*Article* **Specificity of Anti-Citrullinated Protein Antibodies to Citrullinated** α**-Enolase Peptides as a Function of Epitope Structure and Composition**

**Ilaria Fanelli 1, Paolo Rovero 1, Paul Robert Hansen 2, Jette Frederiksen 3, Gunnar Houen 3,4 and Nicole Hartwig Trier 3,\***


**Abstract:** Rheumatoid arthritis (RA) is an autoimmune disease affecting approximately 1–2% of the world population. In addition to the first discovered serologic markers for RA, the rheumatoid factors (RFs), anti-citrullinated protein antibodies (ACPAs) are even more specific for the disease compared to RFs and are found in 70–80% of RA patient sera. RA etiopathogenesis still needs to be elucidated, as different factors are proposed to be involved, such as Epstein–Barr virus infection. Hence, understanding the interaction between ACPAs and their citrullinated peptide targets is relevant for a better knowledge of RA pathophysiology and for diagnostic purposes. In this study, a cohort of RA sera, healthy control sera and multiple sclerosis sera were screened for reactivity to a variety of citrullinated peptides originating from α-enolase, pro-filaggrin, proteoglycan and Epstein–Barr nuclear antigen-2 by enzyme-linked immunosorbent assay. ACPA reactivity to citrullinated α-enolase peptides was found to depend on peptide length and peptide conformation, favouring cyclic (disulfide bond) conformations for long peptides and linear peptides for truncated ones. Additional investigations about the optimal peptide conformation for ACPA detection, employing pro-filaggrin and EBNA-2 peptides, confirmed these findings, indicating a positive effect of cyclization of longer peptides of approximately 20 amino acids. Moreover, screening of the citrullinated peptides confirmed that ACPAs can be divided into two groups based on their reactivity. Approximately 90% of RA sera recognize several peptide targets, being defined as cross-reactive or overlapping reactivities, and whose reactivity to the citrullinated peptide is considered primarily to be backbone-dependent. In contrast, approximately 10% recognize a single target and are defined as nonoverlapping, primarily depending on the specific amino acid side-chains in the epitope for a stable interaction. Collectively, this study contributed to characterize epitope composition and structure for optimal ACPA reactivity and to obtain further knowledge about the cross-reactive nature of ACPAs.

**Keywords:** anti-citrullinated protein antibodies; citrullinated peptides; epitopes; rheumatoid arthritis

#### **1. Introduction**

RA is a systemic and chronic autoimmune disease with a worldwide prevalence of approximately 5 per 1000 adults, affecting women two to three times more often than men. RA disease onset may occur at any age; however, the peak incidence is in the sixth decade [1–3]. RA is characterized by infiltration of monocytes, B cells and T cells in the synovial membrane in joints and ultimately cartilage degradation and erosion of the underlying bone [2]. In addition to joint damage, some systemic features are associated

**Citation:** Fanelli, I.; Rovero, P.; Hansen, P.R.; Frederiksen, J.; Houen, G.; Trier, N.H. Specificity of Anti-Citrullinated Protein Antibodies to Citrullinated α-Enolase Peptides as a Function of Epitope Structure and Composition. *Antibodies* **2021**, *10*, 27. https://doi.org/10.3390/antib10030027

Academic Editor: Ohad Mazor

Received: 22 May 2021 Accepted: 14 July 2021 Published: 21 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

with RA, for instance pulmonary, cardiovascular, psychological, and skeletal disorders [4,5]. Hence, RA has increased morbidity and mortality rates when left untreated [1,6].

RA is diagnosed according to EULAR/ACR classification criteria revised in 2010 which, along with clinical disease manifestations, comprise serological biomarkers such as anti-citrullinated protein antibodies (ACPA) [7]. ACPAs are detected in 70–80% of RA patients and in approximately 1–2% of the healthy population. Moreover, they have been reported to be present up to 14 years before the manifestation of clinical symptoms, making ACPAs good biomarkers for RA [8,9]. Ultimately, it has been reported that ACPApositive RA patients experience increased joint damage and low remission rates, indicating that these individuals have more severe disease courses compared to ACPA-negative RA patients [10]. The occurrence of ACPA-positive RA is related to genetic risk factors that predispose for RA, for instance, protein tyrosine phosphatase nonreceptor type-22 (PTPN22) and MHC class II alleles [11–14].

ACPAs recognize the nonstandard amino acid citrulline, a nongenetically encoded amino acid. Citrullination is the result of a post-translational modification, where the positively charged guanidino group of Arg is substituted by the neutral ureido group. Ultimately, this modification may lead to structural unfolding of the citrullinated protein [15,16]. Citrullination is catalyzed by Peptidyl Arginine Deiminase (PAD) enzymes, which are calcium-dependent metalloenzymes [17].

ACPAs are typically detected with assays, which exploit enzyme-linked immunosorbent assay (ELISA) methods and synthetic citrullinated peptides [18–22]. The first generation of assays was based on a synthetic linear citrullinated peptide derived from human filaggrin [19]. In order to improve assay sensitivity, the linear peptide was replaced by a cyclic version (Cyclic Citrullinated Peptide, CCP, containing a disulfide bond), as the cyclic peptide yielded higher sensitivity and specificity compared to the linear version. This assay is also referred to as CCP1 [20]. Screening of peptide libraries has led to the selection of other antigens and generation of second and third generations of CCP assays [18]. While the previously mentioned assays only detect IgG ACPAs, the CCP3.1 detects both IgG and IgA isotypes. Despite this, the golden standard for ACPAs detection is the CCP2 assay [18,22].

In the commercial ACPA assays, different citrullinated peptides are employed, which is in accordance with the cross-reactive nature of ACPAs. ACPAs are able to recognize several citrullinated targets, preferably containing a Cit–Gly motif [19,21,23–27]. Besides a critical Cit–Gly motif, charged amino acids in the C-terminal have been proposed to be essential for a stable interaction between ACPAs and citrullinated peptide targets [23]. The amino acids surrounding citrulline have been analyzed in several studies, which revealed that substitutions in positions -x-x-Cit-Gly-x- do not influence antibody reactivity. This finding demonstrates the crucial role of the Cit–Gly motif for a stable antibody–antigen interaction, even though sometimes other amino acids besides Gly are also tolerated [23,24,26–28]. Examples of ACPA targets that have been reported are collagen, fibrinogen, α-enolase, vimentin, pro-filaggrin, Epstein–Barr nuclear antigen (EBNA)-1, and EBNA-2 [21,22,25].

It has been proposed that ACPAs can be divided into two groups, based on their ability to interact with citrullinated peptides [16,17], one group that appears to recognize a large variety of citrullinated targets and another group that recognizes a very limited number of citrullinated peptides [29]. The first group of ACPAs, also referred to as "overlapping" or "cross-reactive" antibodies, is primarily backbone-dependent, whereas the second group, also referred to as "nonoverlapping" or "epitope-specific", depends on the specific amino acid side-chains of the epitope to establish a stable antibody–antigen interaction [16].

On this basis, we analyzed the interactions between citrullinated targets and ACPAs in order to obtain further knowledge about ACPAs, which is important for the improvement of diagnostic tools, and to elucidate their role in the pathogenesis of RA. Citrullinated α-enolase peptides were used as a point of origin to characterize epitope composition and structure for optimal antibody reactivity and the overlapping and nonoverlapping ACPA reactivities.

#### **2. Materials and Methods**

#### *2.1. Reagents*

Alkaline phosphatase (AP)-conjugated goat-anti-human IgG, streptavidin and AP substrate tablets (*para*-nitrophenylphosphate (*p*NPP)) were from Sigma Aldrich (St. Louis, MO, USA). Tris-Tween-NaCl (TTN, 0.3 M NaCl, 20 mM Tris, 0.01% Tween 20, pH 7.5), carbonate buffer (0.05 M sodium carbonate, pH 9.6) and AP substrate buffer (1 M ethanolamine, 0.5 mM MgCl2, pH 9.8), were from Statens Serum Institut (Copenhagen, Denmark). Synthetic peptides purchased were from Schäfer-N (Lyngby, Denmark) (Table 1) and were generated on TentaGel resin using standard Fmoc-based solid-phase peptide synthesis. The peptides were synthesized as peptide acids.


**Table 1.** Synthetic peptides tested for antibody reactivity. "B" represents biotin.

#### *2.2. Patient Material*

RA serum samples (*n* = 28) and healthy donor (*n* = 28) serum samples (referred to as healthy control (HC)) were provided by Statens Serum Institut Biobank (Copenhagen, Denmark) (*n* = 28), which routinely analyzes patient sera for diagnostic purposes. The samples were tested anonymously, therefore not requiring ethical consent.

Ten multiple sclerosis serum samples from the Multiple Sclerosis Clinic, Department of Neurology, Rigshospitalet Glostrup (Glostrup, Denmark) were used as disease controls. The samples were tested anonymously, therefore not requiring ethical consent.

#### *2.3. Detection of Antibodies by Enzyme-Linked Immunosorbent Assay and Streptavidin-Capture Enzyme-Linked Immunosorbent Assay*

Microtiter plates were coated with 1 μg/mL free peptide in carbonate buffer and incubated overnight at room temperature (RT) on a shaking table (ST). The wells were rinsed with TTN for 3 × 1 min and blocked with TTN for 30 min. Sera were diluted (1:200) in TTN, added to each well, and then incubated for 1 h (h) at RT on a ST. After washing with TTN buffer, AP-conjugated goat-anti-human IgG diluted in TTN (1:1000) was added to each well and incubated for 1h at RT on a ST. Finally, *p*NPP-containing AP substrate buffer (1 mg/mL) was added to each well and AP activity was determined by measuring the absorbance at 405 nm with background subtraction at 650 nm.

Alternatively, microtiter plates were precoated with 1 μg/mL streptavidin in carbonate buffer and incubated overnight at 4 ◦C. Biotinylated peptides (diluted to 1 μg/mL in carbonate buffer) were added to each well and incubated for 2h at RT on a ST. The following steps in the experiment were carried out as mentioned above. All samples were tested in duplicates.

Based on preliminary screening, absorbances of all the results were normalized to a positive RA control pool (*n* = 28) and a peptide-specific cutoff was introduced, tolerating a nonspecific reactivity of 5% and an intra-assay variation of 15%. Readings above the cutoff were regarded as positive, whereas samples below the cutoff were regarded as being negative. Inter-assay variations below 15% were acceptable.

#### *2.4. Statistics*

Statistical analyses and plots were generated using GraphPad Prism 9.0 software. The values obtained in the experiments were compared further by using Student's *t*-test.

#### **3. Results**

#### *3.1. Reactivity of Rheumatoid Arthritis Sera to α-Enolase Peptides*

Various citrullinated protein targets have been identified in RA, such as α-enolase, pro-filaggrin, proteoglycan, and fibronectin [30,31]. In order to further characterize the reactivity of ACPA to citrullinated peptides, RA patient sera (*n* = 28), HC sera (*n* = 28) and MS sera (*n* = 10) were tested for reactivity to a citrullinated α-enolase peptide (KIHARCEIFDS-Cit-GNPTVEC) by ELISA.

As seen in Figure 1, elevated antibody reactivity was found to the citrullinated peptides compared to the Arg-containing control peptide (*p* = 0.0073 for the cyclic and *p* = 0.0003 for the linear). No significant difference in antibody reactivity was found between the cyclic and the linear α-enolase peptides (*p* = 0.2647). Approximately 40% of the RA sera recognized the α-enolase peptides, and reacted significantly to the linear and the cyclic peptide compared to the control peptide (*p* < 0.0001). None of HC sera or MS sera reacted to the citrullinated α-enolase peptides, confirming that ACPA reactivity to the α-enolase peptides was specific for RA.

**Figure 1.** Reactivity of rheumatoid arthritis (RA), healthy control (HC) and multiple sclerosis (MS) sera to α-enolase peptides (KIHARCEIFDS-Cit-GNPTVEC) analysed by traditional ELISA. A linear Arg-containing peptide (KIHARCEIFDS-R-GNPTVEC) was used as negative control. HC and MS sera were used as controls. A.U. were defined as absorbances normalized relative to a positive RA control pool. *p* values less than 0.001 are shown as \*\*\*.

#### *3.2. Reactivity of Rheumatoid Arthritis Sera to Truncated Linear and Cyclic α-Enolase Peptides*

Previous studies describing RA sera reactivity to citrullinated pro-filaggrin peptides indicated that ACPA reactivity is dependent on peptide length and conformation [23]. To determine whether this relates to α-enolase peptides as well, RA sera that were positive for reactivity to α-enolase in the preliminary screening (*n* = 12) were tested for reactivity to cyclic and linear truncated α-enolase peptides by ELISA.

As seen in Figure 2, RA sera recognized all of the linear peptides independent of their length. Sensitivities of approximately 90% were found for all of the linear peptides. In contrast, specific ACPA reactivity was primarily found to the cyclic peptides C-19-Cit and C-10-Cit, obtaining sensitivities of approximately 90% as well. These findings indicate that the peptide conformation affects the antibody reactivity and that it is essential for peptide presentation. Additionally, no specific reactivity was found when screening HC sera. These findings are in accordance with the literature, describing that ACPA reactivity to 19-mer linear and cyclic pro-filaggrin peptides yields similar reactivity, whereas ACPA reactivity to the linear citrullinated peptides is favoured for smaller peptides (<12 amino acids), when compared to the cyclic peptides [23].

**Figure 2.** Reactivity of rheumatoid arthritis (RA) sera and healthy control (HC) sera samples to truncated linear and cyclic citrullinated α-enolase peptides analysed in ELISA. Peptides are plotted from left to right with decreasing length. An Arg-containing peptide was used as control (KIHAR-CEIFDSRGNPTVEC). (**a**) Reactivity of RA sera to linear peptides. (**b**) Reactivity of HC sera to linear peptides. (**c**) Reactivity of RA sera to cyclic peptides. (**d**) Reactivity of HC sera to cyclic peptides. A.U. were defined as absorbances normalized relative to a positive RA control pool.

#### *3.3. Overlapping Reactivities of Anti-Citrullinated Protein Antibody Responses*

As presented, approximately 40% of the RA sera reacted with the α-enolase peptides. To determine whether the ACPA reactivities were specific for the α-enolase peptide, all RA sera (*n* = 28) and HC sera (*n* = 28) were tested for reactivity to various citrullinated peptides in ELISA. Peptides from pro-filaggrin, proteoglycan, fibronectin, and EBNA-2 were tested for antibody reactivity.

As shown in Figure 3, significant RA antibody reactivity was found to the citrullinated pro-filaggrin (*p* = 0.0119), proteoglycan (*p* = 0.0004), fibronectin (*p* = 0.0459) and EBNA-2 (*p* < 0.0001) peptides compared to the HC. The proteoglycan and EBNA-2 peptides obtained the highest sensitivities of 50% and 68%, respectively. One HC serum showed low reactivity to EBNA-2 L (Figure 3b).

**Figure 3.** Reactivity of rheumatoid arthritis (RA) RA and healthy control (HC) sera to a selected peptide panel. (**a**) Screening of RA samples (*n* = 28) to citrullinated peptides from pro-filaggrin, proteoglycan, fibronectin and EBNA-2. An Arg-containing pro-filaggrin peptide was used as control. (**b**) Screening of HC sera (*n* = 28) to citrullinated peptides from pro-filaggrin, proteoglycan, fibronectin and EBNA2. An Arg-containing pro-filaggrin peptide was used as control. A.U. were defined as absorbances normalized relative to a positive RA control pool.

Thorough analysis of the reactivities of the RA sera showed that approximately 14% of the RA sera samples recognized all 4 peptides, whereas 18% reacted to 3 peptides. Note that approximately 54% of the samples recognized 1 or 2 peptides, whereas 14% of the RA cohort did not show reactivity to any of the peptides (Table 2, first column).


**Table 2.** Reactivities of the RA cohort to citrullinated peptides originating from pro-filaggrin, proteoglycan, fibronectin and EBNA-2.

Of the 9 RA sera that only reacted with 1 peptide (Table 2, first column), 22% reacted with the proteoglycan peptide (*n* = 2), 22% with the fibronectin peptide (*n* = 2) and 56% (*n* = 5) with EBNA-2 peptide.

When dividing the complete RA cohort into α-enolase-positive and -negative sera, it was observed that the RA samples in the α-enolase-positive cohort were prone to have a higher degree of overlapping antibody reactivities compared to the α-enolase-negative cohort. For instance, 50% of the RA sera in the α-enolase-positive cohort recognized 3 or 4 peptides compared to 19% in the α-enolase-negative cohort. Similarly, 42% of the α-enolase-positive cohort recognized 1 or 2 peptides compared to 63% for the α-enolasenegative cohort.

When examining the reactivity of the 3 cohorts (complete cohort, α-enolase-positive cohort, α-enolase-negative cohort) to the three most reactive peptides, pro-filaggrin, proteoglycan, and EBNA-2 L, similar results were obtained. As presented in Figure 4, significant overlapping reactivities were found between the 3 cohorts.

**Figure 4.** Venn diagram illustrating overlapping anti-citrullinated protein antibody reactivities to EBNA-2, pro-filaggrin and proteoglycan. (**a**) Reactivities of the complete rheumatoid arthritis (RA) cohort to the selected peptides (*n* = 28). (**b**) Reactivities of α-enolase-positive RA serum samples (*n* = 12). (**c**) Reactivities of α-enolase-negative RA serum samples (*n* = 16).

> Sera that only reacted to one peptide primarily recognized EBNA-2, as described earlier. Moreover, the most significant overlap in antibody reactivity was found between EBNA-2 and proteoglycan (Figure 4a,b), which is in accordance with the fact that these peptides obtained the highest sensitivities, as presented in Figure 3.

#### *3.4. Reactivity of RA Sera to Linear and Cyclic Peptide Versions. Is the Effect of Cyclization on Antibody Sensitivity General?*

Previous findings indicated that the cyclic version of the α-enolase peptide had a higher sensitivity, although not statistically significant when compared to the linear peptide. To determine whether this effect is general or peptide-specific, RA reactivity to linear and cyclic citrullinated pro-filaggrin and EBNA-2 peptides were tested in ELISA.

Figure 5 illustrates the reactivity of RA and HC sera to the linear and cyclic Profilaggrin and EBNA-2 peptide. ACPA reactivity to the citrullinated peptides was significantly elevated compared to the HCs (*p* = 0.0044 for EBNA-2 L, *p* < 0.0001 for EBNA-2 C, *p* = 0.0028 for pro-filaggrin L, *p* < 0.0001 for pro-filaggrin C).

**Figure 5.** Reactivity of rheumatoid arthritis (RA) samples and healthy control (HC) samples to pro-filaggrin (25 RA and 25 HC samples) and EBNA-2 (28 RA and 28 HC samples) peptides in their linear and cyclic conformation. (**a**) Reactivity of RA and HC sera to citrullinated cyclic and linear EBNA-2 peptides. (**b**) Reactivity of RA and HC sera to citrullinated cyclic and linear pro-filaggrin peptides. A.U. were defined as absorbances normalized relative to a positive RA control pool. \*\*\* *p* < 0.001, \*\* *p* < 0.01, \* *p* < 0.05.

Moreover, cyclization was observed to increase assay sensitivity, as the cyclic EBNA-2 and pro-filaggrin peptides had higher sensitivitities compared to the linear peptides (*p* = 0.0331), 67% and 44% of the RA sera reacted to the cyclic and linear EBNA-2 peptide, respectively, whereas 68% and 44% of the RA sera reacted to the cyclic and linear profilaggrin peptides, respectively. No statistically significant difference in antibody reactivity to the linear and cyclic pro-filaggrin peptides was found, although the antibody reactivity to the cyclic version was elevated (Figure 5b).

Collectively, these findings indicate that peptide conformation and peptide length affect antibody reactivity.

#### **4. Discussion**

In the present study, we analysed the reactivity of peptide-specific ACPAs to citrullinated epitopes and confirmed that factors such as peptide length and conformation notably influence antigen presentation.

The sensitivity of the full-length α-enolase peptide was determined to be approximately 44%, which is supported by earlier findings described in the literature [32]. ACPA reactivity to α-enolase was originally described by Lundberg et al., who showed that a cyclic peptide obtained the highest antibody reactivity among the human α-enolase and *Porphyromonas gingivalis* enolase peptides tested [32]. Similarly, no reactivity was found to the Arg-containing control peptide, confirming that the ACPAs are citrulline-specific (Figure 1). In addition to this, MS samples were tested for reactivity to the linear α-enolase peptide, as it has been described that citrullinated protein levels are elevated in MS [33], however, no reactivity was observed, confirming that ACPA reactivity is specific for RA.

The above mentioned experiments using α-enolase peptides were conducted in the absence of reducing agents, thus the linear peptides may in theory be able to cyclize during coating; however, as the effect of cyclization for α-enolase peptides was the same for pro-filaggrin and EBNA-2 peptides, where no cysteines were found in the linear peptides, we have reason to believe that the α-enolase peptides were found in a linear form during coating [23].

Screening of the truncated α-enolase peptides showed that all of the linear peptides (L-19-Cit, L-14-Cit, L-12-Cit and L-10-Cit) were roughly recognized to the same extent by the RA cohort, indicating that the length of the linear peptides is less important compared to the cyclic peptides. These findings confirm that the Cit–Gly motif in combination with a peptide backbone of approximately 10 amino acids, perhaps even shorter, is sufficient for antibody binding, as previously proposed [23]. In terms of reactivity, L-14-Cit, which

is the second-longest peptide, showed the highest reactivity. In a study conducted using pro-filaggrin peptides, it was found that a linear 21-mer peptide and a linear 14-mer peptide were significantly recognized by ACPAs. Moreover, the 14-mer peptide obtained the highest sensitivity, which conforms to this study, favouring peptides of approximately 10–14 amino acids [27]. This effect may relate to the peptide structure. Even though a small number of amino acids are flanking the Cit–Gly motif in the shorter peptides, the peptides are still able to fold and thus acquire a specific conformation that appears to bind to ACPAs more efficiently. The exact reason remains to be determined.

Increased sensitivities were obtained for the longest (C-19-Cit) and the shortest (C-10- Cit) cyclic α-enolase peptides compared to the C-14-Cit and C-12-Cit peptides, indicating that peptide length and conformation are crucial for peptides containing 12–19 amino acids. These findings are supported by the literature, where it has been reported that the reduced number of amino acids in the cyclic structure may constrain the peptide in a more locked conformation, reducing the flexibility of the peptides and hence negatively influencing the ACPA reactivity [16]. The fact that the smaller peptide (C-10-Cit) was as sensitive as the longest peptide (C-19-Cit), and thus more sensitive than the mid-length peptides (C-12-Cit and C-14-Cit), is intriguing. This interesting reactivity pattern to truncated cyclic peptides could be further investigated by performing crystallography studies of the ACPA binding groove. Ultimately, these findings regarding truncated peptides (both linear and cyclic) confirm that peptide length and conformation are essential for antibody reactivity.

Concerning the aspect of cyclization, it has been reported that peptide cyclization has a positive effect upon antibody reactivity [14]. Our studies of pro-filaggrin and EBNA-2 cyclic and linear peptides revealed an evident increase in the antibody reactivity to the cyclic peptides compared to the linear peptides. These results are consistent with the longest α-enolase peptides (L-19-Cit and C-19-Cit). Nevertheless, for the α-enolase peptides, the reactivity to linear peptides did not depend on peptide length, whereas ACPA reactivity to cyclic peptides was length-dependent. This effect is in direct contrast to earlier findings using cyclic and linear pro-filaggrin peptides, where ACPA reactivity to both the linear and the cyclic peptides appeared to be length-dependent [27]. This may in theory be ascribed to the peptides used, as the pro-filaggrin and EBNA-2 peptides were biotinylated, whereas the α-enolase peptides had free terminals, hence the absence of biotin may have influenced the peptide coating. Thus, further analyses are necessary to confirm these results.

When focusing on the peptide sequence, a high degree of homology is found between pro-filaggrin and EBNA-2 in the C-terminal region, where positively charged and small amino acids are present (Table 1). The high degree of sequence homology may explain the similar sensitivities that the peptides yielded, suggesting that peptide sequence influences antibody reactivity. However, the Proteoglycan Cit peptide sequence does not have homology in the C-terminal end to pro-filaggrin and EBNA-2 and still yields a sensitivity of approximately 50%. Since several sequence patterns along with a Cit–Gly motif can be found among the peptides tested, these observations led to the hypothesis that a structural homology could be shared by the citrullinated peptides recognized by ACPAs. This remains to be elaborated.

Screening of the RA and the HC cohorts on a peptide panel, including four citrullinated peptides from pro-filaggrin, proteoglycan, fibronectin and EBNA-2, revealed a significantly different reactivity between RA and HC samples (Figure 3). The citrullinated peptides obtained the following sensitivities: Pro-filaggrin 32%, Fibronectin 36%, Proteoglycan 50% and EBNA-2 68%. In addition, more than 50% of the RA samples reacted with more than one peptide of the panel, confirming the ability of ACPAs to bind to several citrullinated targets. Further investigation analysing the overlapping reactivities of the RA sera within the peptide panel (Figure 3) showed that approximately 32% reacted with 3 or 4 peptides, 21% with 2 peptides, 14% did not react at all and, lastly, 32% of the cohort only showed reactivity to one peptide, which in 56% of the cases was the EBNA-2 peptide, suggesting the presence of EBNA-2-specific ACPAs. The fact that most of the RA samples interacted with more than one peptide supports the theory of the overlapping reactivity of ACPAs, highlighting the central role of the Cit–Gly motif together with the surrounding amino acids for antigen–antibody binding.

The previous findings led us to compare the ACPA reactivities between the peptide panel and the α-enolase peptides, dividing them into α-enolase-positive and -negative ones. As shown in Table 2, within the α-enolase-positive cohort (*n* = 12), all the RA sera reacted with EBNA-2 peptide and no monospecific reactivity was detected to pro-filaggrin and proteoglycan peptides. One RA sample of the α-enolase-positive cohort only showed reactivity to the α-enolase peptide and not to the whole peptide panel, indicating that it was specific for the side-chains of the α-enolase epitope rather than the actual backbone in combination with a central Cit–Gly motif. In the α-enolase-negative cohort, more monospecific reactivities were observed. These findings are in accordance with the literature, which reports that approximately 15% of the RA sera are monospecific [30]. Additionally, although a small part of the RA sera reacted with only one citrullinated peptide, these results confirm the theory of overlapping and nonoverlapping ACPA reactivities. Here, the α-enolase-positive cohort was regarded to have overlapping ACPAs, which are considered as backbonedependent, whereas the α-enolase-negative cohort was accounted to the nonoverlapping group of ACPAs, which depend on the flanking amino acid side-chains to establish a stable antigen–antibody interaction [16]. Collectively, the RA sera that were positive for α-enolase turned out to be more overlapping within the peptide panel compared to the α-enolase-negative cohort.

#### **5. Conclusions**

RA is an autoimmune disease that affects many people and reduces their quality of life. Thus, early diagnosis of the disease is fundamental to undertake therapy as soon as possible and to prevent disease progression. One of the most important diagnostic criteria is the detection of autoantibodies directed to a variety of citrullinated antigens in the serum of the patient. The origin of these autoantibodies is still unknown, and thus, gaining knowledge about the epitopes that ACPAs are able to recognize is important in the development of more sensitive and more specific diagnostic tools and to obtain a better understanding of the pathophysiological mechanisms. For these purposes and to elucidate the etiology of RA, interactions between ACPAs and potential candidate (auto)antigens were analysed.

Regarding the experiments to further investigate the structure and composition of ACPAs, these results confirm their cross-reactive nature. The analysed RA sera showed a different reactivity pattern to the citrullinated peptide panel in relation to the α-enolasepositive and -negative cohorts. This result indicates that ACPAs can be divided into two categories based on their ability to bind to a wider or a more limited number of citrullinated targets, which, respectively, reflects the peptide backbone and peptide side-chains dependencies. Especially, the ability of ACPAs to react with various citrullinated targets can be explained by a similar structure of citrullinated epitopes. Early studies showed that positively charged and small amino acids in the C-terminal end relative to citrulline yield high sensitivities. However, not all the peptides tested in this study have these features but can still interact with ACPAs (Table 1). Therefore, a structural homology, rather than sequence homology, should be important for ACPA recognition of the citrullinated targets, and this would, in addition, support the theory that the overlapping group of ACPAs is backbone-dependent. To obtain a better knowledge of the structure and to find a pattern that brings together all the citrullinated targets that interact with ACPAs, circular dichroism analyses could be performed.

Within the peptides tested on the RA cohort, the epitope originating from α-enolase, a known autoantigen in RA, was confirmed to have a lower sensitivity compared to available commercial assays and other citrullinated peptides, such as the EBNA-2 peptide, which was confirmed to be a highly sensitive substrate, even though contribution of EBV infection to RA onset still needs to be clarified.

Collectively, this study contributes to the understanding of the nature of ACPAs.

**Author Contributions:** Conceptualization, G.H., P.R.H., P.R. and N.H.T.; methodology, I.F., G.H., N.H.T. and P.R.; software, I.F. and N.H.T.; validation, I.F., G.H. and N.H.T.; formal analysis, I.F. and N.H.T.; investigation, I.F., G.H., N.H.T. and P.R.; resources, G.H., J.F. and P.R.; data curation, I.F. and N.H.T.; writing—original draft preparation, I.F. and N.H.T.; writing—review and editing, I.F., G.H., N.H.T., P.R.H. and P.R.; supervision, G.H., N.H.T. and P.R.; project administration, G.H., N.H.T., P.R. and P.R.H.; funding acquisition, G.H., N.H.T. and P.R. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Lundbeck foundation, grant number R231-2016-3622.

**Informed Consent Statement:** Patient consent was waived as the samples were tested anonymously when tested for diagnostic purposes. The tested samples were not traceable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Recombinant Antibody Production Using a Dual-Promoter Single Plasmid System**

**Stefania C. Carrara 1,2,†, David Fiebig 1,2,†, Jan P. Bogen 1,2,†, Julius Grzeschik 2, Björn Hock <sup>3</sup> and Harald Kolmar 1,\***


**Abstract:** Monoclonal antibodies (mAbs) have demonstrated tremendous effects on the treatment of various disease indications and remain the fastest growing class of therapeutics. Production of recombinant antibodies is performed using mammalian expression systems to facilitate native antibody folding and post-translational modifications. Generally, mAb expression systems utilize co-transfection of heavy chain (*hc*) and light chain (*lc*) genes encoded on separate plasmids. In this study, we examine the production of two FDA-approved antibodies using a bidirectional (BiDi) vector encoding both *hc* and *lc* with mirrored promoter and enhancer elements on a single plasmid, by analysing the individual *hc* and *lc* mRNA expression levels and subsequent quantification of fully-folded IgGs on the protein level. From the assessment of different promoter combinations, we have developed a generic expression vector comprised of mirrored enhanced CMV (eCMV) promoters showing comparable mAb yields to a two-plasmid reference. This study paves the way to facilitate small-scale mAb production by transient cell transfection with a single vector in a cost- and time-efficient manner.

**Keywords:** monoclonal antibodies; promoters; bidirectional; antibody production; upstream processing

#### **1. Introduction**

With the growing interest in monoclonal antibodies (mAbs) for therapeutic applications, advances in antibody production have improved drastically over the last decades. Due to the more complex structure of antibodies, their production requires host cells capable of natively folding and modifying the mAb. Modifications include post-translational glycosylation, which is, among other functional properties, critical to reduce their immunogenicity [1]. For this purpose, mammalian cells fulfil the requirements as appropriate hosts for antibody production [2,3]. Advances in transfection protocols and cell engineering have boosted the use of suspension cell lines with the ability to grow at high densities, and increased production yields [4,5]. Further within the drug discovery and development process, stable cell lines are generated for the most promising candidate(s), while transient transfection is performed at earlier stages to yield research quantities of mAbs, sufficient for characterization of lead candidates. The accessibility of commercially available transfection reagents with high efficiencies and the use of disposable materials makes transient expression an efficient and cost-effective strategy during early drug discovery [6]. Human Embryonic Kidney 293 (HEK293) and Chinese Hamster Ovary (CHO) cells are commonly used for transient antibody expression, due to their high expression yields and human-like glycosylation patterns [1,7,8].

**Citation:** Carrara, S.C.; Fiebig, D.; Bogen, J.P.; Grzeschik, J.; Hock, B.; Kolmar, H. Recombinant Antibody Production Using a Dual-Promoter Single Plasmid System. *Antibodies* **2021**, *10*, 18. https://doi.org/10. 3390/antib10020018

Academic Editor: Itai Benhar

Received: 24 March 2021 Accepted: 10 May 2021 Published: 13 May 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

The basis of successful antibody production is the correct folding of individual chains, followed by their accurate assembly, resulting in functional heterotetrameric glycoproteins. Misfolded or partially folded antibodies are degraded by the host cell's intrinsic quality control system, resulting in low production yields. Furthermore, antibodies with undesired folding are not able to effectively engage their target antigen or to mediate effector functions, have unfavourable pharmacokinetics, and tend to aggregate. Besides these biological limitations, purification of antibody products contaminated with aggregated or misfolded mAbs is a major hindrance in the downstream processing of therapeutic molecules and is currently the topic of numerous studies [9–12].

Antibody folding starts upon co-translational translocation into the endoplasmic reticulum (ER) [13]. Following the homodimerization of the two heavy chains (HC), the light chains (LC) are associated and covalently linked via disulphide bonds [14]. The glycosylation at Asn297 is linked to the CH2 backbone in a co-translational manner [15]. During translation, chaperones are involved to ensure the correct folding of the individual domains, as well as the final assembly of the tetrameric mAb [16].

At present, the largest part of transient production of mAbs is carried out using a two-plasmid system, also known as co-transfection, for the expression of *lc* and *hc*, with each gene driven by its own promoter and transcribed separately [17,18]. These are carried out, for the most part, with an equimolar ratio of heavy chain and light chain genes. Nonetheless, contradictory results have been reported using Expi293-F cells. While some publications report that an equimolar gene ratio results in the highest yield of fully assembled IgGs [8], others have described optimal expression with a 1:2 ratio of heavy and light chain genes, respectively [19].

A large drawback of the two-plasmid system is the moderate control over the relative expression of *lc* and *hc*, with fluctuating cell-to-cell transfection efficiencies [17]. On the other hand, a bidirectional (BiDi) vector with a dual-promoter can ensure the introduction of both *hc* and *lc* genes into each cell in equal amounts. However, for some applications it could be beneficial to have a stronger expression of one chain over the other. By choosing two suitable promoters controlling the transcription of *hc* and *lc*, different ratios can be achieved [20]. While diverse approaches have been developed throughout the years to advance and facilitate recombinant antibody production, the design of expression vectors plays a large role for optimization of expression yields. Recently, Bayat and colleagues (2018) compared the use of three different vector design strategies for the expression of IgG1 antibodies in CHO cells, namely using the conventional two-vector approach with *hc* and *lc* encoded separately, a bicistronic vector based on internal ribosome entry sites (IRES), and a dual-promoter single vector approach. All vectors were under the control of a human cytomegalovirus (CMV) promoter. Expression analysis revealed that the dual-promoter vector system resulted in the highest mAb yield [21].

Andersen and co-workers have previously shown the ability of the CMV enhancer to control two core CMV promoters simultaneously, resulting in efficient antibody expression [22]. With this basis, we sought to investigate different promoters in a bidirectional format to facilitate transient transfection, avoiding co-transfections. We also sought to simplify *hc* and *lc* gene cloning by establishing a one-step Golden Gate cloning procedure that relies on the simultaneous plasmid incorporation of *hc* and *lc* genes together with the bidirectional promoter sequence. By analysing *hc* and *lc* gene expression, as well as through subsequent quantification of fully-folded IgGs, promoter and enhancer element combinations were compared. Here, we show the use of a dual-promoter, single plasmid approach using divergent promoters for the transient expression of two FDA-approved antibodies, Durvalumab and Avelumab, using both Expi293-F and ExpiCHO-S cell systems. This work lays the foundation to facilitate small-scale mAb production in drug discovery programs in a more efficient manner.

#### **2. Materials and Methods**

#### *2.1. Plasmids & Cloning of Constructs*

To allow the individual exchange of different variable domains, the utilization of κ and λ isotypes, as well as the usage of different BiDi promoter combinations, the backbone of the mammalian destination (MD) vector was built in a cassette-like manner. For vector amplification in *E. coli*, a chloramphenicol resistance was utilized, adjacent to the colE1 and the f1 origins. A stuffer sequence, flanked by *Esp*3I restriction sites, was downstream of an inverse-orientated SV40 polyA sequence that was intended to be a terminator signal for the light chain cassette. Upstream of the stuffer sequence, a partial hinge followed by the CH2 and CH3 domains of a human IgG1 were encoded. Again, a SV40 polyA signal sequence served as a terminator signal (Figure 1A). The plasmid was de novo designed in silico and ordered at GeneArt (Regensburg, Germany).

**Figure 1.** Schematic illustration of BiDi promoter system for antibody production. (**A**) The MD vector was designed to exhibit a 200-bp stuffer, flanked by *Esp*3I restriction sites (*Esp*3I sites A and B), adjacent to a SV40 polyA signal sequence and the regions encoding for hinge-CH2-CH3, terminated by a SV40 polyA signal sequence. (**B**) VL-CL and VH-CH1 amplicons can be inserted into MD by Golden Gate cloning utilizing *Esp*3I restriction sites (*Esp*3I sites A and B). The BiDi promoter can be chosen individually and is flanked by *Bbs*I sites (*Bbs*I sites A–D), compatible with the VL and VH sequences. (**C**) Golden Gate assembly results in a fully functional and re-circularized vector, with the light chain under the control of promoter I and the heavy chain under the control of promoter II. (**D**) Schematic representation of the resulting heterotetrameric IgG1 antibody using the same colour code as for the genetic elements.

The selected promoter sequences were either ordered as gene strings at Twist Bioscience (EF-1α, minCMV-enh-CMV (GenBank: MK764037) or PCR-amplified from the pTT5

CMV promoter cassette between bases 42–1185 (hereinafter referred to as eCMV) [23]. To allow for the correct orientation of the promoter sequences, individual primers were used to introduce the respective *Bbs*I Golden Gate cloning (GGC) signature overhangs. Genes for VH-CH1 and VL-CL of Durvalumab (Imfinzi, κ light chain) and Avelumab (Bavencio, λ light chain) were also ordered as gene strings, already bearing suitable signature sequences for *Esp*3I and *Bbs*I as well as their respective leader sequences. The 200-bp stuffer used between individual promoters consists of the non-functional 3 coding region of the amp resistance gene for beta-lactamase, followed by ~40 bp of non-coding DNA. Assembly of the MD expression constructs was conducted with 75 ng destination vector and equimolar amounts of the respective fragments, 20 U *Bbs*I-HF, 10 U *Esp*3I, and 200 U T4-DNA ligase (NEB, Frankfurt, Germany) for 30 cycles (1 min; 16 ◦C; 37 ◦C). For the reference constructs, VH and VL genes were amplified incorporating *Sap*I restriction sites and then inserted into a pTT5-derived vector utilizing CH1-CH2-CH3 or κ/λ entry vectors using GGC as described before [24,25]. PCR reactions were performed utilizing Q5 polymerase (NEB) according to the manufacturer's protocol and purified using the Wizard SV Gel and PCR Clean-up System (Promega, Walldorf, Germany). All primers can be found in Table S1. The DNA sequence for the 2xeCMV BiDi construct can be found in Sequence S1 in the Supplementary Information.

*E. coli* XL1-blue were transformed utilizing the Golden Gate reaction mixtures and cultivated on chloramphenicol or ampicillin DYT agar plates for MD or pTT5 constructs, respectively. Resulting colonies were sequenced at MicroSynth SeqLab (Göttingen, Germany), and positive clones were utilized to inoculate 50 mL overnight cultures. Plasmid DNA for transient transfection was isolated using the PureYield Plasmid Midiprep System (Promega, Walldorf, Germany).

#### *2.2. Cell Lines*

Expi293-F and ExpiCHO-S cells were obtained from Thermo Fisher Scientific. Cells were incubated at 37 ◦C, 8% CO2, 110 rpm, and sub-passaged every 3–4 days in their respective expression media, as described in the manufacturer's protocol (Thermo Fisher Scientific, Schwerte, Germany). Cell count and viability were measured using an automated cell counter (Bio-Rad TC-20) based on trypan blue staining. Cell densities were maintained between 0.3–4 × 106 cells/mL and 0.2–6 × 106 cells/mL for Expi293-F and ExpiCHO-S, respectively.

#### *2.3. 24-Well Transfection*

For gene expression and protein quantification, small-scale transfections using Axygen 24-well deep-well plates (Corning, New York, NY, USA) were performed. One day prior to transfection, cells were seeded into wells at a final cell density of 1.8 × <sup>10</sup><sup>6</sup> or 3 × 106 viable cells/mL in 2.5 mL expression medium for Expi293-F or ExpiCHO-S, respectively, and incubated under shaking conditions in a humified atmosphere at 37 ◦C, 8% CO2, 225 rpm. The following day, the cell density was adjusted to 3 × <sup>10</sup><sup>6</sup> or 6 × 106 viable cells/mL in 2.5 mL expression medium for Expi293-F or ExpiCHO-S, respectively. DNA:Expifectamine complexes were incubated at room temperature with either 3 μg BiDi plasmid or 2 μg heavy and 2 μg light chain plasmid for co-transfections (two-plasmid reference) for 20 or 1 min for Expi293-F or ExpiCHO-S, respectively, before adding dropwise to the cells. Feeding procedures were carried out according to manufacturer's instructions. For gene expression analysis, cells were harvested 3 days post-transfection, while protein quantification was carried out 6 days post-transfection.

#### *2.4. RNA Isolation*

Three days post-transfection, Expi293-F or ExpiCHO-S cells were harvested by centrifugation and cell pellets were processed through a QIAshredder column (QIAGEN, Hilden, Germany). Total RNA extraction was carried out using RNeasy Mini Kit (QIAGEN) following the manufacturer's instructions. RNA concentration was determined spectroscopically using a NanoDrop One (Thermo Fisher), ensuring pure RNA was isolated with a A260/280 ratio of 2.0.

#### *2.5. Gene Expression Analysis by Reverse Transcription Quantitative Polymerase Chain Reaction (RT-qPCR)*

Expression levels of heavy and light chain (both κ and λ) were analysed using 100 ng RNA per well in Hard-Shell® 96-well PCR plates (Bio-Rad, Hercules, CA, USA) and iTaq Universal SYBR Green One-step Kit (Bio-Rad) with designed SYBR Green primers (Sigma Aldrich, Munich, Germany) using a CFX96 qPCR instrument (Bio-Rad). Relative expression levels were analysed using the integrated software from Bio-Rad (CFX Manager, Hercules, CA, USA) and normalized to housekeeping genes GAPDH and RPLP0 (IDT, Coralville, IA, USA). The primers used can be found in Table S2.

#### *2.6. Protein Purification*

To purify the antibodies from small-scale transfections, cells were harvested by centrifugation and cell culture supernatants were purified using Protein A HP SpinTrap columns (Cytiva, Freiburg im Breisgau, Germany) following the manufacturer's protocol. Antibodies were eluted in 0.1 M glycine-HCl, pH 2.7. Protein concentration was determined using a NanoDrop One (Thermo Fisher) using the corresponding molecular weights and extinction coefficients.

#### *2.7. Protein Quantification and Affinity Determination Using Biolayer Interferometry (BLI)*

Six days post-transfection, cells were harvested by centrifugation and the cell culture supernatants were sterile-filtered (0.45 μm). BLI experiments were performed on an Octet Red96 (FortéBio, Fremont, CA, USA). Using Protein A biosensors (Sartorius, Göttingen, Germany) for quantification, mAb concentration was measured from the cell culture supernatants. An in-house produced mAb was used as a standard within the range of 3.13–400 μg/mL.

For affinity determination, anti-human Fab-CH1 2nd generation (FAB2G) biosensors (Sartorius) were used. Purified antibodies were loaded onto the tips at 10 μg/mL until a layer thickness of 1 nm was reached. Association was measured using a serial dilution of His-PD-L1-TwinStrep (produced in-house). Kinetics were determined using Savitzky-Golay filtering and a 1:1 Langmuir binding model.

#### **3. Results**

To produce full-length antibodies in a bidirectional manner, we first designed the respective vector in silico. This vector encoded the fragment crystallizable (Fc) region of an IgG1, which is the most common isotype found in therapeutic antibodies [26]. In order to allow the flexible use for a variety of binders, the fragment antigen binding (Fab), which can be of the κ or λ isotype, was not encoded on the plasmid. Instead, a stuffer sequence, flanked by *Esp*3I sites was incorporated (Figure 1A).

As a reference antibody to establish different promoter combinations, the Fab of Durvalumab, an FDA-approved anti-PD-L1 antibody of the κ type, was chosen [27]. PCR amplicons of VH-CH1 and VL-CL were generated introducing *Esp*3I and *Bbs*I restriction sites. By utilization of a BiDi promoter system, flanked by *Bbs*I sites, a Golden Gate reaction resulted in a re-circularized vector (Figure 1B,C). In this process, the stuffer that was included in the parental MD vector was replaced by a CL-VL-PromoterI-Stuffer-PromoterII-VH-CH1 sequence. Owing to this cloning strategy, a straightforward exchange of different Fabs and different BiDi promoters is feasible. The resulting vector exhibited the functional ORFs for the heavy and the light chains, resulting in the production of full-length antibodies (Figure 1D).

#### *3.1. Cloning Promoter Combinations*

Based on the findings from Andersen et al. [22] that a single enhancer adjacent to a bidirectional CMV promoter allows for efficient antibody production using a single expression plasmid, six different bidirectional promoter complexes were generated comprising not only the minCMV-CMV cassette, but also combinations of other strong promoters including their individual enhancer elements, such as the CMV with its major immediate early enhancer (MIE), the optimized CMV cassette from pTT5 (denoted as eCMV in this study), and the human translation elongation factor 1 alpha (EF-1α) promoter, that were selected based on their capability to produce fully-folded IgG molecules.

The MIE-CMV promoter, being one of the most commonly used in mammalian expression vectors, is designated as a strong driver for recombinant protein expression [28,29]. As shown before by Andersen and colleagues [22], the MIE enhancer is also capable of facilitating elevated expression levels in the divergently oriented minCMV promoter, although to a lesser extent. This was based on the presumption that the formation of the large transcription complex might be sterically hindered. As a reference, we selected a similar bidirectional promoter setup lacking the unique sequence upstream of the enhancer. Additionally, we went for a mirror-symmetric approach comprised of two individual CMV promoters, each having adjacent MIE enhancers that were separated by a 200-bp-stuffer. The eCMV cassette comprises—besides the MIE enhancer and core promoter—several additional regulatory elements that have been described to increase expression levels. These elements include the non-coding adenoviral tripartite leader sequence (TPL), the adenovirus major late promoter enhancer (MLP enh.), as well as distinct splicing sites allowing for prolonged mRNA stability [30,31]. Furthermore, we also utilized the strong human translation elongation factor 1 alpha promoter (EF-1α) that has proven to be advantageous over the CMV promoter in some cell types and in the expression of distinct proteins of interests [32,33].

Based on the modular setup of our MD vector and utilization of GGC, we generated different combinations of the aforementioned promoters to analyse for highest full-length antibody expression levels and product yield of Durvalumab. A schematic overview of the BiDi combinations is depicted in Figure 2.

**Figure 2.** Overview of the different bidirectional combinations tested. The 200-bp stuffer sequence is marked in red for each construct. Abbreviations: minimal CMV (minCMV), cytomegalovirus promoter (CMV), enhanced CMV (eCMV), major immediate early enhancer (MIE), human translation elongation factor 1 alpha (EF-1α), adenoviral tripartite leader sequence (TPL CDS), adenovirus major late promoter enhancer (MLP enh.), splicing donor site (SD), splicing acceptor site (SA), light chain (LC), heavy chain (HC).

#### *3.2. Gene Expression Analysis in Mammalian Cells*

Transient transfections of all six promoter combinations were performed in 24-well plates using Expi293-F cells, an established cell line for transient antibody production. Three days post-transfection, cells were harvested, and RNA was isolated. Relative gene expression of *hc* and *lc* mRNA levels was measured by RT-qPCR (Figure 3).

**Figure 3.** Gene expression analysis of heavy and light chain genes after transient transfection of Durvalumab in Expi293-F cells. (**A**) Bar chart representing heavy (dark blue) and light (light blue) chain mRNA expression in the different constructs. Values are relative to the CMV-minCMV construct and normalised to housekeeping genes GAPDH and RPLP0. Error bars represent the standard error of the mean of technical triplicates. (**B**) Heat map representation of gene expression analysis. The relative normalised gene expression for light and heavy chain mRNA is shown on the right.

The data was set relative to the CMV-minCMV construct and normalised to housekeeping genes. Looking at relative mRNA levels, both variants with minCMV and CMV, independent of the promoter orientation, did not yield high mRNA expression for either *lc* or *hc*. Remarkably, the combinations with the EF-1α promoter and the enhanced CMV cassette (eCMV) showed significant differences in expression levels depending on their orientation. Steering *lc* expression with the EF-1α promoter (EF-1α-eCMV) resulted in a 9.5-fold upregulation in *lc* and 8-fold upregulation in *hc* mRNA levels. Conversely, having the more potent eCMV in the light chain direction and EF-1α in the heavy chain direction (eCMV-EF-1α) led to a 9-fold *lc* upregulation, and low relative *hc* expression levels. Based on these results, mirrored constructs containing both promoter and enhancer cassettes in both directions were tested, one with the traditional CMV promoter (2xCMV), and the other with the eCMV cassette (2xeCMV). Interestingly, the 2xCMV construct did not result in increased *lc* or *hc* mRNA levels compared to the bidirectional construct containing two identical eCMV promoters or constructs containing two different promoters. The BiDi combination of two mirrored eCMV promoter cassettes (2xeCMV) yielded in a 7-fold upregulation of *lc* mRNA and a modest upregulation in *hc* levels.

As promoter strength may vary depending on the cell line, particularly for the CMV promoter [33], ExpiCHO-S cells were also investigated. Gene expression analysis resulted in similar results as in Expi293-F cells, with the 2xeCMV complex showing the

strongest upregulation in both *hc* and *lc* mRNA levels compared to the other promoter combination, namely a modest 10-fold upregulation of *hc*, and a 20-fold upregulation of *lc* mRNA (Figure 4).

**Figure 4.** Gene expression analysis of heavy and light chain genes after transient transfection of Durvalumab in ExpiCHO-S cells. (**A**) Bar chart representing heavy (dark blue) and light (light blue) chain mRNA expression in the different constructs. Values are relative to the CMV-minCMV construct and normalised to housekeeping genes GAPDH and RPLP0. Error bars represent the standard error of the mean of technical triplicates. (**B**) Heat map representation of gene expression analysis. The relative normalised gene expression for light and heavy chain mRNA is shown on the right with their respective scales.

#### *3.3. Protein Yield Determination via BLI*

As mRNA transcript levels do not indicate successful secretion of fully functional recombinant antibodies, protein quantification studies were performed. Expi293-F cells were transiently transfected and the amount of secreted therapeutic antibody Durvalumab [27] was quantified by biolayer interferometry (BLI) using sterile-filtered cell culture supernatants six days post-transfection. The mAb concentrations for the different constructs were interpolated from a standard curve generated using an in-house produced mAb. In line with the gene expression analysis, protein quantification resulted in a clear ranking of the different promoter combinations (Figure 5).

**Figure 5.** Protein quantification of Durvalumab in cell culture supernatants from transfected Expi293-F cells. (**A**) Table showing the mAb concentrations from 24-well transient transfections, listed according to their rank. The ranks 1–5 were set based on the mAb concentration of the different BiDi combinations. (**B**) Bar chart representation of BiDi mAb concentrations.

From the bidirectional combinations tested with transient transfections, the 2xeCMV BiDi construct showed the highest mAb concentration with 353 μg/mL (Figure 5). Interestingly, mRNA expression was higher for EF-1α-eCMV compared to 2xeCMV for both *hc* and *lc*, but antibody yield was significantly reduced in Expi293-F cells (Figure 3). This may likely be due to the fact that EF-1α-eCMV displayed similar *hc* and *lc* mRNA expression patterns, while 2xeCMV revealed a higher accumulation of the *lc* compared to the *hc* mRNA. It is well known that the expression of excess light chain over heavy chain is often beneficial for antibody production [34–37]. These findings corroborate that 2xeCMV resulted in the most promising bidirectional promoter combination, particularly in view of the fact that the usage of this promoter combination resulted, both in Expi293-F and ExpiCHO-S cell lines, in significantly enhanced mRNA synthesis with an excess of *lc* over *hc*.

#### *3.4. Correlation of mRNA and Protein Levels*

After performing both mRNA and protein studies, the correlation of the cycle threshold (Ct) values from RT-qPCR and the mAb concentrations from protein quantification in Expi293-F was determined (Figure 6A). Overall, a Pearson's coefficient of −0.6407 was calculated, indicating, as expected, a negative correlation between Ct values and mAb concentration. Looking at the respective Ct values of either the *hc* or *lc* for 2xeCMV, one can observe they have a low Ct value and resulted in the largest mAb yield. While some claim that abundance of *hc* may hinder productivity, it appears that the excess of *lc* allowing for correct mAb folding and assembly is sufficient for higher mAb yields. On the contrary, the Ct values of eCMV-EF-1α *lc* is similar to that of 2xeCMV, with the only difference being in the *hc* expression, ultimately resulting in much lower yields. Thus, this further shows the importance of a fine-tuned *hc* and *lc* expression, substantiating the potential of 2xeCMV for a BiDi antibody production system.

**Figure 6.** (**A**) Correlation of mAb concentration and Ct values for both heavy and light chain expression in Expi293-F for production of Durvalumab. (**B**) Quantification of antibody concentration for the production of Durvalumab and Avelumab using either a 2-plasmid reference or the BiDi 2xeCMV construct. Error bars represent the standard error of the mean of biological triplicates, while the symbols represent the individual measurements.

#### *3.5. Antibody—And Light Chain-Independence*

To ensure the promoter combination used for our BiDi technology was not antibodyor light chain isotype-dependent, the FDA-approved anti-PD-L1 antibody Avelumab of the λ isotype was also tested [38]. As 70% of approved antibodies belong to the κ type [39,40], only the most promising 2xeCMV construct was used to validate the established system for a λ-based mAb, compared to the conventional co-transfection approach. For co-transfection, Expi293-F cells were transfected using a 1:1 ratio of HC and LC DNA, with each plasmid carrying the same promoter and enhancer cassette as the bidirectional vector, namely eCMV. As can be appreciated from Figure 6B, there was no significant variation in antibody yield

between transfections with BiDi and the two-plasmid reference for either Durvalumab or Avelumab. Similarly, no variation was observed in ExpiCHO-S production (data not shown). Kinetics determination of both antibodies produced with either the two-plasmid reference or our established 2xeCMV BiDi plasmid bound to its target PD-L1 with comparable affinities (Table 1, Figures S1 and S2). Additionally, SDS-PAGE analysis under reducing conditions resulted in the expected heavy and light chains bands (Figure S3).

**Table 1.** Affinity determination of Avelumab and Durvalumab using either co-transfection or 2xeCMV BiDi plasmid.


Thus, this indicated the established system is compatible with different binders and can be employed for both κ- and λ-light chain isotypes.

#### **4. Discussion**

This work describes the generation of a bidirectional vector construct for recombinant antibody expression using two eCMV promoter cassettes, controlling the expression of *lc* and *hc* individually in each direction. By performing a thorough analysis of different bidirectional promoter combinations with varying lengths and strengths, the 2xeCMV combination showed the most promising *hc* and *lc* mRNA synthesis in two regularly used mammalian cell lines, and, more importantly, the highest yields after protein quantification comparable to those using the conventional two-plasmid system. Other BiDi constructs showed potentially promising gene expression profiles, such as EF-1α-eCMV combination, with high relative mRNA levels of both *hc* and *lc* mRNA. Nonetheless, having higher levels of *hc* mRNA in the cells appear to curb productivity of fully folded IgG formation, resulting in drastically decreased antibody yields compared to both the two-plasmid reference and the 2xeCMV BiDi vector.

While using a bidirectional approach takes away some flexibility in terms of being able to alter the *hc:lc* ratios during co-transfection, the overexpression of both genes and, especially the excess *lc* expression, results in sufficient material for antibody hit screening. For convenience, we established a one-step cloning procedure for simultaneous plasmid incorporation of the heavy and light chain encoding segments obviating the need for generating two separate plasmids. Not only does this approach lower plasmid preparation efforts, but it also increases handling for transfection of numerous mAbs during screening and characterization as only a single plasmid is required. The use of two FDAapproved antibodies with either a κ- or λ-light chain shows there is no antibody- or light chain-dependence using this system, indicating that it can be implemented ubiquitously. Further options remain to increase the yields of IgG molecules, such as optimization of the stuffer region between the two eCMV promoter cassettes, potentially reducing any steric hinderance and increasing transcription efficiency [22,41].

In conclusion, this work displays the benefits of using a one-plasmid bidirectional system with 2xeCMV promoters for fully folded IgGs within drug discovery. In terms of practicality, handling of a single plasmid for antibody production may be superior to the conventional way. Moreover, yields of fully folded IgGs are comparable between the two systems. Future directions for this technology go beyond recombinant production of classical antibody formats, as reduction of the number of plasmid constructs could also be considered feasible for the expression of bispecific antibodies and other antibody formats in the frame of antibody drug discovery.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/antib10020018/s1, Table S1: Primers used for cloning of bidirectional constructs, Table S2: RT-qPCR primers for HC and LC constant regions, Figure S1: Affinity determination by BLI of antibodies produced in Expi293-F, Figure S2: Affinity determination by BLI of antibodies produced in ExpiCHO-S, Figure S3: SDS-PAGE analysis of purified antibodies, Sequence S1: DNA sequence of the designed Durvalumab-2xeCMV insert.

**Author Contributions:** Conceptualization, S.C.C., D.F., J.P.B., J.G. and H.K.; methodology, S.C.C., D.F., J.P.B.; investigation S.C.C., D.F. and J.P.B. data curation, S.C.C., D.F., J.P.B. and J.G. supervision B.H. and H.K. writing original draft, S.C.C., D.F. and J.P.B. writing review & editing S.C.C., D.F., J.P.B., J.G., B.H. and H.K. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded by the Ferring Darmstadt Labs at the Technical University of Darmstadt and by GPRD at Ferring Holding S.A., Saint Prex.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available within this article and its Supplementary Materials.

**Acknowledgments:** S.C.C., D.F. and J.P.B. contributed equally to this work. The authors would like to thank GPRD for funding. The funders had no role in study design, data collection, data analysis, decision to publish, or preparation of the manuscript. We acknowledge support by the Deutsche Forschungsgemeinschaft (DFG—German Research Foundation) and the Open Access Publishing Fund of the Technical University of Darmstadt. Figures were created with Biorender and data was processed using GraphPad Prism 8.

**Conflicts of Interest:** J.G. and B.H. are employees of Ferring Pharmaceuticals, while S.C.C., D.F. and J.P.B. are employed by the Technische Universität Darmstadt in frame of a collaboration with Ferring Pharmaceuticals. All authors declare no conflicts of interest.

#### **References**

