Structure‐based discovery of small molecules improving stability of human broadly‐neutralizing anti‐HIV antibody 2F5 in plant suspension cells

The production of biopharmaceuticals in engineered plant‐based systems is a promising technology that has proven its suitability for the production of various recombinant glyco–proteins that are currently undergoing clinical trials. However, compared to mammalian cell lines, the productivity of plant‐based systems still requires further improvement. A major obstacle is the proteolytic degradation of recombinant target proteins by endogenous plant proteases mainly from the subtilisin family of serine proteases. In the present study, the authors screened for putative small molecule inhibitors for subtilases that are secreted from tobacco BY‐2 suspension cells using an in silico approach. The effectiveness of the substances identified in this screen was subsequently tested in degradation assays using the human broadly‐neutralizing anti‐HIV monoclonal antibody 2F5 (mAb2F5) and spent BY‐2 culture medium as a model system. Among 16 putative inhibitors identified by in silico studies, three naphthalene sulfonic acid derivatives showed inhibitory activity in in vitro degradation assays and are similar to or even more effective than phenylmethylsulfonyl fluoride (PMSF), a classical inhibitor of serine proteases, which served as positive control.

ing to the MEROPS classification, subtilases belong to the S8A family of serine proteases. [8][9][10] A couple of secreted subtilases have been identified from tobacco suspension cells and their encoding genes have been cloned. [11,12] Inhibitor studies using intercellular washing fluid of Nicotiana benthamiana (Domin) leaf samples or conditioned tobacco BY-2 cell culture medium demonstrated the involvement of serine protease in the degradation process of recombinant human antibodies. [13][14][15][16] The subtilase family in plants is diverse, comprising 56 members in both Arabidopsis thaliana (L.) Heynh. [17] and N.
benthamiana. [5] Currently it is not known which particular members of the subtilase family interfere with the production of recombinant proteins, and therefore a broad inhibition is desirable. As the interference of a given small molecule inhibitor with the enzymatic activity differs among members of the subtilase family, broad inhibition might require the use of multiple inhibitors that are applied simultaneously. [15,18] Therefore, a screen for novel types of subtilase inhibitors is required to suppress unwanted proteolysis of recombinant proteins.
In the present study we used an in silico approach to screen for potential subtilase small molecule inhibitors. Taking advantage of the known three-dimensional structures of SBT3 from Solanum lycopersicum, [19] we generated structural models of two selected subtilases from tobacco that are particularly abundant in spent BY-2 culture medium. [7] Using a structure-based virtual screening approach, putative small molecule inhibitors were identified. In subsequent biochemical degradation assays, three out of sixteen candidate molecules were proven to effectively stabilize the mAb2F5 in conditioned culture medium of tobacco BY-2 cells.

Three-dimensional modeling of tobacco subtilases
Sequences for subtilisin-like proteases SBT1.1 (GenBank ID: ABQ58079) and SBT1.5 (GenBank ID: AIX97848) from Nicotiana tabacum were retrieved from the UniProt database. NtSBT1.1 and NtSBT1.5 sequences were initially edited by removing the first 23 amino acids, which were annotated as signal peptides. Since SBTs were known zymogens, sequences were further trimmed to obtain the mature active form, by removing the propeptides spanning the regions from amino acid 24 to His110 and His106, respectively.
Three-dimensional protein structures were generated by homology modeling using the Modeller (v9.1) program. [20] SBT3 from S. lycopersicum (PDB ID: 3I6S chain A, resolution 2.5 Å and 3I74-chain A, resolution 2.6 Å, [19] were chosen as modeling template, due to the highest sequence similarity (NtSBT1.1: 57%, NtSBT1.5: 55%). The model quality was assessed using MolProbity. [21] The protein models were prepared by adding hydrogen and adjusting ionization states of amino acid side chains using H ++ with standard options. [22] Afterwards, the models were further refined with a minimization procedure involving gradually releasing atom constraints using SYBYL-X 2.0 (Certara Inc.).

2.2
In silico screening for potential inhibitors  Transgenic tobacco BY-2 cell lines expressing mAb 2F5 were generated as described [24] and screened by Western blot analysis. The Prior to inhibitor studies the cells were transferred to D11b culture medium [25] to facilitate optimal antibody secretion.

In vitro degradation assays
Test compounds (Table 1)   The gels were subsequently blotted onto nitrocellulose membranes using the Semi-Dry device (BioRad, Germany). The mAb 2F5 chains were detected on the membrane by incubation with HRP-labeled goat anti-human IgG (H+L) affinity-purified antibodies (Dianova, Germany).
HRP activity was revealed using the SuperSignal West Pico PLUS chemiluminescent substrate (Thermo Fisher, Germany) and the light signal detected with a CCD camera using the UVP ChemStudio PLUS imaging system (Analytik Jena, Germany).

Modeling of tobacco subtilases and virtual screening for inhibitors
The sequences of NtSBT1.1 and NtSBT1.5 share high sequence similarity with S. lycopersicum SBT3 (57% and 55%, respectively). Using a standard homology modeling approach, we generated structural models for both enzymes ( Figure S1). Analysis of the stereochemical quality of the models using a Ramachandran plot revealed seven (1.1%, NtSBT1.1) and six (0.9%, NtSBT1.5) residues located in the unfavorable region. All of these are located in the loop regions.  (Table S1).
Eventually, ten compounds were selected for experimental validation of which eight were purchasable and soluble in DMSO ( Figure 1A).

In vitro degradation assays of mAb 2F5
The broadly neutralizing monoclonal anti-HIV antibody 2F5 has been proven unstable in spent culture medium from tobacco BY-2 suspension cells. Therefore, the spiking of intact, recombinantly produced, mAb2F5 into spent culture medium from BY-2 cells, provides a   Figure 5B).

DISCUSSION
The application of in silico approaches for the identification of novel bioactive substances has become a standard approach in various research areas. [26][27][28] Here we combined protein modeling of two subtilases with a virtual screen for novel enzyme inhibitors. This led to the identification of a novel compound that proved to be effective in our biochemical assay and prevented the degradation of mAb2F5 heavy chain in conditioned BY-2 culture medium. A hit expansion approach using a similarity search procedure further identified two even more potent compounds.
Our docking studies suggest that the 4,5-dual substituted naphthalene sulfonic acid scaffold seems to be particularly suited to closely The three molecules that we have identified as potent inhibitors have previously not been analyzed for their potential to inhibit serine proteases. Rather, they have a long history as reactants for the synthesis of azo dyes in organic chemistry. [29] A striking difference regarding their inhibitory potential has also been observed for the classical serine protease inhibitors PMSF and AEBSF ( Figure 2 ). While their structure is rather similar, with the AEBSF molecule possessing an ethylamino group in the para position ( Figure 1), only PMSF functions as an effective inhibitor in our biochemical degradation assay. Given these strong differences in inhibitor functions of structurally similar molecules, it is advisable to screen multiple different inhibitors to safeguard a target protein in a given biological matrix. In this respect, it is of particular interest to note the striking difference between PMSF and AEBSF as these inhibitors are often regarded as being equally effective and AEBSF is generally advertised as being preferable due to its lower toxicity compared to PMSF. [30] To assess the implementation of the identified inhibitors in a pro- could be used as recently described for the identification of apoplastic subtilisin-like proteases from N. benthamiana leaves. [16] Once the target proteases for CAS 83-64-7 have been identified it might be possible to optimise the inhibitor structure so that this molecule targets preferably those proteases involved in mAb degradation while preserving the activity of other proteases that are not involved in the degradation of recombinant proteins.