The first therapeutic monoclonal antibody (mAb), muromonab, was licensed in 1986 for the treatment of acute allograft rejection in renal transplant patients, but it wasn’t until the late 1990s that the potential of therapeutic mAbs began to be realized. Following the licensure of muromonab, 12 mAbs were approved in the 1990s; 7 therapeutic mAbs (2 for oncology) and 5 imaging mAbs (4 for imaging tumors to determine the location and extent of the tumor).

The first therapeutic mAb for an oncology indication was rituximab, approved in 1997 for the treatment of patients with relapsed or refractory low-grade or follicular, B-cell non-Hodgkin’s lymphoma. Next was the approval of trastuzumab in 1998 for the treatment of patients with metastatic breast cancer whose tumors overexpress the HER2 protein and who have received one or more chemotherapy regimens for their metastatic disease. Since their original approvals, both products have added additional indications, including autoimmune indications for rituximab, and both have had a major impact in improving outcomes in the oncology indications.

The early failures of most mAbs were attributed to insufficient characterization of the specificity and function or intended mechanism of action (MOA) of the mAb, poor preclinical development, inadequately designed clinical trials and poor target selection (1, 2). In addition, the first mAbs used clinically were derived from murine hybridomas, which have a short half-life in humans, are inefficient at eliciting effector functions, and induced human anti-mouse antibodies (HAMA) (3, 4).

The lessons from these early failures led to the genetic engineering of chimeric mAbs, murine or other non-human variable regions (V region) expressed with human constant regions. This was quickly followed by the development of humanized mAbs, which contain murine or other non-human complementarity determining regions (CDRs) grafted onto human framework regions of V region genes expressed with human constant regions. In direct contrast to their murine counterparts, chimeric and humanized mAbs are predicted to be less immunogenic (5), exhibit longer half-lives, and efficiently promote effector functions in humans (3, 6, 7).

Subsequently, the development of mAbs containing entirely human-derived sequences was facilitated by two approaches: the engineering of transgenic mice to express germline human heavy and light chain genes (8–10); and the development of display technologies to express human antibody genes derived from human donors or constructed synthetically based on human germline gene sequences and the frequency of their use (11, 12). Human mAbs share the advantages of chimeric and humanized mAbs regarding Fc-mediated effector functions and half-life and are predicted to be less immunogenic than chimeric and humanized mAbs. Today, many anti-viral mAbs and a few mAbs directed against human targets are isolated directly from human B cells (13, 14). With B cell repertoire profiling in tumor microenvironments (TME) and in some cases expressing antibodies that contribute anti-tumor activity, isolation of B cells from cancer patients may become a source of novel mAbs for oncology indications (15–17).

Most approved mAbs use an IgG isotype, although several are antibody V region fragments (Fab, F(ab’)₂, scFv, single domain abs). V region engineering progressed beyond reducing immunogenicity to include the optimization of affinity, developability and manufacturability (7, 18, 19), and to extend half-life (20). Studies early in product development help identify potential amino acid sequences that could result in unfavorable quality attributes, such as aggregation-prone motifs, immunogenic motifs and amino acids prone to post-translational modifications, including V-region glycosylation (7).

IgG1 is used if the MOA includes Fc-mediated effector functions, whereas IgG2 and IgG4 are used when effector functions are not desired. One of the earliest types of Fc engineering is the S228P mutation in the IgG4 hinge region to prevent Fab arm exchange with endogenous IgG4. Gemtuzumab ozogamicin, which was developed in the 1990s and initially approved in 2000, was the first approved IgG4 mAb and incorporated this mutation (21).

Amino acid residues in IgG1 that bind to complement, FcγR or the neonatal Fc receptor, FcRn, have been identified and mutated in mAbs to achieve the desired outcome. Some mutations enhance binding to FcγRIIIA for improved antibody-dependent cellular cytotoxicity (ADCC) and/or to FcγRIIA to improve antibody-dependent cellular phagocytosis (ADCP) activities, while other mutations that reduce binding to FcγRs are used when effector functions are not wanted (22–25). Cell line engineering is also used to produce afucosylated glycan structures to enhance ADCC (26).

Some mutations may increase binding to all variants of FcγRII and FcγRIII, while others may increase binding to FcγRIII but decrease binding to FcγRII, particularly the inhibitory receptor, FcγRIIB (24). Margetuximab, an anti-HER 2 mAb approved in 2022, was Fc-engineered to enhance binding to FcγRIIIA and reduce binding to the inhibitory receptor, FcγRIIB (27). It did not have an overall survival advantage over trastuzumab treated HER2-positive advanced breast patients but did show an advantage in the subset of patients who were homozygous for the lower affinity FcγRIIIA-158F variant (27, 28). Additional pre-clinical and clinical studies will be needed to understand the lessons to be learned from the Phase III clinical trial.

Increasing binding to FcγRIIB has been shown to improve the performance of agonist immunostimulatory mAbs, likely due to cross-linking agonist mAb bound to its target on the cell surface (29, 30). Many reviews provide a more in-depth discussion of isotype selection, Fc engineering approaches for FcγR, FcRn, or complement binding and the mechanisms by which these modifications achieve the desired outcome to promote efficacy and safety of the mAb (22–25, 31, 32).

Many approved products are no longer marketed; however, by the end of 2023, over 160 novel and biosimilar mAbs were approved by the FDA ( Figure 1 ). Approximately 42% of the approvals are for oncology indications: 4 imaging agents, 12 antibody-drug conjugates (ADCs, including one bacterial toxin conjugate), 9 bispecific antibodies (BsAbs), one co-formulated mAb combination, and 13 biosimilar mAbs. Several of the approved products incorporate Fc-engineering or are expressed in glyco-engineered cell lines. While most mAbs for oncology indications are delivered by intravenous infusion, several products have been reformulated with recombinant human hyaluronidase for subcutaneous injection. These products allow a shortened administration time compared to the time needed for intravenous infusion (33).

Since a rational design can be applied to antibody-based products (34, 35), developers should formulate a quality target product profile (QTPP) early in development that considers the intended clinical use, dosage form and strength, route of administration, delivery systems, and critical quality attributes (CQAs) that impact sterility, purity, stability, safety, efficacy, and pharmacokinetics. A CQA is defined as “a physical, chemical, biological, or microbiological property or characteristic that should be within an appropriate limit, range, or distribution to ensure the desired product quality.” (36).

It is expected that the MOA of mAb based products will be characterized to the extent possible. All mAbs, including ADCs, BsAbs and mAb-fusion proteins must bind their target. Characterization of a mAb should determine the affinity for the target, including on- and off-rates, and binding of the homolog from other species for non-clinical assessments. Depending on the antigen, antigen density and indication, the affinity for antigen can influence effector function (37) and affect localization and tumor penetration (38, 39). In some cases, reducing affinity of mAbs may result in fewer or less severe adverse events, such as the anti-CD3 arm in BsAbs (40, 41), or enhance activity of immunomodulatory products (42).

All mAbs, including ADCs, BsAbs and mAb-fusion proteins, should be assessed for their ability to carry out Fc-mediated effector functions (43). This can be done once to demonstrate the effectiveness of the engineering for mAbs with low to no effector function, but for mAbs intended to have effector function, assays to assess this activity should be performed routinely.

One of the biggest gaps in knowledge for mAbs with Fc-mediated effector functions is demonstrating a specific effector function in vitro and understanding what the predominant mechanism may be in vivo for each indication. In addition, some adverse events, such as first infusion reactions, may be related to engagement of FcγR on neutrophils (44), however infusion reactions can be managed.

In patients, the predominant Fc-mediated mechanism will depend on the effector cells at the site of the tumor, the FcγRs expressed on those effector cells, or if tumor cells express complement inhibitory proteins. It is also dependent on the epitope of the target antigen: epitopes closer to the membrane favor ADCC activity while epitopes further away favor antibody-dependent cellular phagocytosis (ADCP) activity (45, 46). Antigen density (37, 47), as well as the angle of binding the antigen may influence MOAs, which has been shown for rituximab and obinutuzumab (48, 49). Optimization of the Fc-region for a specific oncology indication should consider the target (tumor antigen versus immunoinhibitory or immunostimulatory receptors) and the types of effector cells and the FcγRs they express in the TME (31, 32).

Despite this uncertainty, it is important to distinguish potential in vivo MOAs from assays used for quality control (QC) of a product. The MOA of a mAb and features that may impact binding to its target, such as glycosylation (50) and other post-translational modifications (PTMs) in the CDRs (51–53) and constant regions (54), are considered CQAs. The absence of an Fc-glycan reduces complement dependent cellular cytotoxicity (CDC) activity and abrogates FcγR-mediated activity (55), but more detailed knowledge of the impact of specific glycan structures on Fc-mediated effector functions, specifically afucosylated glycans, became known after the approval of several mAbs, including rituximab, trastuzumab, alemtuzumab and cetuximab (56–59). It is well established that afucosylated glycan structures have enhanced ADCC activity because afucosylated IgG1 binds better to FcγRIII than fucosylated IgG1 (58) and that galactosylation, high mannose and sialylated forms also contribute, but not to the same extent as afucosylated species (60–67). However, there is likely still more to learn about the effect of specific glycoforms on ADCC and other Fc-mediated effector functions.

If a mAb has multiple Fc-mediated effector functions, or if the predominant mechanism is unknown or may differ per indication, it may not be feasible to assess all mechanisms for product release and stability testing. The purpose of QC testing is to ensure the consistency of each lot of a product. Therefore, a potency assay or assays for release and stability that reflect a relevant MOA, if not the predominant mechanism, should be developed and all attributes that are thought to affect the MOA should be adequately controlled in this assay or by other methods. Historically, complement dependent cellular cytotoxicity (CDC) assays were used since cell lines were not available for ADCC assays and using effector cells isolated from donors is challenging for QC purposes. CD16-expressing cell lines or surrogate reporter-gene ADCC assays became available in the last 10-15 years, so now most mAbs with effector function include an ADCC assay for release. ADCP assays are still challenging but improving. Many are in development, including reporter gene surrogate assays (68–70).

Using an FcγRIIa surrogate assay, Kuhns et al. (70) also showed that terminal galactosylated, afucosylated and high mannose glycoforms have a different or no impact on ADCP activity, depending on the IgG1 mAb and target antigen. In contrast, the effect of different glycoforms on ADCC activity appears to apply across IgG1 mAbs, regardless of target. More work is needed to fully understand the relationship between Fc-glycan structures and ADCP. Furthermore, since there appears to be opposite effects of some glycoforms on ADCC and ADCP activity, for each mAb it is important to understand which mechanism is likely the predominant activity for a given indication and to optimize the glycoforms for each mAb that may favor one activity over the other or optimize the glycoforms to adequately promote both activities.

From a regulatory perspective, based on current knowledge that ADCC activity varies with levels of fucose on the N-glycan structure, ADCC assays are considered the most sensitive to detect differences among lots. Methods that assess binding to FcγR, glycan structures and other relevant PTMs can also be used to control CQAs relevant for potency. As more knowledge is gained about the relationship between glycan structures and other effector functions, regulatory expectations may change depending on the predominant predicted MOA for a mAb.

IgG (IgG1 >> IgG4 > IgG2) is the main antibody isotype of approved and clinical stage products. However, there are different isotypes, different types of V region only constructs and various formats of bi- and multi-specific antibodies that lack an Fc region. Engineering IgG and V region fragment-based products beyond the standard mAb format includes more complex structures, such as BsAbs, ADCs, imaging agents, and bifunctional fusion proteins (71).

Antibodies are inherently immunogenic as suggested by the anti-idiotype network theory (142). The immunogenicity of mAbs is due to both patient and drug related causes (143). Patient-related causes include the nature of the disease, genetic background and concomitant medications. Drug-related causes include the mAb origin and engineering design, target antigen, post-translational modifications, impurities, formulation, dose, route of administration, and frequency of dosing. MAbs of rodent or other non-human origin are generally the most immunogenic, followed by chimeric, humanized and human mAbs.

Immunogenicity cannot be directly compared across products because the ADA assays are specific for each mAb. In addition, today’s assays are more sensitive compared with the ADA assays used 20 years ago and can better detect ADA than older methods. However, given the prediction that the more human a mAb, the less immunogenic, an analysis of ADA responses by origin suggests that human and humanized mAbs may not be significantly less immunogenic overall when compared with chimeric mAbs. This is likely due to patient factors as well as the specific amino acid sequences and engineering of mAbs. Some murine mAbs have low rates of ADA, which may be due to the target and dosing regimen.

Based on information provided in product labels, the formation of anti-drug antibodies (ADA) in oncology indications generally appears to be low. This may be due to concomitant immunosuppressive medications and use in late-stage cancer indications. However, higher rates of ADA can be seen for some products.

Using data taken from the labels of approved therapeutic oncology mAbs (see Supplementary Table 1 ), ADA for murine (4), chimeric (9), humanized (24), and human mAbs (18) highlight several points ( Figure 3 ). For murine antibodies, an example of product factors may be seen with two anti-CD20 radioimmunoconjugates, ⁹⁰Y-ibritumomab tiuxetan and ¹²⁵I-tositumomab, which had 3.8% and 70% ADA, respectively. The dosing regimen for both includes pre-dosing with cold antibody to ensure the radiolabeled product targets the tumor rather than normal B cells. Patients treated with ⁹⁰Y-ibritumomab were pretreated with rituximab, the chimeric version of ibritumomab, while patients treated with ¹²⁵I-tositumomab were pretreated with unlabeled tositumomab. The lower levels of ADA seen with ibritumomab may be due to pretreatment with a chimeric mAb rather than a murine mAb.

An example of patient factors is seen with chimeric and humanized antibodies approved for both oncology and autoimmune diseases. Oncology patients treated with rituximab, and rituximab co-formulated with recombinant hyaluronidase have ADA levels ≤2.0%, while in the autoimmune indications, patients treated with rituximab have ADA ranging from 11-56%. Similarly, alemtuzumab, a humanized mAb, has low rates of ADA in the oncology indication (2%), but high rates in patients with multiple sclerosis (83%). On the other hand, the human mAb ofatumumab, which is also approved for oncology and multiple sclerosis has low levels of ADA in both patient populations. It is not clear if the human origin of ofatumumab is the only reason for low levels of ADA in the multiple sclerosis indication, but levels of ADA detected for other human mAbs approved only for autoimmune indications range from <1% up to 61% (data not shown). The variability in rates of ADA in autoimmune indications for human mAbs likely reflects both product and patient factors.

Several approved humanized and human mAbs target immune checkpoints (CKI) such as PD-1 (6 mAbs), PD-L1 (3 mAbs), CTLA-4 (2 mAbs) and LAG-3 (1 mAb). These mAbs are approved for many different cancer indications as monotherapy and some are approved in combination with other CKI mAbs (nivolumab plus ipilimumab, nivolumab plus relatlimab and durvalumab plus tremelimumab). Relatlimab and tremelimumab are only approved for combination use. The rates of ADA for any given CKI mAb vary per indication and treatment regimen.

In Figure 3 , only the highest rate of ADA for a given CKI across indications is shown. For the humanized mAbs the highest incidence of ADA across indications for each mAb was <4.0% for pembrolizumab, dostarlimab, retifanlimab, and toripalimab, and 36% for atezolizumab. For the human mAbs given as a single agent, the highest incidence of ADA across indications was 5.4% for ipilimumab, 11% for nivolumab, 3% for durvalumab, 19.1% for avelumab, 2% for cemiplimab.

However, for CKI mAbs used in combination, the rates are higher compared to their use as a monotherapy, For the combination of nivolumab and ipilimumab, the highest incidence of ADA for nivolumab was 56% in hepatocellular carcinoma and 13.7% for ipilimumab in malignant pleural mesothelioma. For the combination of durvalumab and tremelimumab, the highest incidence of ADA for durvalumab was 10% and 14% for tremelimumab in metastatic non-small cell lung cancer. These data show that in oncology indications where the products activate the immune system to target the cancer, an activated immune system can generate ADA against the therapeutic mAb, even of human origin.

Fc-engineered mAbs and mAb-based products that are not basic Ab structures are not found in nature and may provoke an ADA response against the engineered amino acids or different subunits of the construct, whether it be a BsAb, conjugate or fusion protein. The assays should be designed to detect ADA against each component of the product. There may be pre-existing ADA against elements of some constructs that are based on a platform. These complex molecules should be tested against normal human serum to determine if there are pre-existing ADA against any component of the product. For example, there may be pre-existing anti-hinge region antibodies, especially in rheumatoid arthritis patients, that may bind to mAbs and Fc-fusion constructs (144) or against the CH3 domain of IgG4 (145).

Finally, the impact of pre-existing and emergent ADA targeting different domains of mAbs, complex mAb-based products or engineered Fc region mutations on the safety and efficacy of a product should be considered when planning clinical trials. Depending on the mAb and indication, the ADA may affect PK, may be neutralizing or may have no effect. For most oncology mAbs, there has been no identified clinically significant effect of binding or neutralizing ADA on PK or safety but the effect on efficacy in any given patient is unknown.

Next generation IgG mAbs, fragment-based mAbs, and mAb-based multi-specific conjugate and fusion proteins will build on the lessons learned from both failed and successful products. Some will be more complex constructs such as bispecific ADCs, bispecific fusion proteins, or ADCs with linkers containing many drugs or more than one type of drug. The next generation mAbs will have new challenges for manufacturing, characterization, understanding the CQAs and development of an appropriate control strategy.

In addition to these products, other isotypes, such as IgM, IgE, IgA and IgG3 have been in preclinical and clinical development over the years, including for oncology indications (146–153). The CQAs for these isotypes are not as well understood as those for IgG1, IgG2 and IgG4 mAbs, leading to challenges for both the manufacture and development of a control strategy for these isotypes.

Among the IgG sub-isotypes, IgG3 has the best binding to FcγRs and the highest CDC activity but has 29 allotypes, including different lengths of the hinge region, and most allotypes have a short half-life of ~7 days. It is O-glycosylated and depending on the allotype, can have up to 11 inter-chain disulfide bonds between H chains (154, 155). These features make IgG3 mAbs more challenging to design, characterize and manufacture with consistency. The short-half life can be circumvented by using one of the natural allotypes with a histidine at position 435 in CH3 that has a half-life comparable to IgG1 (156). This substitution also improves binding to Protein A and in combination with two other mutations in CH3 that reduce aggregation, the manufacturability of IgG3 mAbs can be improved (157).

IgM (146), IgA (158) and IgE (159) all have more complex structures than IgG including: 4 constant region domains for IgM and IgE; multiple N-glycosylation sites on all 3 isotypes; O-glycosylation on IgA and IgE; pentameric and hexameric IgM; dimeric IgA; J chain complexed with pentameric IgM and dimeric IgA: and secretory component derived from the polymeric Ig receptor in addition to the J chain complexed with secretory IgM and IgA. All three isotypes have their own sets of Fc receptors (152, 158, 160, 161), which are expressed or may be induced on different effector cells, as well as cell types other than those that express FcγR (See Table 1 in Liu et al. (162)).

As these isotypes progress through clinical development, a thorough characterization of the mAbs, including a determination of potential Fc-mediated effector functions and effector cells engaged by these isotypes will be needed to understand those quality attributes that should be included in the control system and how best to engineer the isotype for a specific purpose in the way that is now possible for IgG mAbs.

The success of mAbs comprises many bench-to-bedside, back to the bench and back to the beside stories. Lessons learned from each generation of product are applied to the next generation, leading to more approved products. But challenges and failures with each generation of product remain. If each new generation of mAbs successfully incorporates designs that overcome previous challenges, more patients will benefit from mAbs.

MS: Writing – review & editing, Writing – original draft.