Proteolysis and oxidation of therapeutic proteins after intradermal or subcutaneous administration

Abstract

The intradermal (ID) and subcutaneous (SC) routes are commonly used for therapeutic proteins (TPs) and vaccines; however, the bioavailability of TPs is typically less than small molecule drugs given via the same routes. Proteolytic enzymes in the dermal, SC and lymphatic tissues may be responsible for the loss of TPs. In addition, the TPs may be exposed to reactive oxygen species (ROS) generated in the SC tissue and the lymphatic system in response to injection-related trauma and impurities within the formulation. The ROS can oxidize TPs to alter their efficacy and immunogenicity potential. Mechanistic understandings of the dominant proteolysis and oxidative routes are useful in the drug discovery process, formulation development, and to assess the potential for immunogenicity and altered pharmacokinetics (PK). Furthermore, in vitro tools representing the ID or SC and lymphatic system can be used to evaluate the extent of proteolysis of the TPs after the injection and prior to systemic entry. The in vitro clearance data may be included in physiologically based pharmacokinetic (PBPK) models for improved PK predictions. In this review, we have summarized various physiological factors responsible for proteolysis and oxidation of TPs after ID and SC administration.

1. Introduction

Monoclonal antibodies (mAbs), human growth hormone (hGH), insulin, recombinant subunit vaccines and various other proteins are important therapeutic proteins (TPs).¹ These TPs are mainly administered via subcutaneous (SC) and intradermal (ID) routes due to their convenience compared to the intravenous (IV) route.² However, incomplete bioavailability is a major disadvantage for TPs given via non-IV routes.³ Proteolysis at the site of ID or SC injection and within the lymphatic system are major elimination pathways for TPs that reduce their systemic bioavailability.^3–6 The SC tissue consists of cells such as adipocytes, fibroblasts, T cells, mast cells, dendritic cells (DCs) and macrophages, which contribute towards proteolysis of the TPs (Figure 1). The SC tissue has a low level of basal proteolytic activity; however, the penetration of an injection needle and the subsequent hydrostatic pressure due to the injection volume can alter the SC tissue leading to secretion of serine proteases, threonine proteases, and metalloproteases by the tissue matrix, DCs, and other SC tissue cells.^7,8 Finally, intracellular processes like proteasomal catabolism after ubiquitin binding and lysosomal degradation may also lead to reduced bioavailability of the TPs.^7,9,10 In addition, immune reactions (e.g. generation of antibodies against the administered TP, allergy, anaphylactic reaction, accumulation of immune cells around the injection site) can enhance the proteolytic activity in SC tissue. These immune reactions are caused by the presence of non-self antigens, partially degraded (e.g. oxidized or deamidated) TP, manufacturing impurities, and aggregates of the TP.

Figure 1:

Physiological processes at the SC injection site.

a) Migration of DCs from the dermis to SC tissue due to penetration of the SC injection needle. b) Potential oxidation of the TP by the ROS. c) Proteolytic and lysosomal degradation of the TP in the SC interstitial and intracellular compartments. d) Uptake of the TP by lymphatic capillaries (lymph flow) and blood vessels (blood flow).

(TP: Therapeutic protein, O-TP: Oxidized TP, DC: Dendritic cells, M: Macrophages, ROS: Reactive oxygen species, PD: Proteolytic degradation, LD: Lysosomal degradation, SCC: Subcutaneous cells.)

Along with the SC injection site, the lymphatic system can also play a vital role in the proteolysis of TPs. Proteins with molecular weights greater than 16–20 kDa are usually absorbed at least partially via the lymphatic capillaries upon SC administration^3,11,12, from which they are trafficked through the lymphatic vasculature and lymph nodes (LNs) before entering the systemic circulation at the junction of the left subclavian vein and internal jugular vein or at the beginning of left brachiocephalic vein. The nodal sinuses of lymph nodes are predominant sites of antigen presentation, due to the large populations of antigen presenting cells (APCs) and naïve and mature B andT cells. The structure of LNs is responsible for a molecular sieving of passing materials, resulting in extensive interaction of proteins with the APCs.^3,13–15 Therefore, the endocytic uptake and lysosomal degradation may contribute towards loss of the TPs in the LNs. In addition, the immune cells in the lymphatic system may secrete protease enzymes¹⁶ leading to increased proteolysis of the TPs. Hence, the lymphatic system plays a major role in proteolysis of the TPs and reduction in their bioavailability.^3,13–15

Oxidation of TPs within the skin, SC space, and lymphatics may lead to altered biological activity, increased immunogenicity, modified secondary and tertiary protein structure, increased aggregation potential, altered pharmacokinetics (PK), and decreased neonatal Fc receptor (FcRn) binding affinity for the mAbs.^17,18 The amino acid residues methionine, phenylalanine, tryptophan, tyrosine, histidine, and cysteine are highly susceptible to oxidation by ROS generated by respiratory bursts of immune cells, especially the monocytes and neutrophils.¹⁹ However, we note that all amino acids are prone to oxidation by hydroxyl radicals²⁰, which can be produced by secondary processes from ROS generated during the respiratory burst, i.e. via the Fenton or Fenton-like reactions.^21,22 Furthermore, cellular damage, e.g. physical rupture by the injection needle, osmotic and hydrostatic stress, can generate signals of wounding which include secretion of proinflammatory cytokines (IL-1β, TNF-α and IL-6). In addition, DCs and macrophages can initiate an innate immune response. The DCs and macrophages continuously ingest proteins within the tissue and display protein fragments as potential antigens. Localized irritation and immune reactions are frequently observed in the SC tissue after injection of the TPs. The mast cell population is approximately 7000/cm³ and can be responsible for immune responses against allergens possibly present in the TP formulation.⁷ A specialized subpopulation of DCs termed as the Langerhans cells (LCs) constitute around 1% of the total epidermal cell population; however, they occupy approximately 20–25% of the surface area. Various cells in the SC tissue are also involved in the adaptive immune response and antigen presentation to naïve T cells in the LNs.^23–25 The innate and adaptive immune response against the TPs is usually disadvantageous, as it may result in increased proteolysis and generation of ROS leading to the oxidation of TPs (Figure 1). However, in the case of proteins derived from pathogenic agents, this process is used successfully for recombinant protein and sub-unit based vaccines to enhance the immune response and pharmacological action.²³

2. Proteolytic enzymes in the skin, SC tissue and lymphatic system

The proteolytic enzymes in the ID or SC injection site and lymphatic system are largely responsible for reduced bioavailability of TPs.^3–6 This section describes various proteolytic enzymes present in the skin, SC tissue and lymphatic system. In addition, impact of species, age, and pathological conditions on expression of the proteolytic enzyems is also summarized. The altered expression of proteolytic enzmes due to disease or age can lead to changed bioavailability and PK after ID or SC administration.

Proteolytic enzymes (peptidases) are broadly categorized as endo-peptidases and exo-peptidases based on whether they preferentially cleave the peptide backbone distant from the termini (Supplementary Figure 1a) or at the penultimate or terminal peptides, respectively. Proteolytic enzymes can be further classified based on the amino acid present at the catalytic site of the enzyme. The serine, cysteine, aspartic acid, and metallo-proteases are the major categories having different catalytic sites (Supplementary Figure 1b). Detailed classification of proteolytic enzymes was reported elsewhere.^26–29

2.1. Proteolytic enzymes in skin and lymphatic system

More than 30 proteolytic enzymes are present in the skin and SC tissue. Skin has two major layers, the epidermis and the dermis. In the epidermis, cysteine and serine proteases form the major fraction of the total proteolytic enzymes, followed by aspartic and metalloproteases (Supplementary Figure 2).³⁰ The rat epidermis has approximately 10 times more proteolytic activity as compared to the dermis, and proteolytic enzymes are found in the epidermal cells. In the case of dermis, proteolytic enzymes are present in the infiltrating immune cells, sebaceous glands, and follicular cells.³¹

The skin is densely packed with various immune cells and is an immune organ itself.²⁴ The cutaneous surface of normal adult humans has around 20 billion T lymphocyte cells, which can secrete proteolytic enzymes leading to degradation of the administered TPs.^30,32 The proteolytic enzymes such as tryptase, cathepsin G, neutrophil elastase originate from the T cells and other infiltrating immune cells.³⁰ Caspase-14 is a major cysteine protease, expressed in the human epidermal keratinocytes. Serine protease HtrA1 is found ubiquitously in many organs; but increased amounts are found in the mature layers of the epidermis.³³ Cathepsin D and E are found in various organs including the LNs, spleen, kidney, lung, and liver.^34,35 Cathepsin G has been identified in the rat spleen and LNs.³⁶ Dipeptidyl peptidase IV (DPP-IV) was expressed more in the cultured human dermal lymphatic cells compared to the blood vascular endothelial cells.³⁷

The results of recent proteomic analyses of human and animal lymphs have shed greater insight into the composition of the lymph and shown that its composition is distict from the interstitial fluid and plasma. ³⁸ In addition, the proteins in lymph change qualitatively and quantitatively due to pathological conditions.³⁸ Hansen et al. reported that 17% of the proteins found in rat mesenteric lymph were protease inhibitors,^38,39 and levels and activity of matrix metalloproteases and cathepsin increased in rat lymph after hemorrhagic shock and trauma.^38,40,41 The ratio of serine protease to antiprotease in lymph was also altered due to hemorrhagic shock.

2.2. Altered levels of proteolytic enzymes due to species and age

Preclinical safety testing of the TPs is typically conducted in rodents, one or more large animals like the dog, and possibly non-human primates, although these preclinical models have been found to poorly represent human bioavailability and PK of numerous TPs.⁴² Kageyama et al. reported that the expression of cathepsin D and E in spleen and LNs is variable across the species like rat, monkey and humans. The monkey spleen tissue had around two-fold higher levels of cathepsin D compared to the rats. In the case of monkey LNs, cathepsin D was 4-fold higher as compared to rat LNs. In contrast, the rat spleen had around 40-fold higher expression of cathepsin E compared to the monkey spleen. The rat LNs had around 11-fold higher cathepsin E expression than the monkey LNs.³⁵ These differences in protease expression may contribute towards variations in overall clearance of the TPs leading to different values of bioavailability across the species. Further, an age dependent increase in elastase-like endopeptidase activity was seen in the mouse skin. This increase in peptidase activity is important during the first 6 to 15 months of the life and can be attributed to the metallo-endopeptidases.⁴³

2.3. Altered levels of proteolytic enzymes due to the pathological conditions

We hypothesize that the altered skin and lymphatic levels of proteolytic enzmes due to disease conditions can impact the fate of TPs in terms of bioavailability. Unfortunately, there is little literature-based evidence to substantiate this claim. However, changes in expression of proteolytic enzymes are known to occur in diferent disease states as described in this section. The altered levels of proteases can affect the fate of the TPs after ID or SC administration in various diseases as compared to the healthy condition.

Expression of proteolytic enzymes in the skin and SC tissue is changed during various diseases as summarized in Table 1.^30,54 Papillon-Lefèvre syndrome (PLS) is associated with periodontitis and thickening of epidermis.³⁰ Lysosomal cysteine proteases such as cathepsins-B, -H, -L and -S are overexpressed in the tumor cells of Morbus Hodgkin’s lymphoma, and expression of cysteine proteases increases with increasing malignancy of the lymphoma.⁴⁵ Ichthyosis hypotrichosis syndrome of skin leads to loss of function of matriptase (a serine protease).⁴⁶ In the case of inflammatory skin and bowel disease, metalloprotease 17 was not expressed.⁴⁷ In another example, epilysin, or matrix metalloprotease-28, expression is induced after injury to the epidermis as part of the wound healing process.⁴⁸ Epilysin expression was also increased in the cartilage and synovium of the patients with osteoarthritis.⁵⁵ Furthermore, expression of matrixmetallprotease 1 was increased in the recessive dystrophic epidermolysis bullosa.⁴⁹ Leucine-N-aminopeptidases and arginine-N-aminopeptidases, found on the surface of fibroblasts, were elevated up to 1.5- to 6.6-fold, respectively, compared to healthy fibroblasts in patients with psoriasis and rheumatoid arthritis. While, DPP-IV was elevated by 4.5-fold compared to the healthy ibroblasts.⁵¹ In another example, the proteolytic activity in rat liver and kidney cytosol was impaired by diabetes.⁵⁶ Lymphatic expression of proteases and enzymes is also altered during some diseases. The LNs were abnormally enlarged in the mouse model of lymphadenopathy. In addition, the serum DPP-IV levels increased in the enlarged LNs, and the levels were directly proportional to the weight of the LNs.⁵⁰

Table 1:

Proteolytic enzymes altered due to various pathological conditions (Adapted and modified using a review published by Veer at al³⁰ with permission)

Pathological condition	Proteolytic enzyme	mRNA or protein content or activity of the peptidase(s) compared to the healthy condition	Organ or tissue or cell population	Reference
Papillon-Lefèvre syndrome	Cathepsin C	Approximately 85% reduction in the enzyme activity, possibly due to genetic modifications	Leucocytes	Romero-Quintana et al.44
Morbus Hodgkin’s lymphoma Ichthyosis hypotrichosis syndrome	Cathepsin B, H, L, S (cysteine proteases)	Increased expression	Lymph nodes	Kirschke et al.45
	Matriptase (a serine protease)	Mutation leading to loss of function	Skin	Basel-Vanagaite et al.46
Inflammatory skin and bowel disease	Metalloprotease 17	Mutation leading to loss of function	Skin and bowel	Blaydon et al.47
Skin injury	Matrix metalloprotease 28 (epilysin)	Induction after skin injury	Skin (epidermis)	Saarialho-Kere et al.48
Recessive dystrophic epidermolysis bullosa	Matrixmetallprotease (MMP) 1	Elevated expression	Skin (epidermis)	Titeux et al.49
Lymphadenopathy	*DPP-IV (a serine protease)	Increased expression in enlarged lymph nodes	Lymph nodes (aberrant T cells without CD4 or CD8)	Kubota et al.50
Psoriasis and rheumatoid arthritis	Arg-, Ala-, and Leu- N-aminopeptidases and DPP-IV	Increased expression	Fibroblast	Raynaud et al.51
Psoriasis and seborrheic dermatitis patients	Caspase-14	Inactive procaspase-14 along with the active caspase-14	Immature parakeratotic skin	Fischer et al.52
Allergic contact dermatitis and bullous pemphigoid	#Tryptase (a serine protease)	Higher levels in the skin blister fluid	Mast cells in the skin	Brockow et al.53

^*Increased DPP-IV levels in serum were correlated with enlargement of the LNs. MRL-lpr mice were used as a disease model for lymphadenopathy

^#Higher tryptase levels compared to the blister fluids obtained from insect bite reactions, erysipelas, burns, toxic epidermal necrolysis

(DPP-IV: Dipeptidyl peptidase IV, LN: Lymph nodes)

Caspase-14 is a predominant caspase in the human stratum corneum, but proteolytic activity and expression levels can be significantly different between healthy and diseased skin.⁵². The immature parakeratotic skin from psoriasis and seborrheic dermatitis patients contained inactive procaspase-14 along with the active caspase-14 form. In contrast, the healthy skin contained only the active form, i.e. caspase-14. Tryptase is a serine protease abundantly present in the mast cells of the skin. The skin blister fluid from the allergic contact dermatitis and bullous pemphigoid patients contained higher levels of tryptase compared to other inflammatory skin conditions, including, insect bite reactions, erysipelas, burns, and toxic epidermal necrolysis. The higher level of tryptase denotes that mast cell degranulation was the pathophysiological mechanism of allergic contact dermatitis and bullous pemphigoid.^53,57 In addition, mouse skin exposed to UVA radiation showed a greater increase (+90 to 122%) in peptidase activity compared to UVB exposure (+72%). The serine endopeptidases and metalloprotease enzymes were affected by the UVA radiation.⁴³ In another study, it was found that UVB exposure increased secretion of matrix metalloprotease-2 by human dermal fibroblasts.⁵⁸

2.4. Use of protease enzymes for targeted drug delivery

Protease enzymes present in the skin, SC and lymphatic tissues may be important for targeting specific cell population or tumors. For example, the protease enzyme tryptase is abundant only in the mast cells. Other white blood cells, such as peripheral blood leukocytes and basophils, have either none or very small amounts of tryptase.^57,59 Solid tumors of epithelial origin are known to express various peptidases including urokinase, matriptase, and legumain. This physiological phenomenon was used to engineer a pro-mAb, which was masked with a peptide attached to the light chain of the mAb cetuximab via a linker peptide, which was a substrate of the tumor specific proteases. This strategy reduced non-specific toxicity of the mAb in primate studies. The drug activation was also observed ex vivo in tumor specimens from patients.⁶⁰ Additional examples of cleavage of antibody-drug conjugates to release the small molecule anti-cancer agent are described in the sections 3.1 and 3.2. Choi et al have listed various cancers where proteolytic enzmes were overexpressed. For example, cathepsin B is overexpressed in head and neck cancer and, melanoma.⁶¹ In another study, it was reported that type 1 matrix metalloproteianase increased by 2.6-fold in the carcinoma-associated fribroblasts compared to normal fibroblasts.⁶² This can be leveraged to design targated delivery of TPs to tumors.⁶³

Key points and unknowns

Protease enzymes play a crucial role in the clearance of TPs and they are altered in various animal species, with different age and disease conditions. The increased or decreased protease activity in the skin, SC tissue and lymphatic system may lead to differences in PK of the TPs. However, impact of the altered expression of protease enzymes on PK of TPs is not clearly known. It is important to identify the proteolytic enzymes responsible for the reduced bioavailability of TPs, because the reduced ID and SC bioavailability of TPs may be mitigated by co-administration of peptidase inhibitors. For example, when an ointment of peptidase inhibitors (nafamostat or gabexate) was applied before SC administration of a TP, the peptidase inhibitors penetrated the injection site and decreased TP degradation.⁶⁴ This strategy has been explored also for oral and pulmonary delivery of proteins and peptides.^65,66

3. In vitro models to evaluate TP proteolysis and predict bioavailability

Various in vitro systems (whole blood, serum, plasma, liver and kidney homogenates), have been used to study proteolysis of the TPs.⁶⁷ The tissue homogenates and recombinant enzymes are most commonly used tools to evaluate stability of the TPs. The metabolism of small molecule drugs (SMDs) in the liver and other organs has been extensively studied using in vitro systems to obtain quantitative data.⁶⁸ Further, this in vitro data is commonly used in the PBPK models to predict PK of SMDs.⁶⁹ However, similar quantitative information about the metabolic processes of TPs in the SC or ID injection sites and lymphatic system is not available.^2,70–73 Therefore, it is difficult to use the in vitro TP metabolic data from currently available models in the prediction of PK after SC or ID administration. In this section, in vitro techniques used to study proteolysis and bioavailability of TPs are summarized.

3.1. In vitro tools to study proteolysis of TPs

Wroblewski et al. used in vitro models to demonstrate that decoy receptor 3 was proteolytically cleaved between amino acids 218 and 219. This information was used to improve the TP stability, by replacing Arg-218 with a glutamine. The modification resulted in reduced proteolytic degradation and increased stability of the TP.⁷⁴ In another report, DPP-IV and neutral endopeptidase were used to find degradation products of human glucagon like peptide-1 (GLP-1) and its modified analogue, liraglutide. Liraglutide degraded into a single metabolite in the presence of DPP-IV, while multiple peptides were formed after incubation with neutral peptidase. Rate of liraglutide degradation was slower compared to GLP-1, while, the cleavage site for liraglutide was similar to that of GLP-1. This study helped to understand comparative stability of various peptide analogues and to find cleavage sites for the peptides.⁷⁵

Proteolysis of TPs usually results in reduced bioavailability after SC administration; however, proteolysis is occasionally important for the pharmacological action TPs, such as antibody-drug conjugates (ADCs). Bessire et al. at Pfizer have developed an in vitro model for ADC metabolism and drug release using lysosomal enzymes from human liver S9 fraction. ⁷⁶ Their results demonstrated the choice of enzymes in release studies impacts the release kinetics and species. Over an 18 hour incubation, the human S9 fraction, human cathepsin B, human cathepsin L and rat liver crude lysosomal fraction (CLF) all released auristatin D conjugated a human anti-her2 IgG1 antibody via a cleavable valine-citrulline-p-amino benzyl carbamate linker (Figure 2), although cathepsin B was significantly slower. Human liver cathepsin D failed to release the drug. Several ADCs utilize non-cleavable linkers that rely on non-selective proteolysis and lysosomal degradation to metabolize the conjugate and release the drug cargo ^76,77. Ado-trastuzumab (Kadcyla) is an ADC approved by the FDA that uses a noncleavable linker.⁷⁸ In addition, an ADC conjugated with monomethylauristatin F (MMAF) using a non-cleavable maleimidocaproyl linker was equally potent compared to an ADC linked with the cleavable Val-Cit-PABC linker.⁷⁹ Bessire et al. studied a maleimide-caproyl-mono-cysteine-conjugated human IgG1 - auristatin F ADC that released the auristatin F over 24 hours in mouse liver CLF, purified human cathepsin B and human liver S9 fraction. ⁷⁶ However, cathepsin B resulted in the least release, and cathepsins D and L failed to release substantial drug. The group then went on to generate novel ADCs utilizing clevable and non-cleavable linkers with different drugs. Cathepsin B was the least effective in cleaving the drugs, especially in the case of the non-cleavable lysine linked (amide) ADCs. Bessire et al. also identified different fragmented peptide metabolites of the antibodies in the various enzyme cocktails. These results demonstrate that the choice of enzymes and source used for in vitro assays strongly influence the release kinetics of the ADC payload and antibody metabolites.

Figure 2:

Release of MMAE from the ADC by cathepsin B or lysosomal proteolysis

3.2. Bioavailability of TPs after ID or SC administration

Recombinant peptidase enzymes (e.g. DPP-IV and neutral endopeptidase), tissue homogenates and tissue fractions including human liver S9 fraction were used to examine proteolysis of TPs.⁷⁵ A higher rate of in vitro proteolysis generally correlated with lower bioavailability. For example, leuprolide acetate had a SC bioavailability of 32%, and in vitro incubation of leuprolide acetate with a rat skin homogenate lead to around 90% degradation within 120 min. This may indicate that the skin is mainly responsible for the proteolysis of certain TPs.⁸⁰

Erythropoietin is a TP with about 20–30% SC bioavailability in humans and 30–38% in rats.^13,81 Wang et al. used rat SC tissue homogenate and a LN cell suspension to study the degradation of erythropoietin. The incubation of erythropoietin with the SC tissue homogenate and LN cell suspension resulted in appearance of smaller fragments of the TP and decreased intensity of the parent erythropoietin after 24 h as shown by SDS-PAGE analysis, which indicated that the reduced bioavailability was probably due to proteolysis in the SC tissue and LNs.¹³ In another study, after SC injection of human growth hormone (hGH) to a lymph duct cannulated sheep, 61% of the hGH was collected in the peripheral lymph, while around 8% was collected in the central lymph. Total recovery of hGH by the lymphatic system was around 93%, indicating minimal contribution of SC injection site in the proteolysis.⁸², and leading to a conclusion that most degradation occurred within the lymphatics.

Kinnunen et al. reported a novel in vitro platform for simulating the SC injection site. The platform consists of a dialysis based SC injection chamber, which can be filled with multiple non-cellular SC tissue components such as hyaluronic acid. Another chamber with a physiological buffer served as an infinite sink representing rest of the body. However, this system did not utilize proteolytic enzymes or cellular components responsible for proteolysis of the TPs. The platform may be used to predict behavior of the TP formulations, and to measure rate of release of the TPs from the SC injection site.⁸³

Key points and unknowns

Various in vitro methods are useful to examine proteolysis of TPs, but rarely to date provide quantitative proteolysis information. Therefore, bioavailability cannot be mechanistically estimated using the available in vitro systems.⁷³ These quantitative data can be used to define the rate of metabolism from the ID or SC injection sites and the lymphatic system. In order to develop in vitro system for quantitiative understanding of bioavailability, it is necessary to understand various types of cells like macrophages, lymphocytes, SC tissue cells to which the TPs are exposed. The quantitative data from appropriate in vitro systems can be used to develop a PBPK model for PK predictions.

4. Factors affecting PK of TPs after ID or SC administration and PBPK modeling

This section describes various physiological parameters related to skin, SC tissue and lymphatic system, which can potentially control the fate of TPs. These parameters and impact of disease condition, exercise, species, and site of injection on these parameters can be used to build a hypothesis-driven mechanistic PBPK model.

4.1. Effect of molecular weight on lymphatic uptake of TPs

After ID or SC administration TPs usually travel through the lymphatic vessels and LNs before reaching the systemic circulation.⁸⁴ The uptake of TPs via lymphatic capillaries after ID or SC administration is dependent on the molecular weight, and TPs greater than 16 kDa are generally absorbed via lymph vessels.¹⁵ Supersaxo et al. reported a linear correlation between molecular weight and percent dose recovered in the lymph fluid of sheep after SC administration. (Figure 3). rIFN alpha-2a (19 kDa) had a percent lymphatic recovery of approximately 60%.¹¹ In another study, hGH (22 kDa) had lymphatic recovery of around 62% in the sheep model.⁸² Wu et al reported that half-life (time for 50% loss of protein from the SC injection site) was proportional to the molecular weight of the protein in mice. Therefore, molecular weight of the protein is inversely proportional to the rate of absorption via blood capillaries and lymph vessles. The authors tested proteins with molecular weight ranging from 23 to 149 kDa.⁸⁵ The molecular weight of TPs determines the percent lymphatic uptake and rate of absorption from the injection site. All of these studies indicate that molecular weight is important factor for the lymphatic transport and to understand disposition of the TPs after SC administration. In general, TPs with molecular weight greater than 24 kDa would have more than 80% absorption via lymphatic capillaries. Further, for TPs with molecular weight greater than 23 kDa, with increasing molecular weight, the rate of aborption would decrease. Therefore, molecular weight can be leveraged to design TPs with desired PK properties.

Figure 3:

Effect of molecular weight on % lymphatic recovery of various proteins after SC administration to sheep.

These data were originally reported by Supersaxo et al.¹¹ and Charman et al.⁸² The lymph was collected from the right popliteal lymph node of a sheep after SC administration of the protein on lower side of the right hind leg.

(Reproduced with permission from a review article by Porter et al.³)

4.2. Effect of animal species on PK of TPs

The PK of TPs may be highly variable between different animal species.⁸⁶ A minipig animal model may be a more suitable preclinical species for the human PK predictions compared to rodents and other small mammals, because minipigs and humans have a similar skin structure and lymphatic plexus. The thicknesses of the epidermis and stratum corneum, and lipid contents of the stratum corneum are similar in humans and minipigs. However, the absorption rate of TPs from the SC injection site was 2 to 5-fold higher in minipigs compared to humans. In addition, the bioavailability of adalimumab was different in minipig and humans (Table 2).^2,87 The major differences between SC space of human and other animal species were mentioned previsouly by McDonald et al.⁸⁶ In brief, these differecens are mainly linked with the panniculus carnosus muscle. This striated muscle is absent in humans, while it is prominent in lower species, and reduced in non-human primates. Further, the mobility of skin is higher in scruff animal species compared to humans.^2,86 These physiological differences in skin and SC space can potentially lead to inaccurate prediction of PK using allometric scaling. Hence, mechanistic understanding of proteolysis and absorption processes in the SC tissue of each species is important. In addition, it is known that the catalytic products of TPs do not differ substantially between various species. However, rates of proteolysis of TPs are generally higher in lower animal species as compared to humans.⁸⁸

Table 2:

Bioavailability, T_max, and C_max of various TPs and their relationship with the injection site, population, species or external conditions

TP	PK Parameter	Route	Injection site (Condition/Species)	Reference
	Bioavailability (%)
Mepolizumab	64	SC	Abdomen (Human)	Ortega et al.93
Mepolizumab	75	SC	Arm (Human)
Mepolizumab	71	SC	Thigh (Human)
Mepolizumab	81	IM	NA (Human)
Rituximab	44	SC	Abdomen (Rat-10 mg/kg)	Kagan et al.5
Rituximab	31	SC	Back (Rat-10 mg/kg)
Rituximab	18	SC	Back (Rat-40 mg/kg)
Adalimumab	64	SC	NA (Human)	Zheng et al.87
Adalimumab	83	SC	Inguinal (Minipig)
Darbepoetin alpha	106	SC	Interdigital (Sheep)	Kota et al.12
Darbepoetin alpha	85	SC	Abdomen (Sheep)
Darbepoetin alpha	92	SC	Shoulder (Sheep)
	Tmax (min)
Insulin	92	SC	Thigh (Human)	Berger et al.92
Insulin	76	SC	Arm (Human)
Insulin	63	SC	Abdomen (Human)
Insulin	30	SC	Thigh (Hot water bath-Human)
Insulin	135	SC	Thigh (Cold bath- Human)
Insulin	38	SC	Thigh (Local massage)
Insulin lispro	60	SC	Abdomen (Normal weight-Human)	de la Peña et al.94
Insulin lispro	90	SC	Abdomen (Obese Human)
	Tmax (h)
Darbepoetin alpha	6–8	SC	Interdigital (Sheep)	Kota et al.12
Darbepoetin alpha	18–72	SC	Abdomen (Sheep)
Darbepoetin alpha	6–12	SC	Shoulder (Sheep)
	Cmax (pmol/L)
Insulin lispro	822	SC	Abdomen (Healthy normal weight-Human)	de la Peña et al.94
Insulin lispro	560	SC	Abdomen (Obese Human)

4.3. Effect of the injection site on PK of TPs

The SC injection site may be partially responsible for the proteolysis of TPs by secreting proteolytic enzymes in the extracellular matrix as well as by intracellular proteolysis. According to the published studies, hGH and insulin were degraded at the injection site.^89,90 Further, the bioavailability of hGH in humans was higher when injected in the abdomen compared to the thigh.⁹¹ The bioavailability of insulin was different after SC injection at various injection sites (abdomen, thigh, and arms) (Table 2). The rate of absorption (indicated by T_max) was higher for abdominal injection (63 min) compared to the arm (76 min) and thigh (92 min).⁹² In another study, bioavailability of mepolizumab was 64, 75, 71 and 81% after SC abdomen, SC arm, SC thigh and intramuscular (IM) administration, respectively.⁹³ The bioavailability of rituximab (10 mg/kg) in rats was greater after abdominal SC injection (44%), while it was lower (31%) after SC injection in the back. The higher SC dose (40 mg/kg) of rituximab in the back lead to bioavailability of 18%. This indicates saturation of FcRn-mediated protection from proteolysis.⁵

The rate of absorption of darbepoetin alpha was different for various SC injection sites like interdigital space, abdomen and shoulder. T_max for interdigital space, abdomen and shoulder was 6–8, 18–72, and 6–12 h., respectively. These differences may be attributed to alterations in the lymphatic vasculature and characteristics of the SC tissue at various injection sites. In addition, the authors reported that for the SC injection via interdigital space, the lymphatic pathway was a major absorption route (around 90%). While in the case of abdominal SC injection, the lymphatic absorption was around 67%.¹²

The studies described above indicate that injection site is an important factor governing bioavailability of the TPs. Change in the bioavailability with different injection sites may be due to differences in the proteolytic degradation in the ID or SC tissue and lymphatic system or variation in the uptake of the TP by lymphatic plexus. A few reports showed that bioavailability, and PK profile of the TP was similar when various injection sites such as abdomen, arm and thigh were compared.^95,96 Nonetheless, ID or SC injection site and the lymphatic system are important for proteolysis of the TPs and can impact their bioavailability and other PK parameters.

4.3. Physiological factors responsible for absorption and proteolysis of TPs

The physiological factors related to the SC injection site are important to govern the transport of TPs. For example, obese rats had a bioavailability of around 19%, while normal rats had a bioavailability of around 45% after SC administration of PEGylated erythropoietin. The lymphatic uptake in the fatty/adipose tissue may be reduced leading to reduced bioavailability. The proteolytic activity may also change due to the fat content; however, this remains to be investigated further. Higher body weight was associated with lower SC bioavailability in humans¹³ and the SC tissue thickness correlated with the body mass index (BMI). In general, the thickness increased with increasing BMI, and females had greater thicknesses as compared to males (Figure 4).⁹⁷ These differences in the SC tissue thickness may lead to altered rate and extent of the absorption. Injection depth relative to the underlying muscle layer will vary with the SC tissue thickness, but it has been shown that the PK of insulin lispro was independent of the injection depth in normal and obese patients.⁹⁴ In addition, it can be speculated that the body weight and SC tissue thickness are positively correlated with each other. Therefore, it needs to be examined further if reduced bioavailability due to increased body weight was in fact due to increased SC tissue thickness. In another study, it was reported that obese mice had significantly lower lymphatic fluid transport.⁹⁸ This alteration in the lymph flow may also change transit of the TPs via the lymphatic system and affect overall PK. The TPs would generally first travel through the initial lymphatics (lymphatic capillaries), next through the pre-collecting lymphatics and finally through the collecting lymphatics (lymphatic trunks) and lymphatic ducts. Therefore, lymph flow rates and dimensions of the lymphatic vessels are important physiological factors responsible for lymphatic transport of the TPs.⁸⁴

Figure 4:

SC tissue (insulin injection site) thicknesses in type-2 diabetes adult patients

Each data point represents mean and standard deviation. BMI: Body mass index (kg/m²), Male BMI <17 (n=8), female BMI<17 (n=4), male BMI 17–19 (n=11), female BMI 17–19 (n=3), male BMI 19–23 (n=29), female BMI 19–23 (n=14), male BMI 23–25 (n=11), female BMI 23–25 (n=5), male BMI >25 (n=9), female BMI >25 (n=7).

This data was originally reported by Jain et al.⁹⁷ (Reproduced with permission).

The flow rates and physiology of the lymphatic system play an important role in determining the PK of TPs. In a recent report by Varkhede and Forrest, lymph flows and physiological data related to the lymphatic system were used in a minimal PBPK model to understand the clinical PK of several mAbs.⁸⁴ Lymphatic capillaries have diameters of around 10 to 60 µm (human) and are connected to the lymphatic trunks via pre-collecting lymphatics.⁹⁹ The pre-collecting lymphatics have diameters ranging from 35 to 150 µm or sometimes 300 µm (measured in human thigh).¹⁰⁰ The lymphatic trunks have diameters of around 1–2 mm in the lower limbs. Lymphatic trunks originating from the right-hand side of the head, thorax and right arm; join the right lymphatic duct, which meets the systemic circulation in the right subclavian vein. The lymphatic vessels from the rest of the body (left-hand side of the head, left arm, and lower limbs) are connected to the thoracic duct, which meets systemic circulation at the junction of jugular and left subclavian veins.⁹⁹

Cutaneous lymph flow rates in melanoma patients were different in various regions of the body as shown in Table 3. Hence, absorption of TPs may differ when administered at different injection sites. Skin lymph flow rate in the leg and foot region were approximately 7-fold greater than the head and neck region.¹⁰² In another publication, mean cutaneous lymphatic transport rate was reported as highest in the lower limbs (4.07 ± 0.35 cm/min), followed by upper limb (2.67 ± 0.33 cm/min), trunk (1.79 ± 0.47 cm/min), and, head and neck region (1.11 ± 0.22 cm/min) (Figure 5).¹⁰¹

Table 3:

Cutaneous lymph flow rates in various body regions in humans

Site	Average flow (cm/min)
Head and neck	1.5
Anterior trunk	2.8
Posterior trunk	3.9
Arm and shoulder	2.0
Forearm and hand	5.5
Thigh	4.2
Leg and foot	10.2

These data were originally published by Uren RF¹⁰² (Reproduced with permission).

Figure 5:

Cutaneous lymph flow rates in various regions of the human body

These data were originally published by Fujiwara et al.¹⁰¹ (Reproduced with permission)

The pre-nodal lymph flow measured in the healthy human leg joint was 13.2 ± 1.1 mL/24 h.¹⁰³ In another study, average afferent lymph flow rate was 10.4 cm/min (human).¹⁰⁴ In the case of sheep, popliteal efferent lymph flow rate was 2 to 10 mL/h.¹⁰⁵ Quin et al. also reported afferent (4.58 ± 0.6 mL/h) and efferent (3.87 ± 0.58 mL/h) lymph flow rate for the popliteal lymph node in a sheep.¹⁰⁶ Efferent lymph flow for the iliac lymph node in rats was 66 ± 6 µL/h while efferent lymph flow rate from the mesenteric lymph node in rats was 132 ± 41 µL/h.¹⁰⁷ The afferent lymph flow rate in dogs was 19.1 ± 0.3 µL/min.¹⁰⁸

In addition, the thoracic lymph duct and right lymphatic ducts are important because most of the lymph generated in the body is transported through either of the two lymph ducts. The thoracic lymph duct flow rate is variable across various species (sheep-5.4 ± 3.1 ml/min, rat-0.038 ml/min).^109,110 The thoracic duct diameter was 3.6 ± 0.1 mm in healthy humans.¹¹¹ In another report, the diameter was reported as a range of 2 to 5 mm. The length of the thoracic lymph duct is around 45 cm.¹¹² Thoracic duct lymph flow was 0.7 to 1.6 mL/kg/h in humans.¹⁰⁴ Thoracic duct lymph flow in mouse was 0.83 to 1.66 mL/h, while it was 1 to 10 mL/h¹¹³ or 2.3 mL/h¹⁰⁹ in rats. Various literature values of the thoracic lymph duct flow rates were mentioned by Giragossian et al.¹¹⁴ Length of the thoracic lymph duct in rats is 3 to 7 cm.¹¹³ While, internal diameter of the rat thoracic duct was 405.5 ± 10.1 µm¹¹⁵ or 0.4 to 1 mm.¹¹³ This variation may lead to differences in the PK across and within the species.

It is possible that the number of LNs to which the TP is exposed, would differ after SC injections at various injection sites. For example, if the SC injection site were the lower limb or thigh, the TP would travel through the inguinal and iliac LNs. If the TP were administered on the arm, it would pass through the axillary and bronchomediastinal trunk LNs. In both of the cases, number of LNs through which the TP has to pass before reaching the systemic circulation would be different. Due to this differential lymphatic exposure, extent of the lymphatic proteolysis would differ after SC injection at various injection sites.⁸⁴

The cutaneous blood flow rates may be important for the absorption of the smaller molecular weight (< 22 kDa) TPs, although they have minimal effect on the larger TPs (e.g. mAbs). Differences in the skin blood flow may result in altered degree of TP absorption and proteolysis.^86,102,116 Dermal blood flows were different in various animal species and in specific regions of the body such as buttocks, ear, humeroscapular joint, thoracolumbar area, and abdomen (Table 4).^117,118

Table 4:

Dermal blood flow (milliliters/minute per 100 gm) in various animal species and in specific regions of the body

Site	Mouse	Rat	Dog	Pig	Monkey
Buttocks	3.88±0.92	4.20±1.05	2.21±0.67	3.08±0.48	3.12±0.58
Ear	1.41±0.48	9.13±4.97	5.21±1.53	11.7±3.02	20.93±5.37
Humeroscapular area	10.10±3.51	6.22±1.47	5.52±1.31	6.75±2.09	8.49±3.28
Thoracolumbar area	20.56±4.69	9.56±2.17	1.94±0.27	2.97±0.56	2.40±0.82
Ventral abdomen	36.85±8.14	11.35±5.53	8.78±1.40	10.68±2.14	3.58±0.41

Each value represents mean and standard error.

This data was originally reported by Monteiro-Riviere et al.¹¹⁷ (Reproduced with permission).

4.3. Effect of the disease and exercise on the physiological parameters related to the lymphatic system

Exercise and disease state both effect the flowrates of lymph. The thoracic lymph flow rate was 1 mL/min in the case of patients without cirrhosis, while, it was 3 to 6 times higher in the cirrhosis patients.¹¹⁹ The thoracic lymph duct flow rate was also measured in pigs, and the impact of lipopolysaccharide (LPS)-induced sepsis was studied. Normal thoracic lymph duct flow in pigs was around 1.5 mL/min; however, the lymph flow rate was doubled after IV LPS administration. This increase in lymph flow was attributed to increased intra-abdominal pressure. LPS sepsis can also increase permeability of the lymphatic capillaries.¹²⁰ Another study reported that the nodal lymph flow in nodes injected with cancer cells was reduced compared to the normal nodes in rats. In the case of normal LNs, lymph flow was 1.49 ± 0.64 mL/gm/min, while it was 0.5 ± 0.24 mL/gm/min for cancer cell injected nodes.¹²¹ The pre-nodal lymph flow was 13.2 ± 1.1 mL/24 h for healthy humans (measured in leg joint). However, this lymph flow increased to 22.6 ± 3.2 mL/24 h for rheumatoid arthritis patients.¹⁰³ Thoracic duct has diameter of 3.6 ± 0.1 mm in healthy humans. The diameter increased (4.8 ± 0.4 mm) in the case of patients with portal hypertension.¹¹¹

Physical activity or exercise, heat, massage, increase in hydrostatic pressure in the lymph vessels may result in elevated lymph flow. In contrast, cold, lack of movement and external pressure may lead to reduced lymph flow.^86,102,116 The thoracic lymph duct flow in dogs increased while they were exercising. Normal thoracic duct lymph flow was 1.7 mL/min, while, during the exercise (1.5 miles/h on a treadmill), the lymph flow increased by 121% (3.9 mL/min). Further increase in the treadmill speed (10 miles/h) increased the lymph flow by 419% (9 mL/min).¹²² In a PK evaluation, the hot water bath and massage elevated the serum insulin levels, while, the cold water bath delayed the absorption (Table 2).⁹² Overall, the disease states and exercise can alter various lymph flows, which may lead to changes in the PK (Table 5).

Table 5:

Effect of disease or exercise on various lymph flows

Lymph flow	Rate of lymph flow	Species	Disease/exercise	Reference
Thoracic duct lymph flow	1 mL/min	Humans	Healthy	Dumont et al. 119
Thoracic duct lymph flow	3–6-fold higher than healthy	Humans	Cirrhosis	Dumont et al. 119
Thoracic duct lymph flow	1.5 mL/min	Pigs	Normal	Malbrain et al.120
Thoracic duct lymph flow	2-fold than healthy	Pigs	IV LPS administration	Malbrain et al.120
Nodal lymph flow	1.49 mL/gm/min	Rats	Normal	DSouza et al.121
Nodal lymph flow	0.5 mL/gm/min	Rats	Cancer cell injected lymph nodes	DSouza et al.121
Pre-nodal lymph flow	13.2 mL/24 h	Humans	Healthy	Olszewski et al.103
Pre-nodal lymph flow	22.6 mL/24 h	Humans	Rheumatoid arthritis patients	Olszewski et al.103
Thoracic duct lymph flow	1.7 mL/min	Dogs	Normal	Desai et al.122
Thoracic duct lymph flow	3.9 mL/min	Dogs	1.5 miles/h on treadmill	Desai et al.122
Thoracic duct lymph flow	9 mL/min	Dogs	10 miles/h on treadmill	Desai et al.122

4.4. Physiologically based pharmacokinetic (PBPK) model for TPs after ID or SC administration

PBPK models are useful for PK prediction and estimation of the first-in-human dose for new chemical entities (TPs and SMDs). The in vitro clearance data (e.g. liver microsomes) is used extensively in the PBPK models of SMDs⁶⁹, yet these models are less often applied to TPs. In vitro proteolysis data generated using ID or SC tissue and lymphatic system tissues may be useful to develop PBPK models for new TP entities. Giragossian et al. used the serum in vitro stability data to develop a PBPK model for a Fibroblast Growth Factor 21–antibody conjugate to predict PK.¹¹⁴ In a recent study by our group, a minimal PBPK model (Figure 6) was used to understand mAb transport in the lymphatic system and to identify physiological and physicochemical factors governing the absorption after SC administration.⁸⁴ The model described a step-by-step transfer of TPs from the SC interstitial space towards the systemic circulation via lymphatic capillaries, lymphatic trunks, LNs and thoracic duct. We hypothesized that, the TPs are proteolyzed during the lymphatic transit, especially in the LNs. Unfortunately, the in vitro data to describe rate of clearance from the LNs was not available, hence a top-down approach was used with parameter estimation using clinical SC PK data.

Figure 6:

Schematic representation of the minimal PBPK model of mAb after SC administration (Adapted with permission from Varkhede and Forrest⁸⁴)

Key points and unknowns

The physiological parameters related to the lymphatic system are important for transport of TPs from the ID or SC injection site towards the systemic circulation. These parameters may be altered due to the disease conditions. In addition, the physiological parameters are different across various species, which can lead to differences in the PK and bioavailability. As stated in section 3, the in vitro systems representing the ID, SC and lymphatic tissues can be utilized to generate quantitative proteolysis information. The LN cell suspension may be an appropriate system to account for the proteolysis in the LNs. Finally, the PBPK model can be used with the physiological parameters and in vitro proteolysis data to predict PK of the TPs.

5. Mechanisms and case studies of biological oxidation of proteins

Biological oxidation of proteins remains an important focal point among biomedical scientists, as elevated levels of oxidized proteins are associated with aging as well as several pathologies such as, e.g., Alzheimer’s disease, Parkinson disease, diabetes, rheumatoid arthritis and muscular dystrophy.^123–126 Protein oxidation comprises covalent modifications resultant of reactions with reactive oxygen species (ROS) or reactions with secondary by-products of oxidative stress.^127,128 Due to their abundance in vivo and high reactivity with ROS, proteins are one of the major targets for ROS.^125–127 Protein oxidation can result in aggregation, an increase in side-chain hydrophilicity, side-chain and backbone fragmentation, protein unfolding and change in conformation.^125,126 The sources, mechanisms and case studies of in vivo oxidation reactions are discussed in the sections below.

5.1. Nature and properties of ROS

ROS are responsible for oxidation of various biomolecules such as proteins, DNA, and lipids. The term ROS includes both oxygen radicals and certain non-radicals that can be readily converted into radicals.^127,128 All oxygen radicals are ROS, but not all ROS are oxygen radicals. The group of radicals includes superoxide anion (O₂^•-), nitric oxide (NO^•), hydroxyl radical (OH^•), as well as peroxyl (ROO^•) and alkoxyl (RO^•) radicals. Some of the non-radical species comprise hydrogen peroxide (H₂O₂), hypochlorous acid (HClO), singlet oxygen (¹O₂), and organic peroxides.^126–128 A radical is any species which contains an unpaired electron. Radicals can be formed through various mechanisms, such as one-electron oxidation or reduction reactions or through the homolytic fission of a covalent bond.^129,130 A diverse range of radicals are generated in living systems, which differ markedly in their reactivity. The properties and main reactions of some of the biologically relevant ROS are discussed below.

OH^• radicals can be generated in vivo through the metal-catalyzed degradation of H₂O₂, ionizing-radiation, proton-catalyzed decomposition of peroxynitrite, and the decomposition of ozone.^131,132 For example, the reaction of hydrogen peroxide with ferrous ion (Fe^II) can yield OH^• and hydroxide anion (OH⁻) in the classical Fenton reaction (Scheme 1, reaction 1).^133–135 In reality, this reaction is complex, where either OH^• can dissociate from the metal complex or remain bound to the metal complex and react as “complexed hydroxyl radical” or generate higher oxidation states of iron^22,135,136; all these species are reactive, but differ in reactivity and redox properties.^{20,133,134,137,138} Noteworthy, the nature of metal ligands plays an important role in the generation of reactive species from Fenton or Fenton-like reactions, where oxidizing species generated at the central metal ion can directly react with the metal ligands.^136,139,140

Scheme 1:

Generation of ROS in vivo

The OH^• radical is the strongest oxidant which may be generated in vivo, and, hence, reacts rather non-selectively with biological targets in its vicinity.^{20,132,134,141} Due to the high reactivity of HO^• radicals and the high density of biological targets in vivo, OH^• radical-mediated oxidation reactions are often site-specific.¹³² OH^• radicals react through hydrogen atom abstraction, addition and electron transfer reactions. The hydrogen abstraction reactions result in the production of H₂O and carbon-centered radicals (R^⋅) (Scheme 1, reaction 2), which, in presence of oxygen, generate ROO^• (Scheme 1, reaction 3).^19,20,130 In the absence of oxygen, carbon-centered radicals can form covalent cross-links (Scheme 1, reaction 4). A carbon-carbon cross-link frequently detected during protein oxidation is dityrosine, though this cross-link originates from tyrosyl radicals which do not efficiently react with oxygen.¹³⁰

Peroxyl radicals, ROO^•, are moderately strong oxidants (depending on the substituents on R), which are generated by the addition of oxygen to carbon-centered radicals, R^•. ROO^• can also be generated in the absence of oxygen from reaction of transition metal ions with hydroperoxides.^125,130,132 The reaction of two ROO^• generates intermediary tetroxide species, which can decompose via various mechanisms. Scheme 1, reaction 6, shows the decomposition of one tetroxide into molecular oxygen and two alkoxy radicals, RO^• Alkoxyl radicals are usually more reactive than peroxyl radicals, i.e. in hydrogen abstraction reactions, but can also further react via unimolecular processes such as β-cleavage or 1,2-hydrogen-shift; both RO^• and ROO^• react with biomolecules.¹³⁰ The chemical reactivity of ROO^• and RO^• is determined by their α-substituents.^130,142,143

The reaction of O₂^•- with NO^• yields the strong oxidant peroxynitrite, ONOO⁻ (Scheme 1, reaction 8), which can react with wide array of biological entities, including DNA and proteins.^144,145 Protonation of ONOO⁻ forms peroxynitrous acid, which can undergo homolytic cleavage to form OH^• and nitrogen dioxide (Scheme 1, reaction 9b).^130,144,145

Biologically significant non-radical oxidants include H₂O₂, ¹O₂ and HOCl.^19,20,137 H₂O₂ is generated in vivo through chemical and enzymatic reactions catalyzed by several oxidases, such as monoamine oxidase (MAO), NADPH oxidase (NOX),^146,147 xanthine oxidase (XO),¹⁴⁸ and though the dismutation of O₂^•- in presence of superoxide dismutase (SOD) (Figure 7).¹⁴⁹ The direct reaction of H₂O₂ with most biomolecules is slow (an exception is the reaction with peroxiredoxins), but Fenton or Fenton-like reactions can convert H₂O₂ into stronger oxidants.^{133,134,137,141} HOCl is a moderately strong oxidant generated by the reaction of H₂O₂ with chloride (Cl⁻), catalyzed by myeloperoxidase (MPO).¹⁵⁰ HOCl is rather reactive towards biomolecules containing amines and thiols, yielding chloramines (RNHCl) and sulfenyl chloride (RSCl), respectively. The latter are unstable and convert to additional products.^150,151

Figure 7:

Role of NADPH oxidase in generation of ROS

With a variety of ROS produced in vivo, proteins can be oxidized to a manifold of products (Table 6).^{18,125,126,130,137} The generation of protein carbonyls is irreversible and may lead to loss of protein activity and/or the formation of aggregates, which are potentially immunogenic.^18,137,152 Protein carbonyls can be formed in vivo through direct oxidation of mainly Lys, Arg, Thr, Pro, and Glu or through the oxidative cleavage of the protein backbone.¹⁸ Among all the amino acids, sulfur-containing amino acids such as methionine (Met) and cysteine (Cys) are prone to both reversible and irreversible oxidative modification.^18,153–155 Frequently, Met is converted to methionine sulfoxide (MetSO) and Cys is oxidized to form the disulfide cystine.^18,155 Both oxidation reactions can be reversed through enzymes such as MetSO reductases and thioredoxin dependent enzymes, respectively.^153–155 However, oxidation of Met to methionine sulfone (MetSO₂) and Cys to cysteic acid (CysSO₃H) are generally not reversible.

Table 6:

Various oxidative modifications of proteins

Amino acid	Oxidation products
Methionine	Methionine sulfoxide, methionine sulfone
Tryptophan	Kynurenine, N-formyl kynurenine, hydroxy tryptophan
Tyrosine	3,4 dihydroxyphenylalanine (DOPA), 3-nitrotyrsoine, 3-chlorotyrosine
Phenylalanine	o-tyrosine, m-tyrosine
Histidine	2-oxo-histidine
Cysteine	Oxidation of the sulfhydryl group to form sulphenic, sulphenic or sulphenic acids. Formation of disulfide bond in a reduced protein or formation of a thioether linkage
Lysine, arginine, threonine, proline, leucine, isoleucine, valine	Carbonyl products of respective amino acids such as α-aminoadipic semialdehyde, glutamic semialdehyde, 2-pyrrolidone and 2-amino 3-ketobutyric acid from Lys, Arg, Pro and Thr amino acids, respectively

Typically, the oxidation of Met to MetSO is closely monitored in therapeutic proteins.^18,156 In mAbs, primarily Met252 and Met428 in the Fc domain are monitored, as oxidation of these Met residues can affect the serum-half-life of these proteins.^18,137 This topic is more extensively discussed in the next section.

5.2. Biological oxidation of protein therapeutics

Given the technical challenges and limited sample availability, the in vivo characterization of the chemical integrity of protein therapeutics is not performed regularly. There are only a few studies that have reported on the in vivo oxidation of protein therapeutics. Yang et al have designed a study to determine if results from an in vitro model study can be used to estimate the results from in vivo model results.¹⁵⁷ In this study, they have compared the rates of in vivo chemical degradation in mAb1 extracted from clinical serum samples to the rates of in vitro chemical degradation mAb1 extracted from spiked PBS and human serum samples. They observed oxidation of various methionines; Met252 in the Fc region and Met40 in the Fab region was monitored for 6 weeks using peptide mapping and intact mass analysis. There was no increase in Met252 oxidation in the mAb1 extracted from clinical serum samples, which is consistent with previous studies conducted in both rat and monkey models. However, the oxidation of Met40 in the Fab region increased from 1.9% to 6.2% in clinical serum samples and from 4.0% to 12.3% in the spiked serum samples, but there was no further increase in PBS spiked samples over 6 weeks.¹⁵⁷ The reason for this increase in Met40 oxidation exclusively in clinical serum and serum spiked samples could be attributed to ROS and other cellular oxidants present in the serum. Similar findings were reported by Battersby et al. where recombinant human growth hormone (rhGH) injected into a rat model exhibited methionine (Met14) oxidation. In addition, they also reported a trend in increase of Met14 oxidation with increase in incubation time.¹⁵⁸ In another interesting study, Yao et al investigated the global biotransformation of a therapeutic protein using LC-MS differential analysis.¹⁵⁹ In this study a mAb was force-oxidized with 0.1% of tertiary-butyl hydroperoxide (tBTH) at 25°C overnight. The force-oxidized and control mAb were spiked with rat serum to generate a biotransformation sample and a control sample, respectively. Later both sample sets were digested for LC-MS analysis and the generated data sets were processed using bioinformatic tools for differential analysis. On analysis, Met254 and Met430 were found to be oxidized in samples incubated with rat serum, but not in the control sample.¹⁵⁹

Cys is a redox sensitive amino acid¹⁵⁵, prone to the generation of a variety of post-translational modifications such as sulfenic, sulfinic, and sulfonic acid, persulfide, thioether and symmetric and asymmetric disulfides.^155,160,161 Under basic conditions, disulfides can degrade through β-elimination forming dehydroalanine and persulfide (RSSH), which can be reduced back to Cys. The resulting Cys and dehydroalanine residues can cross-link to form a thioether.^160,161 Previously, Zhang et al measured thioether levels in endogenous antibodies to determine if thioether formation is naturally happening in vivo.¹⁶² They identified a thioether cross-link between the original residues Cys214 of the LC and Cys220 of HC in both IgG1λ and IgG1k (endogenous antibodies). The average thioether levels in IgG1k and IgG1λ were 5.2% and 11.0%, respectively. To confirm the formation of thioether linkages in vivo, their levels were monitored over time after administering a single dose of IV injection of a therapeutic IgG1 in humans (2 patients). In patient 1, thioether content increased from <1% to ~ 2.7% and in patient 2 it increased from 0.5% to 2.1% within 14 days, with a formation rate of 0.1% per day.¹⁶²

As outlined in this section, there is only a limited number of studies which have monitored the chemical integrity of therapeutic proteins in vivo. The most commonly measured products of protein oxidation in biological samples are MetSO, dihydroxyphenylalanine (DOPA, oxidation product of tyrosine) and carbonyl derivates of Pro, Arg, Lys and Thr.^{18,126,130,163} Generally, liquid chromatography-tandem mass spectrometry (LC-MS/MS) in combination with bioinformatic tools allows for the identification of oxidative protein modifications. A specific improvement for the resolution and detection of these modifications is the use of very long columns (≥ 1m), coupled to mass spectrometry detection.^164,165

6. Impact of oxidation on pharmacokinetics of TPs

TPs are cleared from the circulation through multiple pathways: target-mediated clearance, degradation by proteolysis, nonspecific endocytosis, Fcγ receptor mediated clearance, and formation of immune complexes (ICs) followed by Fc receptor mediated clearance mechanisms.^166,167 The rate of elimination is determined by molecular weight, along with physico-chemical properties such as lipophilicity, functional groups, glycosylation pattern and conformational structure.^167,168 The dependence of clearance on physico-chemical properties of TPs can be demonstrated by comparison of regular human insulin with rapid acting insulins (lispro and aspart).^169–171 The insulin analogues have an onset of action between 5–15 mins, and are effective up to 6 h when compared to regular human insulin with a later onset of 30–60 min, with an effective duration of 8–10 hrs.^169,172 The differences in clearance of these sets of insulins can be attributed to structural differences between regular human insulin and insulin analogues through amino acid substitution in chain B, resulting in conformational changes that result in changes in binding to the c-terminal region.¹⁷²

The rate of elimination can also be affected by oxidation of specific amino acids in a protein.^173,174 Previously, Iwao et al have investigated the effect of oxidation on in vivo elimination of human serum albumin (HSA). In this study, the oxidized HSA was produced through a MCO system and the pharmacokinetics were evaluated in mice.¹⁷⁴ Iwao et al. observed a gradual increase in elimination with increase in carbonyl content in HSA. The liver clearance of HSA was closely related to the hydrophilicity and net charge of the proteins. Upon oxidation, they observed a slight decrease in α-helical content which was accompanied by an increase in accessibility of the hydrophobic areas and an increase in net negative charge on the HSA molecule which led to an increase in elimination of oxidized HAS, resulting in its low serum half-life.^173,174

Given their large size and higher molecular weight, mAbs have some unique properties which are determined by multiple factors related to antibody’s structure and functions including glycosylation pattern, FcRn mediated recycling, isoelectric point (pI), off-target binding and anti-drug antibody response.^{88,166,175,176} In this section, PK properties and the effect of oxidation on the PK of mAbs will be discussed.

6.1. Impact of methionine oxidation in human IgG1 Fc on FcRn binding

The neonatal FcRn receptor is responsible for the phenomenal long half-lives (20–21 days) of the three IgG (1,2 and 4) subclasses through a protective mechanism against metabolism and elimination of IgG’s.^{88,166,176,177} FcRn is a heterodimeric receptor composed of two polypeptides, a 48–52 kDa class I histocompatibility complex-like protein (α-FcRn) and a 14 kDa β2- microglobulin.^176–179 The interaction between FcRn and IgG occurs strictly in a pH-dependent manner in acidic pH conditions (pH 6.0–6.5). Like all circulating proteins, IgG’s undergo non-specific pinocytosis by vascular endothelial cells and bone marrow-derived cells.^{176,177,179,180} After pinocytosis, FcRn binds to the Fc domain at the CH2-CH3 portions with high affinity at acidic endosomal pH, which salvages them from lysosomal degradation. The resultant IgG-FcRn complex is then transported back to the cell surface and the subsequent release into the blood stream upon exposure of the IgG-FcRn complex to the neutral extracellular pH environment, whereas the unbound IgG is digested to individual amino acids by intracellular lysosomes.^181–183 Based on the in vitro binding studies some hypothesize that the FcRn-IgG interaction occurs in 2:1 ratio, with two receptors binding to the two heavy chains of an IgG.¹⁸⁴

The IgG sequence in the CH2-CH3 interface involved in the FcRn binding contains two conserved methionines located at positions 252 in the CH2 domain and 428 in the CH3 domain (EU numbering).^185,186 Both Met252 and Met428 residues are surface-exposed, which are structurally close to the FcRn binding interface. The binding affinity of the IgGs for FcRn can be affected by post-translational modifications of the two conserved (252 and 428) methionines in the Fc domain.^185–188 One of the most widely observed post-translational modification in IgG is oxidation of methionine to methionine sulfoxide at these two (252 and 428) positions. ^185–190

6.2. Impact of methionine oxidation on pharmacokinetics properties of therapeutic antibodies

The way oxidation of Met in the CH2-CH3 interface affects the binding affinity to the FcRn receptor and consequently, the PK of mAbs is a topic of interest and is being extensively studied. Various reports have shown that oxidation of Met in the CH2-CH3 interface caused decrease in binding affinity to the FcRn receptor, but did not affect the dissociation rate.^185–188 The molecular mechanism causing decrease in binding affinity had been hypothesized to be due the conformational changes as the result of oxidation in CH2-CH3 interface.^190–192 Multiple studies using hydrogen/deuterium exchange mass spectrometry have indicated that oxidation of Met 252 induced subtle conformational changes in the FcRn binding region, which is in the CH2 domain covering residues 243–247.^191–193 Through molecular modeling studies Wang et al.,¹⁸⁷ have suggested that the resultant conformational changes undermined the hydrophobic interactions between oxidized Met 252 and Pro 134 of FcRn and induced repulsion between oxidized Met 252 and Glu 135 leading to the decrease in binding affinity. Unlike Met 252, oxidized Met 428 was hypothesized to indirectly affected the binding affinity by disturbing the hydrophobic core present between CH2-CH3 domains.¹⁹⁴

As FcRn is known to prolong the serum half-life of IgG, decrease in binding affinity of IgG-FcRn interaction can decrease the antibody serum half-life.^181–183 Utilizing human FcRn transgenic mice model system, Wang et al.,¹⁹⁰ have demonstrated that extensive Met oxidation significantly decreased serum half-lives of IgG, whereas moderate levels of Met oxidation had minimal effect on serum half-lives. For example, IgG containing 40% oxidized Met 252 had a half-life like the non-oxidized sample, while IgG containing 80% oxidized Met 252 showed a 4-fold decrease in the half-life. Interestingly Stracke et al.,¹⁸⁸ have identified that IgG sample with only Met 428 oxidation showed increased affinity to the FcRn receptor and remarkably slower plasma clearance compared to non-oxidized sample. However, in the same report they have also suggested that Met 428 oxidation decreased FcRn binding and subsequent decrease in serum half- life only if oxidized Met 252 co-exist in the same heavy chain.¹⁸⁸

7. Conclusion

This review summarizes the current knowledge of physiological factors that can control ID, SC, and lymphatic transport, proteolysis and oxidation of TPs. There exists a need to investigate the proteolysis of TPs using relevant in vitro systems to generate quantitative information, which can be used to build hypothesis-driven mechanistic PBPK models for TPs after ID or SC administration and to aid the prediction of bioavailability. Finally, mechanisms of the biological oxidation of TPs by ROS is described, which can alter their FcRn binding and overall PK.