Considerations for the Characterization and Interpretation of Results Related to Alternative Complement Activation in Monkeys Associated with Oligonucleotide-Based Therapeutics

Introduction

Intravenous (IV) administration of phosphorothioate oligonucleotides (PS ONs) to non-human primates has been widely reported to cause activation of the alternative complement pathway [1,2]. Understanding the nature predictive validity of these nonclinical observations to human safety is of paramount importance. Several investigators have used nonclinical models and clinical experience to assess the nature of the response, its molecular mechanism, downstream sequelae, clinical significance, and methods of monitoring it in laboratory species and humans.

The purposes of this communication are to (1) provide a brief review of the industry's experience with ON-induced complement activation in monkeys, (2) provide advice regarding general approaches for characterizing complement activation in monkey studies, and (3) propose an overall strategy for the standardization of analytes and time points for monitoring complement activation in both nonclinical and clinical studies.

Molecular Mechanism

The acute, transient, and plasma concentration-dependent activation of the alternative pathway complement (APC) cascade in cynomolgus or rhesus monkeys by PS ONs is well documented [3]. Available evidence indicates that PS ONs activate the APC within the blood compartment as a consequence of their interaction with soluble Factor H, a glycoprotein that exerts critical negative regulation on the APC cascade [4,5]. High plasma concentrations of PS ONs have been correlated with decreased plasma concentrations of Factor H and activation of the APC cascade in monkeys.

Presumably, the polyanionic structure of the PS ON binds to Factor H in circulation, but it is unclear whether the apparent decrease in serum Factor H reflects interference with the binding of the antibody in the ELISA with the protein or a temporary sequestration of the glycoprotein. Regardless, the interaction of PS ONs with Factor H interferes with its inhibitory function on the alternative pathway and allows amplification of the proteolytic cascade, which leads to production of the alternative pathway-specific split product, Bb, and the downstream anaphylatoxins C3a and C5a.

ON-induced activation of the APC cascade is dependent upon plasma ON concentration and results in reliable concentration/effect relationships that can be used to set threshold concentrations for activation limits. The effect of a PS ON on APC is typically transient, as rapid clearance of the ON from plasma permits restoration of the regulatory function of Factor H.

It is generally believed that PS ON-induced activation of the APC cascade depends upon achieving a threshold plasma concentration. For example, when a 20-nucleotide long PS ON is administered for 1–2 h by IV infusion, the threshold concentration for activation has generally been reported to be in the range of 50–70 μg/mL, although some PS ONs may induce activation at lower plasma levels. The critical plasma PS ON concentration for APC activation is relatively independent of the nucleotide sequence, but is influenced by both ON chemistry and the infusion rate. For most of the 2′-MOE or 2′-Me, and locked nucleic acid (LNA)-modified PS ONs, the concentration–response curve shifts to the right, with thresholds for some of these ONs ranging from ∼70 to >100 μg/mL [6,7]. This has been attributed to the lower degree of protein binding in ONs with 2-alkyl modifications. The threshold is also influenced by the circulating time of the ON in the blood compartment. This is manifested in lower threshold plasma concentrations associated with either subcutaneous (SC) administration, where the peak plasma concentration occurs ∼3–4 h after injection, or longer-duration IV infusion.

General Investigational Strategy

Evaluation of the potential for ONs to initiate APC activation has become routine in monkey toxicity study protocols for candidate therapeutic ONs. Most investigations of APC activation with PS ONs have been performed with the objective of characterizing the plasma concentration–APC response relationship. Therefore, it is important to understand the plasma pharmacokinetics of the ON associated with the particular route and method of administration, to enable sample collection for analysis of complement split products at time points that coincide with the peak plasma ON concentrations, and at later time points to document recovery of the complement system (regeneration of split products) after the activation has subsided.

Assessment for APC activation should include sampling from control monkeys at the same time points as test article-treated animals, mainly because the stress associated with restraint for dosing and sample collection can produce increases in Bb over time that are independent of drug treatment but could be interpreted as treatment-related in the absence of data from a parallel control group. The stress effect on Bb is particularly prominent when monkeys are restrained for IV infusion of a test article, and there can be two to threefold increases in mean Bb levels associated with chair-restraint for an hour or longer, or if the animals are repeatedly briefly restrained for blood sampling and other procedures. It is also important to place freshly collected samples on ice (if not processed immediately) and to process blood samples to plasma under refrigerated conditions (eg, using a refrigerated centrifuge) to minimize ex vivo activation of the APC.

As mentioned, this testing has been done almost exclusively to date in monkeys, owing to the fact that the rodent complement system is markedly different, such that activation by PS ONs does not occur in rodent species (see Species Specificity and Clinical Relevance section for further discussion). Commercially available ELISA and RIA kits are available for human complement split products that have been validated for the measurement of monkey APC activation [4]. To the Subcommittee's knowledge, there have been very few investigations of complement activation in other non-rodent species. In fact, the propensity for PS ONs to induce complement activation in monkeys is one of the main reasons why this species is so widely employed for safety assessment of this therapeutic [8].

Complement activation is largely independent of the frequency of administration. The plasma kinetics of PS ONs are very similar from one dose to the next, and hence, the potential to activate the APC cascade does not change with repeated ON dosing. There is some depletion of the complement protein factors (eg, C3) from plasma after activation, but these proteins are typically fully resynthesized within 48–72 h, and the consistency in activation after the first or last dose in studies up to 13 weeks has been demonstrated. Upon repeated administration of PS ONs, no cumulative effect on complement activation is observed, nor is sensitization apparent. The APC can be activated to a similar degree with each dose, regardless of whether doses are given daily or less frequently (eg, weekly).

That being said, with more chronic duration of treatment, there does appear to be a progressive decrease in plasma C3 with some ONs. This is associated with lower peak split product production (less protein to be cleaved), and a gradual increase in the predose levels of some of the split products (possibly reflective of more sustained activation). This apparent depletion of the complement pathway, presumably due to the chronic activation by repeated ON treatment, is a recent observation that has implications for the conduct and interpretation of chronic monkey studies.

Because PS ON-induced APC activation results from a direct interaction between the ON and a plasma protein (Factor H), cell-based in vitro assays cannot model the in vivo response. PS ONs can, however, initiate the APC cascade in whole monkey blood or serum. Such ex vivo models yield reproducible assessments of concentration-dependent increases in the specific split products (eg, Bb and C3a), reflecting alternative pathway activation, whereas formation of the primary split product representing classical pathway activation, C4a, is not induced. While, these ex vivo results are consistent with the selective activation of the alternative complement pathway observed in in vivo studies in monkeys, the concentration-effect relationship is often different, presumably due to the differences between a static biochemical assay compared to one with active clearance processes.

Assessment of the utility of serum-based assays is in progress. Various investigators have employed ex vivo models to study species sensitivity, biochemical mechanisms, and structure-activity relationships. As more information is collated, the reliability and predictive validity of ex vivo screening will be determined. Should there be sufficient quantitative correlation between ex vivo and in vivo models, it may be useful for ONs of novel chemistries to be assessed via ex vivo screening before the initiation of in vivo studies. Significantly, no concentration-dependent activation of APC has been observed in ex vivo studies with human whole blood (see Species Specificity and Clinical Relevance section).

Structure Activity Relationship

Several distinct structural classes of therapeutic ONs have undergone standard toxicity testing in monkeys, including the assessment of APC activation. These include single-stranded DNA ONs with phosphorothioate-modified linkages, single-stranded DNA ONs with PS linkages and 2′-alkyl modifications (ie, 2′-O-methyl, 2′-O-methoxyethyl, and LNA), double-stranded RNA molecules that typically do not have PS linkages, ON aptamers with a mixture of backbone chemistries and modifications, CpG-containing PS ONs, and double-stranded DNA, often containing full-length PS modification of both strands. While initiation of the monkey APC cascade has been observed with almost every variety of ON, the structural characteristics that modulate the magnitude of the effect differ among the structural subclasses. The use of complex delivery formulations also has a major impact on the propensity for ONs to activate the APC. One type of ON structural modification that has been shown to have no potential to induce APC activation is the morpholino backbone, which has a neutral charge and generally appears to not interact with cationic sites on proteins [9].

The potential for an ON to initiate the APC cascade in monkeys is correlated with the nonspecific protein-binding character of the molecule. This is dictated, at least in part, by the polyanionic nature of the ON backbone. Although charge is an important determinant of protein binding potential, the phosphorothioate content significantly contributes to the non-specific protein binding. ONs with greater total protein binding generally exhibit a greater potential for binding to Factor H and initiation of the APC cascade.

PS modifications are introduced to stabilize ONs against nuclease degradation, and this structural alteration results in a relatively high degree of plasma protein binding (typically greater than 90%), with a concomitant high propensity to initiate the APC cascade. Conversely, ONs that do not contain PS linkages exhibit much lower plasma protein binding and a greatly reduced potential for APC activation. Similarly, PS ONs with additional 2′-alkoxy modifications (eg, 2′-O-methoxyethyl) exhibit reduced plasma protein binding potential and a lower propensity for APC activation.

For single-stranded PS ONs, complement activation, along with total protein binding, is directly correlated with ON length (Fig. 1). The majority of ONs that have progressed into development in this class have been 19–21 nucleotides or longer, and these ONs consistently cause APC activation in monkeys. More recent in vivo investigations with shorter PS ONs (12–16 nucleotides long) revealed a lesser potential for activation with reduced chain length, and this difference has been confirmed by in vitro studies in monkey serum (Isis Pharmaceuticals; Unpublished Data). This decrease in APC activation appears to be simply attributable to lower overall protein binding.

OPEN IN VIEWER

FIG. 1.

Alternative complement pathway activation in monkey serum is dependent on ON length. Monkey serum was incubated 30 min with increasing concentrations of ON. Compounds examined were 2′_MOE ASO targeting B-glactosamine with length ranging from 14–24 nucleotides. Sequences had common sequence of 14 nucleotides in the center or incrementally longer seqeunces on the 3′ and 5′ ends. Complement activation was determined by C3a ELISA. ON, oligonucleotide.

Effects of Nucleotide Sequence

For “first-generation” 20-nucleotide PS ONs (ie, those with a full-length PS backbone and no other modifications), the plasma concentration–response relationship for APC activation generally appears to be similar among ONs of different nucleotide sequences, as long as the route and mode of administration are the same. Indeed, for the first few deoxy PS ONs studied, the threshold plasma concentration for activation was remarkably similar [7]. Even specific sequence motifs, such as the CpG dimer, did not appear to have much influence on the APC activation by PS ONs. This presumably reflects that, while the CpG sequence increases specific interaction with TLR9, it has little influence of the nonspecific protein-binding or Factor H-binding properties.

However, with the introduction of greater structural diversity among PS ONs, including 2′-methyl, 2′-MOE, or LNA modifications, more heterogeneity with respect to the potency for APC activation has been observed. The structure–activity relationships for ON-induced complement activation have not been fully elucidated, but there is clearly more diversity among ONs than was evident from studies with first-generation PS ONs. Because of this collective variability related to length, chemistry, and nucleotide sequence, it remains important to thoroughly characterize the complement activation potential for each newly advanced PS-containing ON.

Double-Stranded RNA Oligos

Where the relatively new class of double-stranded RNA-silencing ONs falls in the spectrum of APC activation is complicated by the widespread use of specialized delivery formulations for many of these programs. There is little information on the potential for double-stranded RNA (eg, siRNAs) to activate the APC. Most of the ONs in this category, particularly those that are delivered in complex formulations, have little or no PS content, and hence, would not be highly suspect as an APC-inducing ON. In fact, for some of these programs, complement activation has been shown to be a property of one or more of the formulation excipients, that is, attributable to the cationic lipids that are commonly used in lipid nanoparticle formulations. In this case, the complement activation appears to involve the classical pathway and is unrelated to the APC activation induced by PS ONs. Hence, it will not be discussed further herein, as this is not an ON-specific issue.

For other types of RNA ONs that contain PS modifications or other structural alterations that are known to (or are expected to) confer nonspecific protein binding, the same concerns and considerations apply about testing for complement activation, particularly if the ON is not presented in a complex formulation that would be expected to minimize interactions with blood proteins.

Species Specificity and Clinical Relevance

Monkeys (cynomolgus and rhesus) are by far the most widely used non-rodent species for safety assessment of ONs, despite the fact that this species appears uniquely sensitive to ON-induced APC activation.

Available data support the conclusion that the APC cascade is not initiated in mice, rats, or guinea pigs by PS ONs when administered as IV bolus or SC doses up to 50 mg/kg and higher (Isis Pharmaceuticals; Unpublished Data). In fact, the IV LD50 for most PS ONs in rodents (mice and rats) is greater than 800 mg/kg, which suggests that complement activation does not occur even at much higher doses in these species. As of yet, there are no data in rabbit or minipig that are known to the Subcommittee.

While dogs have been used to a very limited extent for safety assessment of ON, the available experience indicates they are substantially less sensitive to ON-induced APC activation than are monkeys (Isis Pharmaceuticals; Unpublished Data). No decrease in total hemolytic complement (CH50) has been observed in dogs after SC administration of several different ON sequences at doses up to 30 mg/kg. In one anecdotal example (Kornbrust; unpublished information), mortality occurred in dogs after IV (bolus) administration of a PS ON at single dose of 80 mg/kg and above; however, it was not confirmed that the cause of mortality was related to complement activation, although the clinical signs were suggestive.

Given this remarkable difference in the sensitivity among species, the question of which is the appropriate model for human sensitivity is important. Monkeys are certainly hyper-responsive relative to the other commonly used laboratory species. As such, characterization of the concentration-effect relationship for APC activation in monkeys has afforded safety to subjects and patients during the conduct of clinical trials. Fortunately, APC activation is something that can be readily monitored by the measurement of plasma split products in clinical trials. In general, clinical safety has been ensured by managing dose escalation such that plasma ON concentrations remain below the threshold for complement activation determined in monkeys. However, in some oncology trials, the doses have been escalated into a range that approaches or exceeds the threshold levels for complement activation characterized in monkeys. To date, there has been generally no evidence of complement activation under these conditions or in other types of clinical trials involving short IV infusion or SC injection of PS ONs [10–18]. The only report of complement activation associated with ON treatment in clinical trials was in an oncology trial that used a 24-h IV infusion of very high doses ≥18 mg/kg. Based on these data, it appears that humans are not more sensitive to oligo-mediated APC activation, and may actually be less sensitive than monkeys. As described previously, data from the serum activation assay suggest that humans, like the majority of experimental animal species tested, are less sensitive to oligo-induced APC activation than are monkeys [3]. In that context, and considering that complement activation may be dose-limiting in monkeys, it is reasonable to question whether macaques are the most appropriate species for nonclinical safety assessment of PS ONs.

Influence of Route of Delivery and Mode of Administration

Activation of the alternative complement pathway by PS ONs has been observed following IV and SC administration, as both delivery routes can result in high blood concentrations of ONs. As discussed above, time may also be a component of the concentration–effect relationship (ie, the potential for complement activation is dependent on both the peak plasma ON concentration (Cmax) and the duration above a particular concentration). While a short exposure to a high plasma concentration of PS ON is associated with significant inhibition of Factor H, a long duration of exposure to low plasma ON concentration (eg, as would occur with SC injection) also results in APC activation, although at a lower magnitude. Thus, the threshold plasma concentration for complement activation is dependent on duration of exposure in plasma and will be lower for a longer duration of infusion or with SC injection. Over time, this “smoldering” activation may result in accumulation of more stable split products, such as Bb, but typically not the labile split products such as C5a.

To the Subcommittee's knowledge, complement activation by PS ONs is not observed with administration by routes other than IV and SC, as would be predicted from the low plasma ON concentrations associated with other routes of administration. Consideration has been given to the localized activation of complement as complement proteins are present in various tissues and fluids, including vitreous humor (in the eyes) and cerebrospinal fluid. However, to date, there has been no evidence of local complement activation at the site of administration. This includes local SC or intradermal injection where the dose can be quite large, but there has been no evidence of activation as determined by local complement C3 deposition. Local inflammatory changes are often observed (ie, cellular infiltration), but this is attributed to stimulation of local innate immune cells (ie, most likely mediated by interactions with Toll-like receptors and/or related proteins). Thus, complement activation by PS ONs appears to be a blood-compartment event, rather than a reaction occurring at local sites.

Consequences of ON-Induced Complement Activation

The downstream sequelae from APC activation in monkeys range from no consequence to pronounced hemodynamic disturbances and death. The intensity of the APC activation and the resulting formation of specific split products with various biological properties is the primary determinant of the magnitude of these effects. It is the C5a split product, formed as the proteolytic cascade proceeds down the common pathway, which is believed to be the main mediator of adverse sequelae. It binds to receptors on neutrophils and causes them to rigidify and adhere to endothelium (a state of activation of these cells), C5a can stimulate other cells to secrete cytokines and other vasoactive agents, and it is known to be directly cardiotoxic [19]. C5a is a split product with a very short half-life and typically does not increase in plasma concentration unless there is burst of complement activation cascading down the common pathway. In the experience some of the Subcommittee members, there is a high correlation between increases in C5a that can be seen with appropriately timed plasma sample collection and the occurrence of hemodynamic disturbances in monkeys. In the absence of C5a accumulation, there are typically no acute adverse consequences to the APC activation induced by PS ONs, although there may be several-fold increases in plasma C3a or Bb concentration. While C3a is generally regarded as an anaphylotoxin, its much lesser potency for mediating hemodynamic and cardiovascular disturbances, relative to C5a. Some members of the Subcommittee have observed that increases in C3a are not well correlated with acute hemodynamic disturbances in monkeys. Therefore, while monitoring of C3a may have utility for monitoring for pathway activation similar to Bb, it is not considered the best marker for predicting the acute hemodynamic effects of complement activation in monkeys.

Thus, if the complement system is intensely activated by a PS ON (eg, with bolus injection or short-duration infusion), the sudden burst of split product formation, particularly C5a, may lead to adverse sequela [20]. However, with long-duration infusion or non-IV routes of dosing, plasma concentrations may rise gradually, and the complement system is less intensely activated, such that only those relatively stable split products with little or no biological activity (such a Bb) will accumulate. Under such conditions, C5a may be formed, but at a low rate, such that it does not accumulate to an extent that elicits adverse effects. A repeated low level of complement activation can lead to gradual consumption of complement factors without accumulation of biologically active split products. Hence, with continuous infusion, monkeys may be effectively “decomplemented,” which could be adverse if this translates into an impairment of innate immunity, but it is otherwise uneventful. As mentioned above, once activation has subsided (eg, after infusion), the levels of complement proteins are regenerated within 24–72 h.

Although not documented in the published literature, the collective experience of the Subcommittee members attests to an apparent attenuation of the severity of hemodynamic disturbances in monkeys with repeated dosing of PS ONs. The nature of this diminished response appears to be unrelated to formation of split products and appears to reflect some form of desensitization to the impact of those split products. The mechanism for this change over time is unknown, but there is anecdotal evidence for downregulation of C5a receptors on neutrophils as one possible means by which the downstream sequelae may be attenuated with repeated ON dosing. Thus, the risk for severe hypotension and mortality in monkey studies stemming from complement activation may be greatest upon the first dose and generally diminishes thereafter, although there is certainly a continued risk, particularly if the dosing is highly intermittent, to an extent that allows full recovery of all components of the response.

Recommended Strategy

Based on the above understanding, the strategy that is recommended for assessment of complement activation of PS ONs, both nonclinically (ie, in monkeys) and clinically, involves collection of blood samples (typically in di-potassium EDTA, processed to plasma) at appropriate time points, including a predose sample and one or more postdosing samples, with the key sampling time point coinciding with the expected Cmax (ie, the time of peak plasma concentration). It may also be of interest to collect one or more additional samples to characterize the recovery of split products and verify the acute nature of APC activation, although this is well understood for PS ONs. As described above, complement activation by PS ONs is a first-dose effect, and so blood sampling for measurement of split products can be scheduled upon the first dose in a monkey toxicity study or clinical trial. It is prudent to also collect samples upon the last dose in a toxicity study, and during initial clinical trials in normal subjects or patients, to verify the consistency of the response in monkeys or the continued absence of a response in subject/patients.

The collected plasma samples should be analyzed for Bb, at a minimum. If an elevation in plasma Bb is observed, it would be logical to then measure C5a (ie, in remaining plasma or in a separate sample), to address the biological significance. As discussed above, elevations in Bb in the absence of any accumulation of C5a are of questionable toxicologic significance, as this likely represents a slow acceleration of alternative pathway split product generation without sufficient common pathway involvement to yield increased levels of biologically active split products.