A landscape of sequence variants in 51 commercial antibody therapeutics

Abstract

Background

Sequence variants (SVs), unintended amino acid substitutions, represent a quality attribute of antibody therapeutics with potential safety and efficacy implications. Although control strategies are evolving, a comprehensive landscape of SVs in commercial antibody products, including biosimilars, innovators, and those from different expression systems or geographical origins, remains largely unexplored.

Methods

We systematically analyzed SVs in 51 commercial antibody therapeutics via optimized high-sensitivity LC–MS/MS peptide mapping. The panel encompassed originators and biosimilars from Chinese and multinational firms, eukaryotic (CHO, NS0) and prokaryotic (Escherichia coli) expression systems, and diverse formats (mAbs, bispecifics, ADCs) and subclasses (IgG1, IgG2, IgG4).

Results

Across all products, 3006 SV occurrences were identified. The vast majority (92.5%) were ≤0.1% per site. Only 7.5% (226 SV occurrences) exceeded 0.1%, 2.4% (72 SV occurrences) exceeded 0.2%, and a minimal 0.4% (12 SV occurrences) exceeded 0.5%. SV levels of Chinese biosimilars and innovators were comparable to international counterparts. While prokaryotic-derived products showed more SVs above 0.2%, SV levels showed no significant correlation with antibody subclasses or formats. Batch-to-batch consistency was excellent across all products.

Conclusions

This study provides the largest empirical SV dataset for marketed antibody therapeutics to date. It demonstrates manufacturing processes maintain SV levels below 0.2% per site and reveals no inherent SV disparity across product origin, subclass, or format. Empirically, this 0.2% threshold serves as a pragmatic benchmark for commercial antibody quality control. These findings provide a science-based reference for establishing cell line screening in-process limits, refining industrial SV control strategies, and guiding regulatory science.

Statement of Significance This large-scale analysis of 51 commercial antibodies defines the real-world profile of sequence variants. The findings guide manufacturing control and support rational regulatory decision-making for antibody therapeutics.

Introduction

Recombinant therapeutic antibodies have revolutionized the treatment of numerous diseases, with their consistent quality being paramount for ensuring safety and efficacy [1–3]. Among the myriad of quality attributes, sequence variants (SVs)—unintended amino acid substitutions arising from errors during DNA replication, transcription, or translation—have garnered significant attention from both biopharmaceutical developers and regulatory agencies [4–6] (Fig. 1). SVs pose a potential risk, as they may alter protein structure, function, stability, or immunogenicity [7, 8]. Consequently, establishing scientifically justified control strategies for SVs throughout development and manufacturing has become an industry imperative.

Figure 1.

Schematic illustration of the origins of SVs in recombinant therapeutic antibodies. SVs arise from errors during DNA replication, transcription, and translation.

Significant progress has been made in understanding the origins of SVs and in developing advanced analytical techniques for their detection, such as next-generation sequencing (NGS) and high-resolution liquid chromatography–tandem mass spectrometry (LC–MS/MS) [4, 5, 8–10]. Furthermore, evidence-based proposals for SV control limits during process development, notably the 0.1% threshold per amino acid site, have been put forward based on studies of approved therapeutics and process capability [6]. However, a critical gap persists in our collective knowledge: a comprehensive, empirical landscape of SVs actually present in a broad array of commercialized antibody therapeutics is lacking. Existing studies have often been limited in scope, focusing on a small number of products, specific subclasses, or a single geographic origin. Systematic comparisons encompassing originators and biosimilars, products from different global regions (e.g. China and multinational companies (MNCs)), diverse expression systems (eukaryotic vs. prokaryotic), and various antibody formats (monoclonal antibodies (mAbs), bispecific antibodies (BsAbs), and antibody-drug conjugates (ADCs)) remain largely unexplored [8, 11]. This knowledge gap hinders the industry's ability to benchmark product quality, refine control strategies based on real-world data, and inform science-based regulatory expectations on what constitutes a typical and well-controlled SV profile for marketed products.

To address this, we performed a large-scale, systematic analysis of SVs in 51 commercially available antibody therapeutics. Regarding marketing authorization, the majority of China-developed biosimilars are approved only by the National Medical Products Administration (NMPA) for sale within China; a few have also received approvals in Europe, the United States, or other regions. In contrast, all innovator antibodies from MNCs are globally launched products with approvals in multiple major regulatory jurisdictions (e.g. Food and Drug Administration (FDA), European Medicines Agency (EMA), Pharmaceuticals and Medical Devices Agency (PMDA)). Most Chinese-developed innovator antibodies are currently marketed exclusively in China, reflecting their recent approvals and the local market focus. This diversity in regulatory origins strengthens the representativeness of our panel. Our panel was meticulously designed to capture real-world diversity, including originator biologics and their biosimilars, products developed by both Chinese and MNCs, and molecules produced in Chinese Hamster Ovary (CHO), murine (NS0), and Escherichia coli expression systems, spanning multiple Immunoglobulin G (IgG) subclasses and modern formats. Utilizing an optimized, high-sensitivity LC–MS/MS peptide mapping method, we generated an extensive dataset of SV identities and relative abundances.

To our knowledge, no previous study has systematically profiled SVs across a broad, representative panel of commercially approved antibody therapeutics. Prior investigations have largely focused on development-stage molecules, single products, or limited numbers of reference materials [8, 11]. Here, we present the first large-scale, multi-dimensional landscape of SVs in 51 marketed antibody drugs, covering biosimilars and innovators, Chinese and multinational manufacturers, eukaryotic and prokaryotic expression systems, and diverse formats and subclasses. By mapping this landscape, we aim to deliver a critical, evidence-based reference for the biopharmaceutical community and regulators, shedding light on the current state of SV control in commercial manufacturing and informing the ongoing evolution of quality assurance paradigms for biologic therapeutics.

Materials and methods

Study design and sample selection

To establish a comprehensive landscape of SVs in commercial antibody therapeutics, a systematic and diverse panel of 51 products was assembled. The selection strategy was designed to capture real-world heterogeneity across multiple critical dimensions:

Product type: Originator biologics and their biosimilars.

Geographical and company origin: Products developed and manufactured by leading Chinese biopharmaceutical companies and MNCs.

Expression system: Eukaryotic systems, including CHO and murine NS0 cells, and prokaryotic systems (E. coli).

Antibody format: mAbs, BsAbs, and ADCs.

IgG subclass: IgG1, IgG2, and IgG4.

The panel included 9 bevacizumab biosimilars and their originator, 2 ranibizumab biosimilars and originator, 2 rituximab biosimilars and originator, and 1 omalizumab biosimilar and originator. Additionally, 33 innovative antibody drugs (15 from Chinese companies and 18 from MNCs) covering the aforementioned formats and subclasses were analyzed. The NISTmAb (RM 8671, Sigma-Aldrich) was included as a well-characterized reference material for method performance assessment. For products with available batches, three independent batch samples were analyzed to assess batch-to-batch consistency. A complete list of products is provided in Tables 1 and 2.

Table 1.

Biosimilar antibody therapeutics selected for SV analysis

INN	Subclass	First approval year	Company	Serial number
Bevacizumab	IgG1	2004	Roche	1
	IgG1	2019	Qilu Pharmaceutical Co., Ltd.	2
	IgG1	2020	Innovent Biologics (Suzhou) Co., Ltd.	3
	IgG1	2021	Bio-Thera Solutions, Ltd.	4
	IgG1	2021	Betta Pharmaceuticals Co., Ltd	5
	IgG1	2021	Boan Biotech	6
	IgG1	2021	BioDlink International Co., Ltd	7
	IgG1	2021	Shanghai Henlius Biotech, Inc.	8
	IgG1	2021	Suzhou Suncadia Biopharmaceuticals Co., Ltd.	9
	IgG1	2023	Chia Tai TianQing Pharmaceutical Co., Ltd.	10
Ranibizumab	IgG1	2006	Novartis AG&Roche	11
	IgG1	2024	Qilu Pharmaceutical Co., Ltd.	12
	IgG1	/	Hangzhou Zhongmei Huadong Pharmaceutical Co., Ltd	13
Rituximab	IgG1	2014	Roche	14
	IgG1	2024	Shanghai Institute of Biological Products Co., Ltd	15
	IgG1	/	Dr. Reddy’s	16
Omalizumab	IgG1	2003	Novartis AG&Roche	17
	IgG1	2024	CSPC Megalith Biopharmaceutical Co., Ltd	18

Table 2.

Innovator antibody therapeutics selected for SV analysis

Innovative antibody drugs from Chinese company
INN	Type	Subclass	First approval year	Company	Serial number
Sevacizumab	mAb	IgG1	2025	Shandong Simcere Biopharmaceutical Co. Ltd	1
Picankibart	mAb	IgG1	2025	Innovent Biologics (Suzhou) Co., Ltd.	2
Teprotumumab	mAb	IgG1	2025	Innovent Biologics (Suzhou) Co., Ltd.	3
Vunakizumab	mAb	IgG1	2024	Suzhou Suncadia Biopharmaceuticals Co., Ltd.	4
Cetuximab β	mAb	IgG1	2024	Taizhou Mabtech Pharmaceutical Co., Ltd	5
Siltartoxatug	mAb	IgG1	2025	Trinomab Biotech Co., Ltd	6
Camrelizumab	mAb	IgG4	2019	Suzhou Suncadia Biopharmaceuticals Co., Ltd	7
Narlumosbart	mAb	IgG4	2023	CSPC Megalith Biopharmaceutical Co., Ltd	8
Stapokibart	mAb	IgG4	2024	Chengdu Kangnuoxing Biopharma, Inc.	9
Zimberelimab	mAb	IgG4	2021	Guangzhou Gloria Biosciences Co., Ltd	10
Firsekibart	mAb	IgG4	2025	Changchun GeneScience Pharmaceuticals Co., Ltd	11
Sacituzumab Tirumotecan	ADC	IgG1	2025	Sichuan Kelun Botai Biomedicine Co., Ltd	12
Becotatugvedotin	ADC	IgG1	2025	BioDlink International Co., Ltd	13
Zanidatamab	BsAb	IgG4	2025	BeOne Medicines Ltd.	14
Iparomlimab and Tuvonralimab	BsAb	IgG1/4	2024	Qilu Pharmaceutical Co., Ltd.	15
Innovative antibody drugs from MNC
INN	Type	Subclass	First approval year	Company	Serial number
Certolizumab pegol	Fab	IgG1	2008	Union Chimique Belge	16
Brolucizumab	scFv	scFv	2019	Novartis AG	17
Sacituzumab	ADC	IgG1	2020	Everest Med	18
Belimumab	mAb	IgG1	2011	GlaxoSmithKline	19
Tislelizumab	mAb	IgG2	2022	AstraZeneca	20
Amivantamab	BsAb	IgG1	2021	Johnson & Johnson	21
Faricimab	BsAb	IgG1	2022	Roche	22
Epcoritamab	BsAb	IgG1	2023	AbbVie	23
Isatuximab	mAb	IgG1	2020	Sanofi	24
Crovalimab	mAb	IgG1	2023	Roche	25
Donanemab	mAb	IgG1	2024	Eli Lilly and Company	26
Zolbetuximab	mAb	IgG1	2024	Astellas	27
Galcanezumab	mAb	IgG4	2018	Eli Lilly and Company	28
Brentuximab	ADC	IgG1	2011	Takeda	29
Enfortumab	ADC	IgG1	2019	Astellas	30
Loncastuximab	ADC	IgG1	2021	Overland	31
Efgartigimod alfa	Fc	IgG1	2021	Argenx	32
Tafasitamab	mAb	IgG1/IgG2	2020	InnoCare Pharma Limited	33

Reagents and materials

Sequencing-grade modified trypsin was purchased from Shanghai Rhinobio Biotechnology Co., Ltd. Iodoacetamide (IAM) was from Tokyo Chemical Industry. Urea, formic acid (FA), and acetonitrile (ACN, HPLC grade) were obtained from Thermo Fisher Scientific. Tris (2-carboxyethyl) phosphine hydrochloride (TCEP-HCl) was from Sigma-Aldrich. All other chemicals were of analytical grade or higher.

Sample preparation for peptide mapping

A standardized peptide mapping sample preparation protocol was employed for consistent and reproducible SV analysis. For each product, ~100 μg of protein (based on mAb concentration) was used. The sample was denatured and reduced in a solution containing 6 M urea and 10 mM TCEP-HCl at 37°C for 20 min. Alkylation was then performed by adding IAM to a final concentration of 20 mM and incubating at 37°C in the dark for 15 min. The urea concentration was diluted to 0.6 M using 50 mM ammonium bicarbonate buffer. A 60 μl aliquot of the diluted sample was digested with trypsin at an enzyme-to-substrate ratio of 1:10 (w/w) at 37°C for 25 min. The digestion was quenched by adding FA to a final concentration of 1%.

Optimized LC–MS/MS analysis for high-sensitivity SV detection

Chromatographic separation was performed on a Thermo Scientific Vanquish Flex UHPLC system. For optimal peptide separation and sensitivity, an ACQUITY UPLC Peptide CSH C18 column (130 Å, 1.7 μm, 2.1 mm × 150 mm; Waters) was used. Mobile phase A was 0.1% FA in water, and mobile phase B was 0.1% FA in 80% ACN/20% water. Peptides were eluted using a linear gradient from 1% to 45% B over 57 min (total run time 73 min including equilibration) at a flow rate of 0.2 ml/min. The column temperature was maintained at 60°C, and the sample compartment at 4°C.

Mass spectrometric analysis was conducted on a Thermo Scientific Q-Exactive Plus hybrid quadrupole-Orbitrap mass spectrometer. Full-scan MS spectra (m/z 300–1800) were acquired in the Orbitrap at a resolution of 70 000 (at m/z 200), with an automatic gain control (AGC) target of 1e6 and a maximum injection time of 50 ms. Data-dependent acquisition (DDA) was performed using a top-5 method. Selected precursor ions were fragmented by higher-energy collisional dissociation (HCD) with a normalized collision energy of 25%. MS/MS spectra were acquired at a resolution of 17 500 with an AGC target of 1e5 and a maximum injection time of 120 ms. Dynamic exclusion was set to 5 s.

Method optimization and sensitivity assessment

Critical parameters affecting SV detection sensitivity, including column chemistry and sample load, were systematically optimized. Performance was benchmarked against the NISTmAb. Compared to a standard BEH C18 column, the CSH C18 column provided superior peak shape and reproducibility, increasing the number of confidently identified SVs in the NISTmAb under identical conditions [12]. Increasing the injection amount from 4 μg to 10 μg on the CSH column significantly enhanced the signal for low-abundance peptides, allowing the identification of 43 SV sites in the NISTmAb, surpassing the number reported in prior literature. This optimized method was further validated using an internal control sample (mab3C) with a known high-level single-site mutation (K148N) and multiple low-level SVs (~0.01%), confirming its robust sensitivity and quantitative capability.

Data processing and SV identification/quantification

Raw LC–MS/MS data were processed using Byos software (version 4.3, Protein Metrics). Database searches were performed against the expected amino acid sequence of each therapeutic antibody. The search parameters included trypsin digestion with up to two missed cleavages, fixed modification of Carbamidomethyl on cysteine, and variable modifications including deamidation, protein N-terminal acetylation, and methionine oxidation. Crucially, the search encompassed all possible single amino acid substitutions resulting from a single nucleotide change. Identified SV candidates were filtered and manually verified using Byos software (Protein Metrics) by examining MS1 mass accuracy (within 10 ppm), MS/MS spectral quality (minimum score of 15), fragmentation patterns (the putative amino acid substitution site should be covered by MS/MS fragment ions), isotopic distributions (the monoisotopic peak of the isotope cluster was inspected to ensure correct m/z assignment), and retention time consistency relative to the native peptide (modified peptides and their native counterparts should display different retention times to rule out interference from in-source fragmentation).

For relative quantification, the abundance of each verified SV-containing peptide was compared to its corresponding native peptide using Byos software (Protein Metrics). The relative percentage of the SV was calculated as: [Area (SV peptide) / (Area (Native peptide) + Area (SV peptide))] × 100%.

Data analysis and visualization

Summary statistics were calculated for the entire dataset. SVs were categorized and analyzed based on product type (biosimilar/innovator), company origin (Chinese/MNC), expression system, IgG subclass, and antibody format. Batch-to-batch consistency for products with multiple batches was assessed by comparing the presence/absence and relative abundance of SVs. Data visualization and statistical analyses were performed using GraphPad Prism (version 8.0) and Hiplot (https://hiplot.com.cn/home/index.html).

Results

Establishment of a high-sensitivity LC–MS/MS method for comprehensive SV detection

Given the typically low abundance of SVs in commercial antibody products, achieving high analytical sensitivity is paramount. We systematically optimized a peptide mapping-based LC–MS/MS workflow, focusing on two critical parameters: column chemistry and sample load. Using the well-characterized NIST mAb as a benchmark, the performance of BEH (Ethylene Bridged Hybrid) and CSH (Charged Surface Hybrid) C18 columns was evaluated. The CSH column provided superior peak shape and reproducibility (Data not shown). Under identical conditions (4 μg load), the number of confidently identified unique SV sites in the NIST mAb increased from 13 using a BEH column to 18 using a CSH column. Further increasing the injection amount to 10 μg on the optimized CSH column significantly enhanced signals for low-abundance species, enabling the identification of 43 unique SV sites in the NIST mAb (Fig. S1a). This performance exceeds the 21–23 sites reported in prior inter-laboratory studies, underscoring the enhanced sensitivity of our method [6]. Our method covers 50% of the SV sites reported in the previous study (Fig. S1c). The majority of uniquely identified SV sites were present at levels ≤0.05% (Fig. S1d), which further confirms the sensitivity of our method. This optimized method was subsequently validated using an internal control sample (mab3C) containing a known high-level single-site mutation (K148N) and multiple low-level SVs (~0.01%), confirming robust sensitivity and reliable quantification (Fig. S1b). The finalized platform was applied to profile SVs across the entire panel of 51 commercial therapeutics.

An integrated landscape of SVs across 51 commercial antibody therapeutics

Integrating data from all 51 products (biosimilars and innovators), a total of 3006 SV occurrences were identified and quantified. The comprehensive distribution revealed that the vast majority, 92.5% (2780/3006 SV occurrences), of all SVs were present at a level of ≤0.1% per individual site. A small fraction, 7.5% (226 SV occurrences), exceeded 0.1%. Within this group, only 2.4% (72 SV occurrences) of the total exceeded 0.2%, and a minimal 0.4% (12 SV occurrences) exceeded 0.5% (Fig. 2). This extensive dataset provides a robust, empirical benchmark for the typical SV burden in contemporary commercial antibody manufacturing, demonstrating that current processes across the industry generally maintain excellent control, with SV levels predominantly below 0.2% per site. Notably, the observed proportion of SVs exceeding 0.1% in our study is moderately higher than that reported in some prior publications [6]. This discrepancy may be mainly attributed to the broader diversity of products included in our panel, which better represents the heterogeneity of the current commercial landscape.

Figure 2.

Global landscape of SV occurrences across 51 commercial antibody therapeutics.

SV profiles in eukaryotic-expressed biosimilars demonstrate high process control

To evaluate SV control in a major therapeutic class, we analyzed the originator and 9 Chinese biosimilars of bevacizumab (IgG1, CHO-expressed), with three batches per product. The number of unique SV sites identified per product varied (Fig. 3a). The originator (Avastin) and one biosimilar (SBP6) exhibited the fewest sites. For most products, the vast majority of individual SV levels were below 0.2% (Fig. 3b). Excellent batch-to-batch consistency was observed, evidenced by high overlap in SV identities across three batches for representative products (Fig. 3c) and minimal inter-batch variability (≤0.05% for the majority of sites across 9 biosimilars) (Fig. 3d).

Figure 3.

SV profiles of eukaryotic-expressed bevacizumab biosimilars demonstrate robust process control. (a) Number of unique SV sites identified per product (each dot represents one batch). (b) Distribution of SV types and their relative levels across the bevacizumab originator and nine biosimilars. (c) Venn diagrams showing SV site overlap across three batches for the originator (Avastin) and representative biosimilars (SBP5, SBP6, SBP8). (d) Inter-batch variability (range) of SV levels for the 362 unique SV sites identified across nine biosimilars (excluding SBP8). (e, f) Distribution of SV types and levels for omalizumab and rituximab originators and their biosimilars. Each dot represents for an individual SV occurrence.

A total of 42 distinct amino acid substitution types were identified across the 10 bevacizumab products (Fig. 3b). All SVs in the originator were present at levels <0.1%. Among the 9 biosimilars, 8 products maintained all individual SVs at ≤0.2%. Of the 992 SV occurrences identified across these 8 biosimilars excluding SBP8, only 11 (1.11%) exceeded the 0.1% threshold (Table S1). These data indicate that SV control and process consistency in these Chinese-developed biosimilars are comparable to the originator standard.

One biosimilar (SBP8) displayed a distinct profile, with 41 SV occurrences exhibiting Tyr → Phe substitutions above 0.1%. This pattern of multi-site, co-localized substitutions is characteristic of tyrosine depletion in cell culture, suggesting a specific, remediable process condition [13, 14]. Analysis of originators and Chinese biosimilars of omalizumab and rituximab further supported these findings. SV levels in the biosimilars were all ≤0.1% and were comparable to or lower than their respective originators (Fig. 3e, f).

Prokaryotically-expressed products exhibit a higher propensity for elevated SVs

To evaluate systems with inherently different translational fidelity, we analyzed the originator and two Chinese biosimilars of ranibizumab (E. coli-expressed), with three batches each. The biosimilars showed slightly fewer unique SV sites than the originator (Fig. 4a), with excellent batch-to-batch consistency (Fig. 4b). Analysis of 237 SV occurrences across the three products revealed 23 substitution types (Fig. 4c). Notably, 3 SV occurrences (1.3%) exceeded 0.2% in relative abundance. In contrast, among 1197 SV occurrences in the 10 CHO-produced bevacizumab products (excluding SBP8), only one (0.08%) exceeded 0.2%. This aligns with the expected higher baseline misincorporation rate in prokaryotic systems [15, 16]. The number of SV occurrences >0.1% in the biosimilars (1 and 6 sites) was less than the originator (12 sites) (Fig. 4d), demonstrating that the SV control capability of the Chinese manufacturers in this prokaryotic system is on par with the originator.

Figure 4.

Prokaryotically-expressed ranibizumab products show a higher propensity for elevated SVs. (a) Number of unique SV sites identified per product (each dot represents one batch). (b) Venn diagrams showing SV site overlap across three batches. (c) Distribution of SV types and their relative levels across the ranibizumab originator and two biosimilars. Each dot represents for an individual SV occurrence. (d) Comparison of the number of SV occurrence >0.1% in the originator versus biosimilars.

Innovator antibodies from Chinese companies show well-controlled SV profiles

We next analyzed 15 innovative antibody drugs from Chinese companies, encompassing mAbs, bispecifics (BsAbs), ADCs, and IgG1/IgG4 subclasses (three batches each). All products demonstrated excellent batch-to-batch consistency in their SV profiles (Fig. 5a). The number of unique SV sites per product was generally low (mostly <20 sites), with no apparent correlation to antibody subclass or format (Fig. 5). Individual SV levels were well-controlled, with all but one product maintaining levels below 0.2% (Fig. 5b). Compared to the earlier-generation biosimilars, these recently developed innovator products exhibited fewer SV sites overall, reflecting the industry's advancing understanding and proactive control of SVs during cell line and process development [3, 5, 17].

Figure 5.

SV profiles of innovator antibodies from Chinese companies are well-controlled. (a) Number of unique SV sites identified per product (each dot represents one batch). (b) Distribution of SV types and their relative levels across Chinese-developed innovator antibodies (mAbs, bispecifics, ADCs; IgG1/IgG4). Each dot represents for an individual SV occurrence.

Comparative analysis reveals no inherent SV disparity based on product origin or attributes

To gain a holistic view and test for inherent biases, we analyzed an additional 18 innovative antibodies from MNCs and compared them with the 15 Chinese innovators. This combined dataset of 33 products covered CHO, NS0, and E. coli expression systems; IgG1, IgG2, and IgG4 subclasses; and mAb, BsAb, and ADC formats.

The number of SV sites showed no apparent differences based on expression system (CHO vs. NS0), antibody subclass (IgG1, IgG2, IgG4), or format (mAb, BsAb, ADC) (Fig. 6a-c). The lower site count for the two E. coli-expressed products is likely attributable to their shorter protein sequences (Fragment of antigen binding (Fab) and single-chain variable fragments (scFv)). Critically, the SV profiles of Chinese-developed innovators were comparable to those from MNCs, with no systematic difference in SV levels or distribution (Fig. 6d).

Figure 6.

Analysis of unique SV sites across different categories of innovator antibodies reveals no inherent bias. (a-d) Number of unique SV sites per product grouped by (a) expression system, (b) IgG subclass, (c) antibody format and (d) geographical origins (each dot represents one batch).

Analysis of SV levels revealed that 27 of the 33 products (82%) maintained all individual SVs below 0.2% (Fig. 7). Six products contained SVs exceeding 0.2%. Four of these exhibited multi-site substitution patterns (e.g. Val → Xle, Thr → Ser, Pro→Ala) suggestive of process-related amino acid misincorporation. The other two contained single-site SVs at ~1% (Arg → Lys, Lys → Asn). The most frequently observed substitution types (Asn → Ser, Pro→Ala, Thr → Ser, Val → Xle) predominantly result from single-nucleotide transitions and often involve amino acids with similar physicochemical properties.

Figure 7.

Distribution of SV types and their relative levels across the 33 innovative antibody therapeutics from Chinese and MNCs. The dashed line indicates the 0.2% level. Each dot represents for an individual SV occurrence.

Discussion

Unlike earlier reports that were restricted in scope, our dataset presents the most extensive empirical survey to date of SVs in marketed antibody therapeutics. By systematically analyzing 51 commercial products—spanning originators and biosimilars, molecules from Chinese and MNCs, and diverse expression systems, formats, and subclasses—we have mapped a comprehensive landscape of SV profiles in real-world, approved drugs. Our data reveal several key findings with significant implications for biopharmaceutical development, quality control, and regulatory science.

First, our results provide a robust, quantitative benchmark for the typical SV burden in contemporary commercial manufacturing. The fact that nearly all products maintained SV levels below 0.2% suggests that this higher threshold may represent a practical and conservative "risk alert" level for commercial products, beyond which more focused investigation and process scrutiny may be warranted. The moderately higher proportion of SVs >0.1% observed here, compared to some prior studies [6], likely reflects the enhanced sensitivity of our optimized method and the inclusion of a broader, more representative product panel, capturing the true heterogeneity of the market [11].

Second, the study demonstrates that SV control in products from Chinese biopharmaceutical companies is on par with international standards. For both biosimilars and innovators, Chinese-developed products showed SV levels, substitution patterns, and batch-to-batch consistency comparable to—and in many cases, indistinguishable from—their originator or MNC-developed counterparts. This finding is particularly significant given the rapid growth and global impact of China's biopharmaceutical sector [18–21]. It indicates that advanced process understanding and control strategies for critical quality attributes like SVs have been successfully implemented by leading Chinese manufacturers [4, 5, 22].

Third, we observed clear system-specific differences consistent with known biology. Prokaryotically expressed products exhibited a higher incidence of SVs exceeding 0.2%, aligning with the inherently lower translational fidelity of E. coli systems [23–25]. This reinforces the importance of tailoring control strategies to the expression host. Conversely, no significant differences in SV profiles were attributable to antibody subclass (IgG1/2/4) or format (mAb/BsAb/ADC), suggesting that these molecular attributes do not inherently predispose a molecule to higher SV when modern platform processes are employed.

Fourth, the data highlight specific, identifiable signatures of suboptimal process conditions. The presence of multi-site, co-localized substitutions (e.g. Tyr → Phe, Val → Xle) in a small subset of products—observed in both biosimilars and innovators from various origins—is a classic indicator of nutrient depletion (e.g. tyrosine, valine) in cell culture [13, 26]. These patterns are not random noise but direct readouts of cellular stress. Their occurrence in marketed products underscores that while overall control is excellent, continuous vigilance and optimization of feeding strategies remain important, especially for high-titer processes.

Finally, this landscape analysis fills a critical evidence gap for regulators and developers. Prior discussions on SV control have largely been based on theoretical error rates, limited datasets, or development-stage molecules [6, 8, 27]. Our study provides a substantial, real-world dataset answering a fundamental question: What do SV profiles actually look like in a wide array of successfully approved and commercialized antibody drugs? The answer is that they are consistently well-controlled, predominantly below 0.1%–0.2% per site. This empirical evidence can help inform more nuanced, risk-based regulatory expectations and guide companies in setting scientifically justified in-process limits and product specifications. It supports the view that SVs below 0.2% are commonly observed in safe and efficacious medicines and may represent a normal, low-risk background level for optimized commercial processes.

This study is cross-sectional and observational; it establishes the “what” but not the “why” for every variant. The clinical impact of specific SVs, particularly those above 0.2%, was not assessed and remains a complex product-specific question. Future work combining such landscape data with in silico immunogenicity prediction, structural modeling, and forced degradation studies could further refine risk assessment frameworks [28, 29]. Additionally, expanding such surveys to include more products over time will allow tracking of industry trends as processes continue to intensify.

In summary, this comprehensive analysis reveals that the current state of commercial antibody manufacturing achieves a high degree of control over SVs. The proposed 0.1% or more per-site target is a relevant and achievable process development goal, while the 0.2% level emerges as a pragmatic upper benchmark for marketed products. The absence of meaningful disparities based on product origin or type is an encouraging sign of globalized manufacturing quality. This dataset serves as a valuable reference point, empowering the industry to benchmark its processes, refine control strategies with greater confidence, and engage in evidence-based dialogues with regulators regarding this important quality attribute.

Author contributions

Gangling Xu (Conceptualization [lead], Data curation [lead], Formal analysis [lead], Funding acquisition [equal], Investigation [lead], Methodology [lead], Resources [lead], Software [lead], Writing - original draft [lead]), Gang Wu (Conceptualization [supporting], Formal analysis [supporting], Funding acquisition [equal], Investigation [supporting], Resources [supporting], Software [supporting]), Chunbo Cui (Data curation [supporting], Formal analysis [supporting], Investigation [supporting]), Sihan Wei (Data curation [supporting], Formal analysis [supporting]), Maoqin Duan (Formal analysis [supporting], Methodology [supporting], Resources [supporting]), Yongfei Cui (Data curation [supporting], Resources [supporting]), Yuting Mei (Resources [supporting], Software [supporting]), Chuanfei Yu (Conceptualization [equal], Resources [equal], Supervision [equal], Writing—review & editing [equal]), and Lan Wang (Conceptualization [equal], Resources [equal], Supervision [equal], Writing—review & editing [equal])

Conflicts of interest

None declared.

Funding

This study was supported by the National Institutes for Drug Control Research Fund for Young and Middle-aged Development (Grant No. 2024B2) and Beijing Municipal Science and Technology Commission (Grant No. Z251100004625003).

Data availability

All data in this study are available on reasonable request.

Declaration of competing interest

The authors declare that there are no conflicts of interest.

Ethics and consent statement

No consent was required.