Utilizing a resource of enrichment profiles in plasma for the systematic assessment of antibody selectivity

There is a strong need for procedures that enable context and application dependent validation of antibodies. Here we describe the foundation for a resource aiding more detailed assessment antibody selectivity for capturing endogenous proteins from human plasma. In 414 immunoprecipitation (IP) experiments with EDTA plasma, data was generated by mass spectrometry (LC-MS) with 157 antibodies (targeting 120 unique proteins). Out of a total of 1,313 unique proteins, 426 proteins (33%) were detected in > 20% of the assays and indicate a background comprised of mainly proteins from the complement system. For all proteins identified either in heat-treated or untreated EDTA plasma, frequencies of occurrence were derived. We determined z-scores for each IP as a measure of enrichment to annotate the antibodies into four categories (ON-target, CO-target, OFF-target and NO-target). For 45% (70/157) of the tested antibodies, the expected target proteins were enriched (z-score ≥3) above background. There were 84% (59/70) of binders that co-enriched other proteins beside the intended target, either due to OFF-target binding or predicted interactions. Comparing several antibodies raised against IGFBP2, the established library allowed us to describe protein complexes in plasma, and we employed multiplexed sandwich immunoassays to confirm these. In summary, the generated resource of plasma enrichment profiles and background proteins adds a very useful and yet lacking starting point for the assessment of antibody selectivity in this clinically important body fluid. The provided insights will contribute to a more informed use of validated affinity reagents and may lead to further advancements of plasma proteomics assays.

6 also affect the sample's composition by causing aggregation and proteins to precipitate. As shown for 151 all IPs in Fig. 1C and Supplementary Figure 4A, heat treatment may influence which proteins are 152 listed among the common contaminants and their frequencies. When we consider LFQ intensities 153 instead of frequencies, we observed that in heat-treated samples particularly fibrinogens (FGA,FGB 154 and FGG) were more abundant (Supplementary Fig. 5). Previous observations state that denaturation 155 of fibrinogen starts at 55ºC (two transitions peaks at 57.7ºC and 96ºC) and particularly affects the D 156 fragment [21]. In heat-treated samples, the affinity of fibrinogens to plastic surfaces has also been 157 reported to increase [22]. Via a similar mechanism, fibrinogen's unspecific binding to surfaces of 158 magnetic beads or to heavy chains of IgG antibodies on the beads may be enhanced [23,24]. Hence, Extending the use of the library containing common contaminants and frequencies, we aimed at 165 classifying the antibodies based on their enrichment profiles in plasma. Considering a protein as being 166 enriched when the LFQ intensity was z-score ≥ 3, all antibodies were annotated according to the 167 following categories (Fig. 3A): 168  ON-target: when a z ≥ 3 was only assigned to the expected target. 169  CO-target: other proteins besides the expected targets were detected with z ≥ 3. Here we will 170 discuss three sub-categories related to the origin go the additional target. 171  OFF-target: proteins other than the expected targets were enriched with z ≥ 3. 172  NO-target: all detected proteins were classified with z < 3. 173 Knowing that common plasma background proteins may differ between assays using untreated and 174 heat-treated plasma, data from the respective assays was analyzed separately and the resulting z-scores 175 were combined for the global assessment. For the 157 antibodies, we made 1173 identifications for 176 681 unique proteins with z ≥ 3. In each of the 414 IPs, we detected about 550 proteins, of which 6-9 177 proteins were enriched above the threshold. The combined outcome of the analyses is shown in Table  178 1, where the classification of 70 out of 157 antibodies (45%) was found supportive by falling into the 179 ON-or CO-target category. When assessing only those antibodies targeting proteins previously 180 annotated in plasma, the fraction increased to 61% (46/75). It is noteworthy that annotating antibodies 181 in non-supportive categories could be due to limited affinity of the antibody, limited assay sensitivity 182 or absence of the target in the used pools of plasma derived from healthy donors, as well as limited 183 peptide detectability. 184 185 Applying IP-MS analysis has been reported to improve the sensitivity of protein quantification [8, 12, 186 25]. Hence, IP-MS may also be used to detect lower abundant proteins, adding to those that presently 187 remain more challenging for other MS protocols. Our investigations lead to the identification of 9 188 extracellular proteins (e.g. CXCL8, TGFA) and 15 cellular proteins (e.g.TP53, CASP2) in plasma, that 189

ON-target enrichment
were not listed in the plasma PeptideAtlas at the time of our study Supplementary Excel Sheet: 190 "Antibodies experim. annotation". For almost 50 % of the polyclonal antibodies annotated as  or CO-target in our studies, the identified peptides aligned with the sequence of the protein fragments 192 used to generate the antibodies or Protein Epitope Signature Tags (PrEST) (Supplementary Fig.1B). 193 As previously discussed [26], affinity purification of polyclonal antibodies may bare the risk of co-194 eluting the target used as bait on the columns and detecting these by MS. Such passenger proteins or 195 peptides may consequently simulate the enrichment of an endogenous target. In our case, this would 196 lead to a false classification of the antibody and may hamper downstream applications. To address this 197 concern, we analyzed beads with immobilized antibodies for the presence of the protein fragment 198 antigen that could cause passenger peptides to appear. Out of 42 tested antibodies (among  or CO-target), only 11 showed the potential presence of passenger proteins (Supplementary note 4, 200 Supplementary Fig. 6). 201 These binders respectively enriched additional members of their protein family, CCL18 and 211 SERPINA6, both sharing sequence homology with the intended target (Supplementary note 7). The 212 proteins CCL16 and CCL18 are estimated to be present at 29 ng/ml and 2 ng/ml levels in human 213 plasma (Peptide Atlas, 2017), hence the estimated abundance differs from the degrees to which 214 HPA042909 captured CCL16 (z = 10.7) and CCL18 (z = 10.1). Both proteins were otherwise rarely 215 observed in other pIPs (ƒ < 4%), have not been predicted to interact directly (Supplementary Fig. 7A) 216 but share sequence to a 27% homology (Supplementary Fig. 7B). In the case of HPA002869, the 217 antibody enriched SERPINA4 (z = 8.2) and SERPINA6 (z = 8.1) in heated plasma. Serpins are a large 218 family of blood proteins highly homolog proteins and no direct integration has been predicted 219 consistent identifications (identified in several replicates) were found for LFQ precursor intensities 257 above this level (Supplementary Fig. 10). One such example is given by the pIP using three 258 antibodies raised against the insulin growth factor binding protein 2 (IGFBP2: HPA077723, 259 HPA045140, HPA004754) of which the latter two were raised against the same antigen. As shown in 260 [c] = 11.0 µg/ml; ƒ = 18%;) and the deoxyribose-phosphate aldolase (DERA: [c] = 0.5 ng/ml; ƒ = 7%) 264 were detected. For BCHE and IGF1, an interaction was indeed previously hypothesized [28,29]. For 265 the third binder (HPA004754), IFGBP2 was only enriched upon prior heat treatment plasma (Fig. 3B). 266

CO-target enrichment and sub-categories
This differential performance of antibodies raised against the same antigen (HPA045140, 267 HPA004754) confirms the necessity to investigate each of the different batches and lots of polyclonal 268 antibodies. 269

270
In order to provide further evidence for an interaction between the identified proteins, we conducted 271 multiplexed sandwich assays. Here, recombinant IGFBP2, IFG1, IGF2 and BCHE were first analyzed 272 in a concentration dependent manner to confirm assay functionality and target specificity (Table 2). 273 First, EDTA plasma was analyzed in a concentration dependent manner, confirming the selectivity of 274 the matched antibody pairs (Supplementary Fig. 11). Then, we investigated if antibody pairs with 275 different selectivity revealed plasma concentration dependent results. As shown in Fig 4, we found 276 pairs of antibodies mixed specificity in the following capture-detection configurations: IGFBP2-IGF2, 277 IGF2-IGFBP2 as well as BCHE-IGFBP2. For IGF1 and IGF2 antibody pairs, it was not possible to 278 obtain a dilution curve with the respective recombinant proteins in solution, but they were functional 279 in plasma ( Supplementary Fig. 11 C,D,M,N,and P)). Also, IGF2-IGFBP2 and IGFBP2-IGF1 280 confirmed the presence of the previously known complex IGFBP2-IGF2 ( Table 2, Fig 4 B-C). 281 Antibody pairs for IGFBP2 and BCHE described a sample dilution depended trend with their 282 corresponding intended recombinant proteins as well as in plasma ( Table 2 and Supplementary Fig.  283 11 A, E,F,H and I)). Since no cross-reactivity was observed towards these two proteins with other 284 antibodies in the assay (Table 2), the functional antibody pair BCHE-IGFBP2 supports the indications 285 provided by IP that a physical interaction between the two proteins in plasma (Fig 4A). An inverted 286 configuration IGFBP2-BCHEand an assay including also the other IGFBP2 antibody indicating an 287 interaction (HPA077723, Fig. 3B) would further support this observation. However, we acknowledge 288 that not all antibodies allowed building mixed sandwich pairs and using the chosen assay protocol, and 289 indeed, HPA004754 and HPA077723 were raised against two different regions of IGFBP2. This could 290 explain their different performance as capture and detection, above all in the presence of complexes 291 containing IGFBP2. HPA004754 was though functional as capture and detection antibody both using 292 heat-treated and untreated plasma for the detection of IGFBP2 and in combination with anti-BCHE 293 All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/158022 doi: bioRxiv preprint first posted online Jun. 30, 2017; ( Supplementary Fig. 11. B and G). HPA077723 was though not functional with anti-BCHE either as 294 a capture or detection antibody, suggesting that the binding of either antibody hinders the other 295 antibody binding to IGFBP2-BCHE complex. Further investigations are needed to investigate if this 296 hindrance is due to a proximity of the two binding sites or other steric effects such as epitope 297 accessibility of a captured complex. 298 More examples of potential protein interactions are provided by those cases in which multiple 299 antibodies were raised against the same target protein using different antigens. Enriching common 300 ON-target and CO-targets provides evidence for protein complexes rather than artefacts. Examples are 301 antibodies targeting FBLN1, or IGF1R (see Supplementary Excel Sheet: "Antibodies against same 302 protein"). 303 OFF-target enrichment category 304 As a last category, we investigate off-target enrichments. Here, we see the plasma abundance of the 305 off-target over the intended analyte as a main reason for failing to enrich the expected target. As the 306 community is starting to acknowledge the fact that the performance of antibodies is indeed sample 307 context and application dependent, certifying off-target interactions may still allow generating novel 308 hypotheses given that these are followed-up and thoroughly validated by appropriate targeted analysis. 309 One example for selective off-target interactions in plasma is presented by the antibody HPA004920, 310 raised against MMP1 (Matrix metalloproteinase 1). We classified this antibody as OFF-target in 311 untreated plasma because it enriched Mannose-binding protein C (MBL2; z = 8.3; ƒ=12%) as well as 312 MMP3 (z = 6.8; ƒ=6%) in the IP assays ( Figure 2C). As described above for CCL16 and SERPINA4, 313 also here a 53% sequences similarity between the intended target (MMP1) and the off-target (MMP3) 314 ( Supplementary Fig. 7F) exists and an interaction between these two proteins has been predicted 315 ( Supplementary Fig. 7E). The other off-target MBL2 and MMP1 have only a 10% sequence 316 similarity (Supplementary Fig. 7G). MBL2 is though estimated to be present at 1.7 µg/ml in 317 circulation, it is almost 1000x more abundant than MMP1 ([c] = 1.1 ng/ml) and MMP3 ([c] = 0.5 318 ng/ml) [14]. MBL2 has also been described in to reside in a complex with the serine protease MASP 319 (MBL-associated serine protease) [

324
In summary, this study describes a resource that was built from our interest in the verification of 325 antibodies selectivity in plasma. The antibodies analyzed in this study include polyclonal and 326 monoclonal that was used for exploratory bead arrays and the development of immunoassays, where 327 either heat-treated or untreated plasma may serve as samples. We have conducted > 400 IP assays in 328 plasma and built a library of proteins with their frequencies of identification in plasma. Constructed on 329 All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/158022 doi: bioRxiv preprint first posted online Jun. 30, 2017; the systematic analysis of 157 antibodies, we described the occurrence of common proteins, denoted plasma background, which allowed us to determine the selectively captured endogenous plasma 331 proteins by mean of z-scores analysis. Our approach, which we also compared with Western blot 332 (Supplementary note 5 and Supplementary Table 1 (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.  (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.

21.
Chen Y, Mao H, Zhang X, Gong Y, Zhao N. Thermal conformational changes of 461 bovine fibrinogen by differential scanning calorimetry and circular dichroism. Int J Biol Macromol. (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/158022 doi: bioRxiv preprint first posted online Jun. 30, 2017; Figure 1A: Workflow and study overview: A set of 157 antibodies targeting 120 genes were covalently coupled to magnetic beads and incubated one-by-one with EDTA plasma.

FIGURES & TABLES
Two to four replicate incubations were performed for each antibody. Following target enrichment, washing and digestion on beads, the obtained data files from LC-MS were searched and normalized by MaxLFQ, z-score analysis was performed to rank proteins specifically enriched by each antibody. Using the resource generated by > 400 IP assays, antibodies were classified based on their enrichment profiles: (1) ON-TARGET, only the target protein was enriched showing a z-score ≥3; (2) CO-TARGET, the target protein was enriched together with other proteins also associated to a z-score ≥3; (3) OFF-TARGET, only proteins other than the expected target were enriched; as well as (4) NO-TARGET, in case no protein was enriched (z-scores < 3).
All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/158022 doi: bioRxiv preprint first posted online Jun. 30, 2017; (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/158022 doi: bioRxiv preprint first posted online Jun. 30, 2017; Figure 1C: Distribution of frequencies of identification. The proteins obtained from assays conducted in heat-treated (red) vs untreated plasma (black) were collected in terms of the number of times they were observed in the IP-MS data. For both sample types, the majority of the 1313 proteins were found in less than 20% of the IPs.
All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
Text: expected target and hypothesized off-targets or interactors. A complete list protein identified and relative z-scores are available in Supplementary Excel Table, Sheet "z-score> 2.5".
All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. Commercial antibodies targeting known plasma proteins of different abundance (C2, CCL16, CST3, IL1A, KLK3, LCN2) were used to establish optimal experimental conditions for volume of plasma and number of beads coupled to antibody. "p"=volume of plasma in µL; "b"= number of coupled beads (500000b = 1.6 µg of antibody).
All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/158022 doi: bioRxiv preprint first posted online Jun. 30, 2017; targeting the plasma proteins of C2, CCL16, CRP, CST3, NGAL and KLK3 (PSA) was analyzed using the pilot library limited to only these 21 IPs (left column,"pilot") as well as the library of more than 400 IP (right). In this experiment, 100 µl of plasma and 500,000 beads were applied.
All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/158022 doi: bioRxiv preprint first posted online Jun. 30, 2017; Supplementary Figure 8. Bar plot representation of the number of proteins to which a zscore ≥ 3 was assigned. Values of z-scores were rounded. All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/158022 doi: bioRxiv preprint first posted online Jun. 30, 2017; Supplementary Figure 9: Bar plot representation of proteins identified as co-targets for more than 3 antibodies. All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.  (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
2 Human EDTA plasma from a pool of individuals (50% females) was obtained from Seralab (Sera 3 Laboratories International Ltd). Aliquots of plasma (0.5 mL) were stored in cryogenic vials at -80ºC 4 and thawed at 4ºC before use. 5 (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.

Target selection
The copyright holder for this preprint . http://dx.doi.org/10.1101/158022 doi: bioRxiv preprint first posted online Jun. 30, 2017; ammonium bicarbonate (Sigma-Aldrich) and 0.25% sodium deoxycholate (Sigma-Aldrich). Proteins 37 were reduced with 1 mM DTT (Sigma-Aldrich) at 56 ºC for 30 min, and alkylated by iodoacetamide 4 38 mM (Sigma-Aldrich), at RT in the dark for 30 min. Alkylation was quenched adding 1 mM DTT. 39 Proteins were digested using a mixture of Trypsin and LysC at 1:25 trypsin-to-protein ratio (Promega, 40 USA) overnight at 37 ºC. Enzyme inactivation and sodium deoxycholate precipitation was obtained 41 adding 0.005% TFA. Peptides in the supernatant were then separated from beads, dried and re-42 suspended in solvent A containing 3% acetonitrile (ACN) and 0.1% formic acid (FA). 43 Ig gamma chain C region from Oryctolagus cuniculus (Rabbit)) because known to be in the dilution 68 buffer andImmunoglobulin variable chains belonging to Homo sapiens (Human), and excluded from 69 the z-score calculations. We considered missing values as missing not at random (MNAR) [5], but 70 missing because of concentrations below the limit of detection (LOD). We therefore used min = 0 as 71 All rights reserved. No reuse allowed without permission.

LC-MS/MS.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/158022 doi: bioRxiv preprint first posted online Jun. 30, 2017; and standard deviation for each protein identified over the population of all experiments, missing 73 values of LFQ intensities were substituted to 1 to allow for log10-transformation before two ways 74 hierarchical clustering and principal component analysis. For z-scores calculation, when duplicate and 75 triplicate experiments were available, we considered only proteins identified in all replicates for 76 further analyses, and calculated average of LFQ intensities Proteins were considered enriched when 77 associated to a z-score ≥ 3. To visualize the enriched proteins for each antibody, z-scores and LFQ 78 intensity values were use and proteins found above the set threshold were annotated accordingly. 79 Raw data produced to assess experimental conditions were analyzed using MaxQuant but excluding 80 the function for LFQ..  Table.  95 Plasma (EDTA Seralab, LOT#BRH1147432) was thawed on ice and centrifuged for 1 min at 2000 96 rpm, and diluted from 1:20 following 4-fold dilutions in PVX casein (PVXC) buffer 10% rIgG. The 97 dilution series consisted of 6 points in duplicate and were heated at 56°C for 30 min. Then, the plasma 98 was incubated with the SBA overnight. The same procedure was carried out with non-heated plasma 99 dilution series. 100 The recombinant proteins used were IGFBP2 and IGF1 were a kind gift from Hanna Tegel and Johan 101 Rockberg (AlbaNova University Center, KTH), IGF-II (R&D systems, catalog # 292-G2-050, lot 102 DS2416011) and BCHE (DuoSet kit R&D systems, Catalog # DY6137-05, lot # 1387842). The 103 dilution series of the proteins in buffer (PVXC 10% rIgG) consisted of 7 points prepared in duplicate 104 and heated at 56°C for 30 min before incubation with the SBA overnight. Non-heated protein dilution 105 series were also tested. The standard curves comprised different concentration ranges depending on 106 All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.  (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/158022 doi: bioRxiv preprint first posted online Jun. 30, 2017; targeting the plasma proteins of C2, CCL16, CRP, CST3, NGAL and KLK3 (PSA) was analyzed 137 using the pilot library limited to only these 21 IPs (left) as well as the library of more than 400 IP 138 (right). In this experiment, 100 µl of plasma and 500,000 beads were applied. 139 140 141 142 All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . http://dx.doi.org/10.1101/158022 doi: bioRxiv preprint first posted online Jun. 30, 2017; We observed that independent IPs performed in the same batch clustered together (Supplementary 144 Figure 3A). Parameters varying between independent batches IPs include different lots of reference 145 plasma, trypsin, and analytical columns. Long-term drift in instrumental response, sample handling 146 and sample heat-treatment may also add additional variability. We found that the main difference 147 between the different assays was due to heat treatment of the samples. Indeed, despite experimental 148 batches, assays using either heat treated or untreated plasma clustered together (Supplementary 149 Figure 4A-B). For this reason, we decide to analyze IPs with heat treated and not heat-treated plasma 150 separately, in order to compare the enrichment profiles from the IPs with similar background. 151 152 153 All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.

198
In plasma proteomics, current options for assessing the selectivity of antibodies can be offered by 199 paired antibodies, protein arrays [20] or Western blot (WB) [21]. Even if the three methods are each 200 valuable in evaluating antibodies quality, validation of a previous discovery or qualification prior to a 201 planned assay should be performed by a method resembling as much as possible the experimental 202 conditions of the previous or intended application. 203 For both protein arrays and WB, the setup is that a surplus of antibodies is diluted in a solution and 204 applied onto supports that present the antigens. Hence, there is generally minimal competition for 205 binding sites between potential on-and off-targets as compared to antibodies being immobilized and 206 applied to a complex solution. For an example given by anti-IL6R binder, we found application 207 dependent recognition for five different antibodies. All five were classified as target-specific using 208 protein arrays, however only three detected IL6R in plasma (Supplementary Excel Table). were depleted of albumin and IgG prior use [22]. Nevertheless, we found that the assessment of 13 out 218 of 104 antibodies (12%) provided supportive evidence by both methods. For antibodies raised against 219 plasma proteins, the success rates for plasma IP (54%) was though higher than for plasma WB (32%) 220 (Supplementary Table 1). When considering cellular proteins, the success rates were the same 33%. 221 For WB however, uncertainty will remain unless other standards or comparative analyses are used to 222 reference the detected bands. Until then, bands detected at the predicted molecular weight may still 223 represent the recognition of an off-target molecule. 224 Consequently, pIP provides an unequivocal identification of the target and could elucidate ambiguous 225 WB results (see C1orf64, CEP162, E2F7,and CCL16 Supplementary Excel Table, sheet: 226 "Antibodies_experim_annotation"). While WB may provide a more accessible technology for some 227 labs, the information added by identification in MS will be required for an in-depth analysis of 228 antibody selectivity. IP-MS can here be one option to enhance our understanding of binding to 229 proteins, in particular for plasma. 230 231 All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint . (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.  All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.  (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint .  "Go enrichment analysis": Differential GO analysis using TOPP cluster between frequent (> 277 20%) and less frequently occuring proteins (< 20%). 278  "z-score >2.5": List proteins for which a z-score > 2.5 was calculated containing targets, off-279 targets and potential interactors. 280  "List of Peptides": Peptides identified in pIP assays for the expected target proteins and 281 overlapping with the antigen used to generate the antibodies. 282  "Experimental Batches": description of major differences in the experimental conditions for 283 the 6 experimental batches analyzed. 284  "Reagents Catalogs_Lot_numbers": List of catalog and lot numbers for affinity reagents. 285  "Antibodies against same protein": proteins enriched by antibodies raised against the same 286 target protein, using same or different protein fragments. 287 288 289 All rights reserved. No reuse allowed without permission.
(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.  (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.