Glycan-intrinsic monoclonal antibody biomanufacturing

Introduction

Monoclonal antibodies (mAbs) are among the most commercially and clinically consequential therapeutic modalities in modern medicine. Antibodies are “Y”- shaped molecules whose function is divided into two parts: the antigen-recognition Fab domain, which binds antigens, and the Fc region, which enables effector functions by binding to immune cells through IgG’s receptors, Fc gamma receptors (FcγRs), as seen in Figure 1A (Kang and Jung, 2019). A classic example is Trastuzumab, which binds to HER2-overexpressing epithelial cells such as breast cancer cells, blocking growth signals. The Fc of the antibody then binds with macrophages, allowing for antibody-dependent cellular phagocytosis (ADCP) and cytotoxicity (ADCC) to eliminate the cancerous cells (Arnould et al., 2006; Shi et al., 2015).

The popularity of mAb-based therapeutics is due in part to technological advances that enable their production via recombinant expression and to engineered cell lines that achieve high titers. Classical methods of selection, such as dihydrofolate reductase (DHFR) and glutamine synthetase (GS), are used for high transgene selection. Moreover, cell-cycle engineering, where induced G1 cell-cycle arrest decouples growth from productivity, boosts cell productivity. These strides in metabolic engineering enable higher yields. However, we know that antibody yield does not correlate with antibody quality (Kaur et al., 2023).

Narrowing into the glycobiology, these antibodies also undergo a post-translational modification in which oligosaccharides bind to the peptide backbone, a process known as N-glycosylation. Importantly, glycosylation is non-template-driven, meaning it is not encoded in the same way as amino acid sequences are encoded by the central dogma of biology (DNA to RNA to protein). Instead, glycosylation relies on a network of enzymes bound by their activity and substrates, including nucleotide sugars and cofactors like manganese within a cell’s endoplasmic reticulum and Golgi apparatus. Depending on the level of cell stress and nutrient availability, antibody glycosylation will vary, making it inherently difficult to control in a bioprocess.

Monoclonals are modified with a single heavy chain glycan on their Fc at the asparagine 297 site (N297). The typical sugar on this site is a complex-type structure with a small amount instead containing a high-mannose glycan structure, as shown in Figure 1B. The terminal glycan sugars augment function. A typical monoclonal produced in Chinese hamster ovary (CHO) cells will contain mainly N-acetylglucosamine (GlcNAc) as the terminal sugar, with a smaller fraction carrying terminal galactose and a minute amount containing sialic acid. This is further augmented with a fucose residue on the innermost GlcNAc residue. Each sugar plays a different role in immune engagement. Loss of core fucose can increase activating FcγR engagement, a key glycoengineered strategy (Yamane-Ohnuki et al., 2004). Terminal galactose increases C1q engagement and complement-dependent cytotoxicity (Peschke et al., 2017). The addition of sialic acids boosts anti-inflammatory activity through lectin engagement, although the effect on antibody pharmacokinetics is limited (Jones et al., 2025). However, the anti-inflammatory mechanism of action remains contentious. Lastly, high-mannose glycoforms are known to be cleared more rapidly from circulation via mannose receptor–mediated uptake and are therefore controlled as a liability attribute, typically increasing under cell stress during production (Goetze et al., 2011; Slade et al., 2016).

As stated earlier, glycosylation is heterogeneous and driven by metabolism and nutrient availability in the culture. Hence, during bioprocessing, small changes in process conditions can alter the glycan pattern (“glycoforms”) of the mAb being produced. Conditions such as medium composition, temperature, dissolved oxygen, pH, feeding strategy, and accumulated waste metabolites are known to alter the final glycan pattern (Kildegaard et al., 2016; Blondeel and Aucoin, 2018). This sensitivity drives variations during bioprocessing, leading to batches with variable glycans that fall outside the release target specification.

Paradoxically, these process sensitivities influence glycosylation and are among the largest manufacturing challenges today, but also the most underexploited therapeutic opportunity. Cell line glycoengineering is part of the answer to this. Knocking out specific enzymes can enhance therapeutic efficacy through receptor engagement. Knocking out fucosyltransferase produced afucosylated mAbs with increased FcγR engagement, allowing for lower drug dosing (Golay et al., 2022). Alternatively, knocking in a sialyltransferase such as ST6GAL1 produces human α2,6-linked sialic acids (Raymond et al., 2015), thereby increasing anti-inflammatory activity (Wang and Ravetch, 2019).

These observations motivate what I term as mAb glyco-betters: antibodies in which a defined glycoform constitutes a deliberate design variable. The established terms “glycoengineered” and “glyco-optimized antibody” describe cases in which a glycan structure was deliberately altered. Glyco-better is intended to denote the end. In other words, a molecule whose defined glycoform is the demonstrated basis of a pharmacological advantage over a comparator, and which is therefore carried forward as a design specification and a release criterion rather than reported as a characterization result. Several such molecules are already approved and in clinical use, and a broader pipeline is advancing through early-stage development (Pereira et al., 2018; Sneed et al., 2024). However, the biomanufacturing infrastructure required to produce these mAbs reproducibly and at scale remains immature relative to the biological knowledge that motivates their development.

Two complementary disciplines are required to close this gap. First, the cell line engineering and feeding strategies that together direct glycosylation towards a desired glycoform, and second, the process knowledge required to control said glycan pattern. Process knowledge is derived from methods for measuring attributes and technology that enables monitoring, collectively known as process analytical technology (PAT). The deployment of PAT for glycosylation monitoring enables near-real-time measurements of glycan profiles that can inform control strategies. The crux of the hypothesis, I argue, is that glycosylation should be treated as a design feature of the mAb rather than a late-stage risk, and thereby integrated into the quality-by-design (QbD) paradigm, in which development proceeds from predefined quality objectives through systematic product and process understanding.

IgG glycobiology and effector function

Afucosylation and enhanced ADCC

Fucose is α1,6-linked to the innermost GlcNAc residue of the glycan by the fucosyltransferase 8 (FUT8). The removal of core fucose from the N-glycan is, to date, the most validated method for increasing antibody effector function through increased FcγR affinity. Falconer and colleagues elucidated the mechanism by which fucosylation inhibits binding to the activating receptor FcγRIIIa (Falconer et al., 2018). Indeed, afucosylation of the antibody increased affinity upwards of 50 fold (Shields et al., 2002) with subsequent studies showing increased efficacy of tumor cell killing in vitro by natural killer cells (Liu et al., 2015). The effects of afucosylation on effector functions beyond NK-mediated ADCC, such as macrophage-mediated phagocytosis, are more variable and remain incompletely resolved (Cvijić et al., 2024).

This mechanism has already been translated into clinical benefit. Mogamulizumab was the first FDA-approved biologic to use a FUT8 knockout cell line in 2018 (Kasamon et al., 2019). Other methods to control for fucosylation levels include overexpression of β-1,4-N-acetylglucosaminyltransferase III (MGAT3), which adds a bisecting GlcNAc to the glycan structure (Umaña et al., 1999), thereby causing steric hindrance during glycan processing (Ferrara et al., 2006). As a result, this decreases total fucosylation, as seen with type II anti-CD20 obinutuzumab, a glyco-engineered mAb approved in 2013 (Cameron and McCormack, 2014). The data show that obinutuzumab plus chlorambucil outperformed rituximab plus chlorambucil, increasing the median progression-free survival from 15.2 to 26.7 months in previously untreated patients with chronic lymphocytic leukemia (Goede et al., 2014). Obinutuzumab and rituximab target the same antigen but differ in both binding mode and glycosylation. These differences act through distinct mechanisms, the type II epitope conferring enhanced direct cell death and reduced complement activation, whereas afucosylation confers increased FcγRIIIa engagement (Mössner et al., 2010). The clinical outcome, therefore, reflects both features, although the effector-mediated component is attributable to afucosylation.

It is, however, important to note that the same FcγRIIIa engagement that drives efficacy also drives cytokine release. For example, obinutuzumab elicits marked elevations in the proinflammatory cytokines IL-6 and IL-8, with infusion-related reactions during the first infusion (Freeman et al., 2015). Effector function thus carries an upper as well as a lower bound, indicating that glycan control is often defined by proximity to a specified target rather than by maximization of a single glycoform.

Sialylation and anti-inflammatory modulation

Terminal sialylation also presents a mechanistic distinction and a clinical opportunity. A minor fraction of endogenous human antibodies carries terminal α2,6-linked sialic acid. This modification is absent in CHO-derived monoclonals, which express α2,3-linked sialic acids while lacking the enzyme sialyltransferase ST6GAL1 required for α2,6-linked human-type sialylation (Lin et al., 2015; Zhang et al., 2019). α2,6-linked sialic acid-containing antibodies have been proposed as a key contributor to the anti-inflammatory activity of intravenous immunoglobulin (IVIG), a second-line treatment for severe autoimmune and inflammatory conditions. Hypersialylation of IVIG has been shown to produce comparable efficacy at a fraction of the typical dose (Anthony et al., 2008). Therefore, a recombinantly sialylated mAb could, in principle, recapitulate that benefit at substantially lower doses as well (Sneed et al., 2024).

However, the mechanistic basis of this activity remains contested. The originally proposed pathway, in which α2,6-linked sialylated Fc engages the lectin receptor dendritic cell-specific intercellular adhesion molecule-3-grabbing non-integrin (DC-SIGN) to upregulate the inhibitory receptor FcγRIIB, was derived largely from murine models. Subsequent work in human systems reported no detectable binding of sialylated human IgG to DC-SIGN, regardless of Fc-glycan composition (Temming et al., 2019). The therapeutic hypothesis that increased sialylation confers anti-inflammatory activity is thus supported by functional data but not yet by a settled receptor-level mechanism, a distinction that matters for any developer pursuing a sialylated glyco-better.

The regulatory logic of glyco-betters

From a regulatory and commercial standpoint, a precisely defined glycoform may offer a differentiation pathway that is both scientifically grounded and strategically durable in an era of intensifying biosimilar competition. Where a glycoform difference is demonstrably and quantitatively linked to a clinical outcome, as in the obinutuzumab precedent, a glyco-better can be positioned as a pharmacologically distinct new entity rather than a structural copy. Indeed, even slight shifts in the glycosylation of FDA-approved biologics can alter receptor binding affinity and jeopardize batch-to-batch compliance (Zhu et al., 2026). The evidentiary bar for glycoform-based differentiation is becoming increasingly tractable as glycan characterization tools and mechanistic understanding mature.

Debottlenecking glycosylation

The cell-line engineering toolbox

Earlier, we discussed loss-of-function through knockout cell lines, as well as single-gene knock-in strategies. In recent years, CHO glycoengineering has moved toward targeted, multiplexed genome editing. Clustered regularly interspaced short palindromic repeats (CRISPR)-Cas9 now enables both multiplexed disruption of genes in the glycosylation pathways and site-specific knock-in of heterologous glycosyltransferases at defined genomic loci, rather than relying on random integration. Nguyen and colleagues performed a systematic screen of 42 glycosyltransferase genes in CHO cells, identified bottlenecks in the N-glycosylation pathway, and alleviated them by overexpressing individual or multiple critical genes (Nguyen et al., 2021). This method led to mAb glycan variants with up to 80% galactosylated or 70% sialylated species, with little change in cell growth, integrated viable cell density, although specific productivity decreased markedly in many of the overexpressing pools. In another method by Karottki et al. (2020), the authors increased transcription of endogenous genes using CRISPR activation. They successfully increased transcription of MGAT3 and ST6GAL1, resulting in increased total sialylation of the mAb (Karottki et al., 2020). Traditional single knock-in and knockout methods carry a risk of metabolic burden decreasing cell productivity (Nguyen et al., 2023). Newer CRISPR methods may help mitigate these concerns. However, this mitigation has not yet been quantified directly against knock-in controls.

Feeding strategies

A complementary approach to increasing the production of certain glycoforms is to feed glycan precursors. The Golgi-luminal availability of guanosine diphosphate-fucose (GDP-fucose), uridine diphosphate-galactose (UDP-galactose), and cytidine monophosphate-sialic acid (CMP-sialic acid) limits glycosyltransferase reaction rates and the relative abundance of glycoforms produced (Hadley et al., 2014). The obvious solution is to remove this bottleneck by directly feeding the required substrate. It has been shown that targeted feeding can produce more homogeneous glycan profiles (Schäpertöns et al., 2026). For example, alternative-day targeted feeding of galactose with manganese, the cofactor for galactosyltransferase, increased mAb galactosylation from roughly 34%–69%, attributable to increased intracellular UDP-galactose pools (Gyorgypal et al., 2025). Indeed, engineering the supply of these activated donors offers control orthogonal to glycosyltransferase engineering; it also connects directly to the bioprocess, since intracellular donor pool sizes respond dynamically to media composition and feeding.

Converging on the ceiling of glycosylation

Cell line-engineered overexpression of the glycosyltransferase and nutrient feeding address opposite sides of the same constraint. Overexpressing a glycosyltransferase relieves the enzyme-side bottleneck, while supplementing the corresponding nucleotide-sugar precursor relieves the substrate-side bottleneck. Combining the two, rather than choosing between them, allows a glycoform to approach its true ceiling. Indeed, this is true for galactosylation. Dekkers and colleagues show that B4GALT1 overexpression combined with galactose supplementation achieved 82% galactosylation without loss of mAb yield (Dekkers et al., 2016). This figure goes beyond feeding alone by raising the ceiling of galactosylation, rather than merely assuming the substrate alone is enough. The same logic could be applied to sialylation and fucosylation as well. These methods cumulatively enable an enriched glycan profile. However, the requirement for control differs by attribute. Fucosylation is close to binary, since FUT8 knockout drives afucosylation to near completion, and that attribute largely ceases to vary. The product does not thereby become homogeneous, since galactosylation, sialylation, and high-mannose content remain responsive to process conditions. The attributes that remain variable after glycoengineering are therefore the ones that require process control.

Process analytical technology for continuous glycan monitoring

The analytical gap

The engineered cell line debottlenecks the cell line’s ability to produce the desired glycosylation, but to produce the product at scale reliably requires monitoring and control strategies. Recent years have seen advancements in PAT that enable the monitoring of glycosylation. Methods rely on high-performance liquid chromatography (HPLC), mass spectrometry (MS), or surface plasmon resonance. The typical gold-standard approach for mAbs is to use HPLC, as this allows rapid batch analysis and enables the development of a method to measure during the bioprocess; automation is required (Tiwold et al., 2023). Others and I have developed near-real-time measurement platforms using LC-MS-based approaches by integrating the analytical methodology with a fluidics system that automates sample preparation, including protein A purification, deglycosylation for free-glycan analysis, and trypsin protease digestion for MS-based analysis (Hajduk et al., 2019; Gyorgypal and Chundawat, 2022; Liu et al., 2022; Gyorgypal et al., 2023). This enables autonomous monitoring of mAb glycans, typically within a few hours of sample collection. However, these types or methods produce complex data that require post-processing and typically incur high maintenance or enzyme costs. This is also bound by the need to integrate with a process control system for feedback, as the current PAT only allows monitoring. Latency of a few hours between sampling and readout is sufficient for a semi-batch process, where glycosylation may not change significantly within that time; however, it would be difficult for a continuous process running for multiple weeks. This makes process control strategies harder to implement due to the lag time.

Other strategies to measure glycosylation include soft sensors, which do not require reactor sampling, such as in-line sampling discussed earlier. Raman spectroscopy has become a popular approach as it allows for continuous monitoring without risk of contamination. Raman has been used for metabolite and titer analysis and, in some instances, for glycosylation prediction (Li et al., 2018; Hamla et al., 2022; Schwarz et al., 2022). These methods typically use partial least squares (PLS) models to generate a standard curve for measuring metabolites, titers, or glycans. These models carry risks, such as poor extrapolation when concentrations are determined below the limit of detection, as well as bias when scaling up if the model is based on bench-scale reactor data (Gibbons et al., 2022). Moreover, in situ methods such as Raman detect vibrational signatures from all molecules in the sample simultaneously. Since the reactor medium is quite complex, this can produce overlapping or co-varying spectral bands that confound the specific signal (Esmonde-White et al., 2022).

The current limitation in analytical characterization is a trade-off between slower, more precise methods and rapid approximation methods.

Continuous bioprocessing and the regulatory challenge

The prevailing direction in next-generation mAb biomanufacturing is the use of continuous perfusion bioreactors, which add fresh medium while bleeding off cells to reach a target cell density for optimal mAb production. A regulatory pathway has developed in parallel: ICH Q13, the harmonised guideline on continuous manufacturing of drug substances and drug products, was adopted by the International Council for Harmonisation in November 2022 and implemented by the United States Food and Drug Administration (FDA) and the European Medicines Agency (EMA) in 2023 (Bhavna et al., 2022). ICH Q13 applies explicitly to therapeutic proteins and recognizes PAT-enabled monitoring and control, alongside real-time release testing, as a core element of the control strategy. ICH Q13 simultaneously raises the evidentiary burden, since a continuous process must demonstrate glycosylation consistency across run durations of weeks to months under conditions that batch development was never required to characterize systematically.

Perfusion bioreactors offer a more favorable environment than fed-batch processes because they avoid the progressive accumulation of metabolic byproducts and depletion of nutrients that otherwise continuously perturb cellular physiology over the course of a fed-batch run. In practice, the quantitative relationships between cell-specific perfusion rate, viable cell density, metabolic state, and steady-state glycoform distribution remain incompletely characterized even for well-established cell lines. Early mechanistic modeling of glycosylation in perfusion reactors, while promising, underscores how strongly glycoform output depends on the dynamic nutrient environment (Karst et al., 2017).

The economics of continuous operation are scale-dependent. Single-use continuous facilities achieve approximately 35% cost savings relative to stainless-steel batch facilities for annual demand in the hundreds of kilograms. This advantage narrows at the multi-ton scale (Ranbhor, 2025). A mAb glyco-better derives its advantage from enhanced potency, which may permit lower dosing. Reduced dosing lowers annual mass demand, which may place such products within the volume range where continuous manufacturing is most economically favorable.

However, the slower-than-anticipated industrial adoption of continuous bioprocessing reflects a multi-factor problem that is largely capital and organizational: substantial sunk investment in validated batch infrastructure and workforce expertise gaps in continuous operation (Kumar et al., 2020). This is further compounded by the need for real-time data management, the greater analytical burden imposed by PAT requirements, and uncertainty about regulatory expectations during a transitional period (Drobnjakovic et al., 2023).

Toward closed-loop glycosylation control

The control strategy of a bioreaction is the planned set of measurements and actions, grounded in process understanding, to keep mAb critical quality attributes (CQAs) within target ranges. PAT supplies real-time measurements that make the control strategy actionable. This idea is further extended with digital twins, in which data-driven and hybrid models have been trained to forecast general cell trajectories (Park et al., 2023). Applied to glycosylation, such a model would allow a controller to compute the feed adjustment required to reach a target glycoform in advance of the next measurement, rather than responding to a deviation after it has occurred. Real-time cell density measurement is conceptually straightforward using tools such as impedance sensors, which enable closed-loop control. In a perfusion reactor, if cell density surpasses the target, the bleed rate increases to bring density back down to the target, as already demonstrated (Richelle et al., 2022).

Applying this concept to glycosylation-based control, consider Figure 2A: if we propose to engineer a mAb with high galactosylation, a wild-type CHO cell will struggle to produce it, or may not be able to at all. Glyco-engineering the cell line to increase galactosylation makes the release target achievable. Monitoring these engineered cell lines then determines whether that target is actually reached and held. In an open-loop system, galactosylation itself is not monitored, so an operator’s decision to increase glycan-precursor feeding cannot guarantee the outcome. In a closed-loop system, feedback is integrated directly: the control strategy relies on PAT readouts of galactosylation and calculates precisely how much glycan precursor to feed to reach and hold the target glycoprofile in an automated fashion, removing the need for operator intervention and substantially increasing the probability of maintaining the target glycoform. Figure 2B shows the current state of closed-loop glycosylation control.

To date, the principal bottleneck is not measurement but the translation of PAT output into control action, which is constrained at each element of the loop. The first is loop dynamics. A change in feeding propagates to the harvested glycoform distribution over one or more vessel residence times, while the measurement is returned hours after sampling, so any deviation persists across a substantial volume of harvest before it is corrected. Control must therefore be predictive rather than reactive. The second is the absence of a validated interface between analytical instruments and the process control layer. Automated integration exists, but glycan results are not routinely delivered as qualified values that a control system can act upon, and emerging data standards remain unimplemented for this purpose. The third is that the system is underactuated. Several glycoform species must be held simultaneously, whereas the validated levers, principally nucleotide-sugar and cofactor feeding, perfusion rate, and temperature, are few and rarely orthogonal to growth and titer. A correction applied to one attribute will therefore perturb others. The fourth is regulatory. Conventional processes demonstrate consistency by operating to a fixed recipe, whereas a controller deliberately varies its actions in response to measurement, such that assurance of quality rests on the control system itself. That system, including the analytical method and any model driving it, must be qualified accordingly, and no such package has yet been assembled for glycosylation.

Discussion

The argument advanced throughout this perspective is, ultimately, one of systems integration. Cell line glycoengineering has delivered tools that expand the pharmacological space accessible through directed glycan design. PAT and continuous bioprocessing have developed analytical and operational frameworks that advance their manufacturing capability. The gap is not in either discipline individually; it is in the absence of their co-design.

Glycoengineered cell lines are typically developed and characterized at bench scale, with little forethought toward how they will perform in a manufacturing-scale or continuous bioprocess. This is more than a translation problem. The genetic modifications themselves can alter cellular metabolism, and an altered metabolic profile may require a different control strategy than the one designed around the parent unmodified cell line. Under these conditions, the glycoform landscape established at bench scale may not translate to a perfusion bioreactor at manufacturing scale. Conversely, a continuous process designed without a cell line characterized under the relevant process stresses leaves the PAT system without a stable biological reference for control. The result is a glycan specification met probabilistically rather than by rational design.

The mAb glyco-better concept makes this integration imperative, both commercially and scientifically. A molecule whose clinical differentiation rests on a defined glycoform cannot tolerate a process that delivers the target only within wide probabilistic bounds. The regulatory and commercial proposition of a glyco-better demands a manufacturing precision that sequential development cannot reliably deliver.

Protein engineering is not an alternative to glycan design so much as a parallel one. Fc substitutions such as YTE and LS extend persistence mediated by the neonatal Fc receptor (FcRn), and others silence or enhance FcγR engagement, but they act on the polypeptide and leave the glycan in place (Saunders, 2019). An Fc-engineered antibody remains a glycoprotein whose glycoform varies with process conditions and continues to influence effector engagement and clearance. Such a molecule can therefore itself be a glyco-better, provided its glycoform is the demonstrated basis of a pharmacological advantage and is carried forward as a specification. The two approaches carry different considerations, since sequence modification can alter biophysical properties such as CH2 thermal stability, with product-specific consequences (Dippel et al., 2023), whereas glycan specification requires process capability. The mechanism through which a product acts determines what is asked of the glycan, not whether it must be controlled.

Glycosylation targets should be defined at the earliest stages of candidate selection, and cell line engineering and process development should then proceed in parallel, with glycoform consistency as a shared primary objective. This requires organizational integration among cell biology, upstream bioprocess engineering, and analytical scientists, together with sustained investment in PAT to deliver glycan data on process-relevant timescales. Defining such a target additionally requires efficacy, pharmacokinetic, and safety data generated across deliberately varied glycoforms, which is resource-intensive and not yet common practice.

It is worth noting that the framework advanced here is not specific to monoclonal antibodies, although the IgG1 backbone is its most tractable case. IgG1 has a single conserved N-glycosylation site per heavy chain, so the control problem concerns a single glycan population. Fc-fusion proteins could be less forgiving. For example, etanercept has three N-glycosylation sites, and its glycosylation has proven a central obstacle to biosimilar comparability (Montacir et al., 2018). Since processing at each glycan site is constrained by its local steric environment, a shared nucleotide-sugar feed will not shift all sites uniformly, and the attributes to be held simultaneously rise faster than the available levers can accommodate. Feedback control of such molecules is therefore a harder problem. Proof of concept should first be established for mAb products, where the constraints are fewer. The resulting strategies would then require adaptation for proteins with multiple glycan sites that must be controlled in parallel.

The near-term horizon is promising. The first integrated glycan control platforms, combining glycoengineered cell lines with perfusion bioreactors operated under at-line, glycan-informed feedback control, may emerge in the coming years, enabled by robust, rapid analytics interfaces and validated hybrid models of CHO N-glycosylation. The biological knowledge to design mAb glyco-betters exists; the engineering knowledge to manufacture them consistently is converging. Cell line engineering sets the ceiling; the imperative is to develop the process capability that reaches it, together and from the start.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. The author declares that generative AI tools were used to support language editing, improve clarity, and fluency. The authors reviewed, edited, and verified all AI-assisted edits and take full responsibility for the accuracy, originality, interpretation, and integrity of the final manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Jayato nayak, Mahindra University, India

Reviewed by: Roger Y. Tam, Health Canada, Canada

Sina Sarsarshahi, SUNY Buffalo State University, United States

Ranjit Ranbhor, Odin Pharmaceuticals LLC, United States

Ren Liu, Merck, United States