19–21), we have used ex vivo assays to uniquely show biologic activity of ChiLob 7/4 on unmanipulated peripheral dendritic cells and conducted a direct comparison with a first-in-human study. Ex vivo, we have shown that ChiLob 7/4 is most potent when cross-linked and is able to activate dendritic cell to the level induced by the potent TLR ligand LPS. The activation and cytokine profiles for cross-linked ChiLob 7/4 in our ex vivo studies are most reflective of the in vivo setting, suggesting that some form of antibody cross-linking may be occurring in these patients as the antibody is provided to the participants in a soluble non–cross-linked form. This is perhaps unsurprising as CD40 and CD154 each form a trimeric protein structure for successful cell signaling (4), which we can simulate ex vivo when cross-linking via the antibody's Fc receptor. This highlights that although ex vivo assays may be able to predict a clinical effect, knowledge of the antibody–target interaction is critical to ensure the assay design is reflective of in vivo mechanisms to correctly assess the potential biologic activity. In addition, several murine studies have highlighted the importance of understanding the role of FcγR cross-linking for monoclonal antibody function. The agonistic activity of anti-CD40 antibody has already been shown to be dependent on the interaction with FcγRIIB, an inhibitory FcR (22, 23), and appears specific to IgG1 antibodies, which may be providing a structural scaffold to enable stable CD40 signaling. Other studies have also shown CRS elicited by CAMPATH-1 to be isotype dependent, with human IgG1 and rat IgG2b antibodies eliciting the strongest response that could be inhibited with an anti-CD16 antibody (FcRγIII), to block cross-linking with the CD16 FcR (24). There is great interest in understanding fully the role of cross-linking and FcR usage, and how this relates to the different IgG isotypes that may be targeting the same antigen. For anti-CD40 targeting, there are different anti-CD40 isotype antibodies already in the clinic and they may trigger different responses. These functional differences between similar therapeutic antibodies may require novel immunomonitoring assays to establish the profiles and mechanism that might be operating in vivo, together with the risk for different types of CRS. Furthermore, the scope for Fc engineering to enhance the efficacy of existing antibodies provides a promising avenue for improved therapy with reduced cytotoxicity.

Immunostimulatory antibodies that are designed to enhance a person's own immune response against cancer provide a promising alternative when conventional treatments are exhausted (25). However, predicting toxicity of these agents is not always successful. TGN1412, an anti-CD28 superagonist antibody, is perhaps the most well-known first-in-human clinical trial of an immunostimulatory antibody in which preclinical assessments failed to predict an almost fatal cytokine storm (1). Although researchers had assessed the risks, the assay designs had not accounted for all the different antibody interactions and species variability, in this case, the differences in CD28 expression on CD4⁺ effector memory T cells (26), which led to a massive cytokine-mediated inflammatory response. In addition, the dose had been determined using the “no observed adverse effect level” (NOAEL) method that calculates dose based on risk, and where no adverse effect is observed in preclinical studies this dose can be overestimated. Since the TGN1412 trial, the “minimally anticipated biological effect level” (MABEL) has been recommended as a more appropriate basis for determining dose and allows dose-escalation studies to be based on minimal activation (27). In the aftermath of the trial, cytotoxicity was investigated by different researchers using different assay setups and found that the cytokine storm could actually have been predicted using human peripheral blood mononuclear cells with adapted assays (28, 29). The reports that followed highlighted the need to advance with caution when translating immunostimulatory antibodies to the clinic and to better assess cytotoxicity of such high-risk agents as a means not only to predict mechanism of action but also to better predict cytotoxicity before use in clinical studies. In particular, ex vivo assays were identified as a potential prerequisite (30). If appropriate ex vivo assays had been conducted and the MABEL method for determining dose applied, the outcome of the TGN1412 trial might have been very different. Nevertheless, there is still a lack of studies that have shown directly comparable ex vivo results with corresponding clinical outcomes. When showing the effect of ChiLob 7/4 on targeting and activating dendritic cells via CD40, the use of unmanipulated peripheral dendritic cells from human ex vivo whole-blood samples presents various obstacles, including low numbers in peripheral blood combined with limited dendritic cell–specific markers and inherent biologic variability.

Nevertheless, in this study, the ex vivo experiments were workable and confirmed that ChiLob 7/4 could activate both mDC1s and pDCs to the same extent as LPS, but with a reduced inflammatory response as shown by a lower CD83 upregulation and lower production of the inflammatory cytokines IL-6 and TNFα. CD83 expression is often used as a dendritic cell early activation marker; however, our results have shown lower CD83 expression on pDCs following ex vivo ChiLob 7/4 stimulation and negligible levels in the patient samples. This finding suggests that CD83 is perhaps not an appropriate marker for pDC activation. Changes in CCR7 expression were measurable and possibly represent a more preferable marker, the role of which in dendritic cell maturation is better understood because of its involvement in chemotaxis and as an indicator of early migration (31). Upregulation of CCR7 in the ex vivo assays would suggest that ChiLob 7/4 is effective in stimulating dendritic cell maturation toward a migratory phenotype, the expression of which was reflected in patient samples. Differences in the upregulation of activation markers are perhaps expected, as the two subsets are known to differ in CD40 expression (32). The proportion of mDC-expressing CD40 is greater than pDCs as well as the individual expression level of CD40, which is higher on the mDC population. This may contribute to the reduced cytokine responses from the pDCs in our study. In addition, different ligands can activate dendritic cells to produce a different dendritic cell outcome (33). This is shown in our study with the pDC population, in which LPS activates the upregulation of CD83 and CCR7 whereas cross-linked ChiLob 7/4 only upregulates CCR7; LPS induces MIP1β and IL-12 production in pDCs but cross-linked ChiLob 7/4 does not.

Cytokine production was also only detectable when ChiLob 7/4 was cross-linked. Two of the cytokines that are often associated with a cytokine storm (TNFα and IL-6) were not detectable in samples from ChiLob 7/4-treated patients. The onset of a cytokine storm is usually fast and immediate, and causative cytokines are measurable within the first 24 hours (25). We used patient samples that were taken at early time points (3–24 hours postinfusion) to monitor for the presence of any immediate cytokine storm. Generally, there was more cytokine present at these earlier time points and these had disappeared at the later day-4 time point. The pattern of cytokine expression in the first 3 hours postinfusion was directly comparable with that seen in the ex vivo 4-hour stimulation assays with the exception of IL-6 and TNFα. It is possible that the time points for the clinical samples may have been after the peak expression of IL-6 and TNFα that may have been during the infusion period, with cytokine being produced and dissipated by the time of the postinfusion blood sampling, in line with other studies showing acutely transient plasma cytokine responses poststimulation (34). The constraints of the clinical study in which the drug infusion is undertaken over 6 hours at the higher dose, compounded by the reduction in infusion rate for those who experienced a CRS, may have contributed to a failure to detect some cytokines. The lack of TNFα and IL-6 in patients may also be explained by the use of corticosteroids that may have dampened the production of cytokines, although MIP1β, MIP1α, IL-8, and IL-12 were still produced in cohort 4, whose subjects were treated with steroids and cohort 6, whose subjects were prophylactically premedicated with steroids. The fact that MIP1β and IL-12 are still produced is particularly encouraging, as both cytokines are known to drive T_H1 responses and suggests a distinct controlled cytokine response in these patients. The effect of corticosteroids on specific cytokine response is an area that clearly requires further investigation. In our study, the 16-mg (3.2 μg/mL) approximates to the 10-μg concentration we used in the ex vivo evaluations in which we first observed a cytokine response following cross-linking of ChiLob 7/4. The emergence of the first symptoms relating to a CRS would be compatible with this approximate equivalence between ex vivo and in vivo evaluations and is supportive of our original starting dose, which was 80-fold less (0.5 mg) than this first adverse event–related dose. At the 160-mg dose, the cytokine responses for MIP1β are lower than those seen at 16 mg, despite the ex vivo experiments showing a significant increase over this 10-fold increase, and this may reflect the introduction of prophylactic corticosteroids to this cohort.

Flow cytometric results proved to be a sensitive method for detecting dendritic cell activation, with changes in CD83 and CCR7 being measured in mDC1s without the need for cross-linking, whereas cytokines were only detected when cross-linked. In addition, in vivo changes in dendritic cell activation markers were seen at a lower ChiLob 7/4 dose before reaching a dose able to produce detectable cytokines. Specifically, patient samples taken at corresponding days 1 and 4 postinfusion time points for dendritic cell flow cytometry and cytokine analysis showed measurable changes in dendritic cell activation markers at a 10-fold lower dose (16 mg) before any measurable cytokine response at a 160-mg dose. Together, these results suggest that lower stimulation can trigger upregulation of dendritic cell activation markers, but stronger stimulation is required to trigger cytokine production.

Together, the results of our study show that ChiLob 7/4 is biologically active and can trigger human peripheral dendritic cell activation and maturation with the release of T_H1 cytokines but without the presence of a classical cytokine storm. We have also shown that in combination with other dendritic cell ligands such as LPS and CpG, there is no augmented response. Interestingly, CP-870-893, another anti-CD40 agonistic antibody, has been shown to have augmented B-cell activation in combination with CpG TLR9 stimulation (35). This IgG2 antibody has also been reported to produce TNFα and IL-6 in clinical studies (11), suggesting that this antibody may be working in a different way than ChiLob 7/4. Finally and most importantly, we show that ex vivo assays provide a useful preclinical tool when correctly applied to help determine the effectiveness of an antibody on their target cell (i.e., ChiLob 7/4 on dendritic cell activation) in addition to monitoring for the presence of a cytokine response and any potential cytokine storm. The ability of ChiLob 7/4 to trigger “safe” dendritic cell activation and cytokine release provides a rationale for subsequent assessment in combination with other clinical agents that may trigger specific effector responses to tumor cells.

M.J. Glennie has served as a consultant/advisory board member and has provided expert testimony. No potential conflicts of interest were disclosed by the other authors.

Conception and design: F. Chowdhury, P.W. Johnson, M.J. Glennie, A.P. Williams

Development of methodology: F. Chowdhury, P.W. Johnson, M.J. Glennie, A.P. Williams

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. Chowdhury, P.W. Johnson, A.P. Williams

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Chowdhury, P.W. Johnson, A.P. Williams

Writing, review, and/or revision of the manuscript: F. Chowdhury, P.W. Johnson, A.P. Williams

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F. Chowdhury

Study supervision: F. Chowdhury, P.W. Johnson, A.P. Williams

The work was conducted with funding from the Experimental Cancer Medicine Centre (ECMC), Cancer Research UK (CR-UK).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.