A Method for Checking Recombinant Protein Quality to Troubleshoot for Discordant Immunoassays

Discordant results of recombinant protein-based enzyme-linked immunosorbent assays (ELISA) and other antigen detection tests are a common and vexing problem in scientific research and clinical laboratories. The reproducibility of immunoassays based on antibody specificity can be adversely affected by cryptic changes in the composition of the analyte (e.g., the protein antigen). By use of sodium dodecyl sulfate-polyacrylamide gel electrophoresis and mass spectrometry analyses, we found varying levels of purity and the extent of post-translational modifications (PTMs), particularly of N-linked glycosylation and phosphorylation, among varied lots of recombinant glucose-regulated protein 78 (GRP78) expressed in Escherichia coli. We expect these lotto-lot variabilities of both purity and PTMs led to inconsistent results stemming from ELISA assays that measured GRP78 autoantibody levels in patient plasma specimens. We present these analyses to draw readers’ attention to a potentially common, yet seldom appreciated problem, in laboratory assays using recombinant proteins as antigens for antibody detection, and propose a workflow to detect and troubleshoot for this typical problem.


27
Poor reproducibility for recombinant protein-based assays can create many challenges for 28 both basic and clinical research, and these challenges deserve more attention to ensure the quality of 29 both endpoint and downstream studies [1][2][3]. Although recombinant proteins are essential tools in 30 laboratory assays, quality control (QC) of these reagents is often overlooked [4]. Without the proper 31 QCs, results derived from these assays may otherwise lead to spurious and/or irreproducible data 32 that waste time and effort, and could potentially lead to therapeutic errors [4].

50
In this project, we have undertaken a simplified and affordable approach that could be 51 similarly adopted by any lab that has access to a MS core facility. As an example, we took 52 commercially available 1-D gels combined with standard MS2 analysis that included proteomic 53 searches, such as cleaved N-glycation sites following Peptide-N-glycodidase F (PNGase F) treatment 54 (i.e. deamination products on glutamine and arginine); mass shifts associated with citrullination of 55 arginine, unexpected cleavage sites, common oxidation events, and O-Linked b-N-acetylglucosamine 56 (O-GlcNac); and phosphorylation changes on serine, threonine, and tyrosine [17,18]. Utilizing the 57 clear overview illustrated herein, one can run a gel for the first QC step, and submit a gel piece for 58 MS2 analysis with any MS core facility.

59
Escherichia coli (E. coli) is the most commonly utilized expression system for the production 60 of recombinant proteins. Although protein PTM is a fundamental and ubiquitous process among all 61 living organisms [19][20][21], it was widely believed that E. coli does not possess PTM capacity. However, 62 recent studies have reported the existence of analogous processes in prokaryotes, particularly in E.

63
coli [22][23][24]. Advances in MS-based proteomics and analysis software have enabled us to identify the 64 presence and quantify the abundance of a particular PTM with high confidence [25]. In this study, 65 besides identifying the varying levels of rGRP78 purity, we also discovered different levels of PTM's 66 in differing lots of this commercially supplied product.

67
PTMs potentially cause many unwanted effects in recombinant proteins [26][27][28] such as 68 improper folding, aggregation, and immunogenicity, which can lead to poor efficacy and toxicity of 69 biosimilar drugs [26][27][28][29]. However, we found fewer reports describing similar problematic effects of 70 PTMs in recombinant protein products used in pre-clinical or basic research immunoassays. To the 71 best of our knowledge, the varying level of purity and PTMs of recombinant protein from commercial 72 vendors as potential sources of inconsistent results in laboratory assays such as ELISA has not been 73 well-documented, and many researchers may be unaware of this issue. In this report we illustrate a 74 simplified experimental GeLC-MS2 workflow, including an informatics approach as an additional 75 tool to QC recombinant proteins, in order to sidestep and/or troubleshoot downstream problems. We 76 hope to raise awareness of this potential complication of recombinant protein-based assays, while 77 helping both clinical and basic scientists along the way.

118
The gel was stained overnight with Colloidal Coomassive. In all samples, molecular weight (MW) 119 bands at ~78kDa were excised and digested with Trypsin Gold (Promega, Madison, WI) prior to 120 nLCMS2 analyses as previously reported [32] . Since we observed different levels of impurities in 121 some samples, we divided the samples into two groups based on different impurity levels ( Figure 2).

122
A sample from each of these two groups was picked at random and excised into 6 MW fractions to 123 test the identities of the impurities.

124
Following trypsin digestion, the peptides were extracted, concentrated until nearly dry 125 under vacuum, and diluted in 0.1% formic acid prior to analysis by 1-D reverse nano-scale liquid 126 chromatographic electrospray ionization multi-stage tandem mass spectrometry (nLC-ESI-MS2).    Figure S1). These inconsistent results 177 led us to conduct troubleshooting assays to fully characterize potential protein contaminants, in 178 addition to the measuring the relative purity of different lots of rGRP78. In these studies, we 179 expressed and purified rGRP78 from E. coli in our laboratory, in order to minimize lot-to-lot variation 180 of the protein antigen.

209
MW bands at ~78kDa from all recombinant protein samples were excised and digested with 210 trypsin prior to LCMS2 analyses. The overall GeLC-MS2 workflow for peptide/protein 211 identification is illustrated in Figure 3. Briefly, after the MS/MS spectra were generated, we used

212
MASCOT distiller to match each MS/MS spectra to a peptide sequence. We found that within the 213 ~78kDa band, there were fairly high levels of purity (up to ~90% of GRP78 spectral counts in the 214 whole lane digestion) among different rGRP78 preparations (Table 1).

221
Based on the SDS-PAGE results, we also divided the samples according to their purity levels 222 into two groups and randomly selected one representative sample in each group to be cut into

275
In troubleshooting to account for discordant ELISA results, we applied a standard GeLC-MS2 276 workflow, which includes the use of SDS-PAGE followed by LCMS2, to verify the quality and purity 277 of the antigen sources, in this case: rGRP78 expressed in E. coli. We found varying levels of purity 278 across each lot of rGRP78 from the same commercial product and one preparation produced in our 279 laboratory. LCMS2 analyses showed that the majority of apparent impurities observed in the 280 different MW bands in stained 1-D Gels were cleaved products of human GRP78. However, variation 281 in purity levels alone cannot fully account for the discordant ELISA results, because while lot#130312 282 and our own lab-generated rGRP78 exhibit similar purity levels, the ELISA concordances of these 283 two preparations are not 100%. We also found varying PTM levels in these preparations, particularly 284 in phosphorylation and glycosylation modifications, although the locations of these PTMs with 285 respect to the particular amino acid residues in the protein do not differ. Therefore, besides varying 286 purity levels, changes in PTMs potentially lead to inconsistent results observed in our immunoassays.

287
In this report, we also presented a standard workflow to assist as a guideline in the testing for 288 the purity and PTM levels of recombinant protein products to help researchers perform QC analyses,

289
and when applicable, troubleshoot problematic laboratory immunoassays. Our report is intended to  Diagnostics 2020, 10, x FOR PEER REVIEW 9 of 11 raise awareness of variations within batches of "purified" proteins and suggests that such samples 291 are routinely tested to ensure the reproducibility and rigor of scientific research applications.

292
Supplementary Materials: Figure S1: High level of reproducibility in a representative ELISA assay to measure 293 GRP78 autoantibodies in COPD plasma; Figure S2: Protein ID's with % of total spectral counts/gel band; Figure   294 S3: The fractions of GRP78 peptides with phosphorylation modification from various rGRP78 preparations; 295 Figure S4A-E: The fractions of GRP78 peptides with glycosylation modifications (as inferred from the changes 296 in deamidation levels) from various rGRP78 preparations; Figure S5: An example of a side-by-side comparison 297 between the locations of various PTMs; Figure S6A-B: The fractions of GRP78 peptides with native deamidation 298 modification (i.e. without PNGase F treatment) in different rGRP78 preparations.

299
Conflicts of Interest: The authors declare no conflict of interest. 300 301