Non-specific interactions of antibody-oligonucleotide conjugates with living cells

Antibody-Oligonucleotide Conjugates (AOCs) represent an emerging class of functionalized antibodies that have already been used in a wide variety of applications. While the impact of dye and drug conjugation on antibodies’ ability to bind their target has been extensively studied, little is known about the effect caused by the conjugation of hydrophilic and charged payloads such as oligonucleotides on the functions of an antibody. Previous observations of non-specific interactions of nucleic acids with untargeted cells prompted us to further investigate their impact on AOC binding abilities and cell selectivity. We synthesized a series of single- and double-stranded AOCs, as well as a human serum albumin-oligonucleotide conjugate, and studied their interactions with both targeted and non-targeted living cells using a time-resolved analysis of ligand binding assay. Our results indicate that conjugation of single strand oligonucleotides to proteins induce consistent non-specific interactions with cell surfaces while double strand oligonucleotides have little or no effect, depending on the preparation method.


Results and discussion
Using a previously reported plug and play conjugation strategy 7 , we prepared several fluorescein-labelled proteins and protein-ON conjugates (see Fig. 1, 5, S1, S2 and Table S1 for syntheses and characterization of the conjugates) with Degrees of Conjugation (DoC) in line with those of our previously-described DNA-linked ADCs 2 (i.e. comprised between 2 and 3). We then evaluated the interaction profiles of these protein-ON conjugates with two cell lines, SK-BR-3 (HER2 + ) and MDA-MB-231 (HER2 − ), using time-resolved analysis of ligand binding assay 8,9 . As an HER2 targeting antibody, we used trastuzumab and as negative controls, we used the anti-CD20 antibody rituximab and Human Serum Albumin (HSA).
In the following figures, we will report association rate constant values k a (M −1 s −1 ), describing the rate of formation of complexes, i.e. the number of fluorescein-labelled compounds bound to cell membranes per second. Thus, high k a value will account for fast binding to cell surface, while low value will account for slow to no interaction. The values of k a in biological systems are typically comprised between 1.10 3 and 1.10 7 M −1 s −1 . We chose to study this parameter to compare specific and non-specific interactions as it was previously shown to be largely insensitive to differences between binding patterns 10 .
In a first set of experiments, we compared the association rate constants of anti-HER2 antibody trastuzumab (T*), trastuzumab-37mer ssON conjugate (T*-ssON), and single-stranded 37mer ssON (ssON*), on both cell lines (Fig. 2). In order to do so, cells were seeded on a cell dish and incubated with fluorescein-labelled compounds (the symbol * indicates fluorescein labelling). The fluorescence intensity was then measured over time and normalized against the background fluorescence of the plastic support (see Fig. S3 to S16). The signal increase was used to extract the association rate constant k a .
Unsurprisingly, trastuzumab displayed a high selectivity profile with an association rate constant k a of 23,800 M −1 s −1 on the HER2 + cell line, while no signal variation was detected for the HER2 − cell line. T*-ssON showed a deteriorated selectivity profile with a reduced k a for HER2 + cells, but more importantly with the appearance of "non-specific" interactions with HER2 − cells. Stochastic conjugation with lysine residues 11,12 for T*-ssON might account for the lower k a with the HER2 + cell line as compared to unconjugated T*. Indeed,   www.nature.com/scientificreports/ random conjugation of payloads, including dyes 11 and small-molecule drugs 12,13 , to an antibody was reported to deteriorate its antigen affinity in some cases. It is noteworthy that ssON* showed an HER2-independant interaction with both cell lines, as evidenced by low but significant values of k a on both SK-BR-3 and MDA-MB-231 cell lines. The order of magnitude of these interactions falls in the same range as that of T*-ssON on HER2 − cells, advocating for the fact that the latter was the result of ON-cell interactions.
To validate this observation, we conjugated the same single-stranded 37mer ON to rituximab (R), an antibody targeting B-lymphocyte antigen CD20 (an antigen that is expressed by neither SK-BR-3 nor MDA-MB-231), and to HSA (Fig. 3).
As expected, both non-conjugated rituximab (R*) and HSA (HSA*) showed no interaction with either cell line. On the other hand, the corresponding ON conjugated R*-ssON and HSA*-ssON were shown to interact to an undeniable extent with both cell lines (Fig. 3), which is consistent with our previous observation that the conjugation of an ON to a protein induces an interaction with cells that is mediated by the ON moiety and not the protein. Interestingly, this effect was more pronounced with R*-ssON than HSA*-ssON, indicating that different proteins might be impacted differently by ON conjugation.
To gain further evidence, we performed a competition assay, where MDA-MB-231 cells, supposedly interacting with T*-ssON via non-specific interactions, were pre-incubated with a 100-fold excess of free 37mer ssON, relative to T*-ssON (Fig. S17).
As expected, we found that the addition of free ssON to the medium almost completely prevented the further association of T*-ssON with MDA-MB-231, suggesting a shielding effect from the free ssON.
We then evaluated the influence of the ONs' length and hybridization on the association rate constant by comparing single and double stranded forms of non-coding and non-structured 20mers, 37mers and 74mers. In all cases, weak interactions with both cell lines were observed, with consistently higher values for single-stranded species ( Fig. 4 and SI).
In the context of drug delivery applications of AOCs 2,14-20 , the ON component is often hybridized with its complementary strand in the form of double-stranded ONs (dsONs). As this could lead to weaker interactions with cell membranes, based on the results in Fig. 4, we set to prepare a dsON version of our trastuzumab conjugate (T*-dsON) in order to evaluate the effect of such structural change. A first method to prepare this conjugate consists in the hybridization of a complementary ssON (cON) to the previously synthesized T-ssON. This is typically done by a brief incubation at 37 °C of the two partners, in order to prevent degradation of the antibody (method 1, Fig. 5). This approach is mostly employed for non-covalent conjugation 21 between two molecules that had been separately conjugated with complementary ssONs 2,20,22,23 . The validity of this approach is supported by the many examples of immunoassays (e.g. immuno-PCR 24,25 , proximity extension assay 26,27 , protein arrays 28 ), which rely upon hybridization steps that proceed under similar conditions. A second method consists in hybridizing the two strands under classical conditions (i.e. by incubation at 95 °C) prior to the bioconjugation step (method 2, Fig. 5). This method has been reported for the synthesis of antibody-siRNA conjugates from commercial chemically-modified duplexed siRNAs 14,16 . We produced T*-dsON conjugates with identical DoC values of 2.9 (determined by SDS-PAGE gel analysis; see Fig. S1) using both methods and compared their interaction rate constants to those of T*-ssON on both cell lines (Fig. 6). www.nature.com/scientificreports/ For the T*-dsON conjugates prepared following the first method, switching from single to double-stranded ONs gave comparable non-specific interactions, and also resulted in a slight decrease in k a , notably on the HER2 + SK-BR-3 cell line. Interestingly, the conjugate prepared following the second method had similar k a with SK-BR-3 cells but did not display non-specific interactions with MDA-MB-231 cells.
In the polymerase chain reaction (PCR), full hybridization of ONs is highly dependent on temperature, and the initial denaturation step is typically performed at 94 °C 29 . Incubation at 90-95 °C is thus commonly used for the hybridization of complementary ssONs into dsONs, as in the case of method 2 (Fig. 5). Performing the hybridization at lower temperature is useful when working with protein-ssON conjugates, since they are prone to undergo thermal denaturation, but it might not be sufficient to reach full hybridization (method 1, Fig. 5). This could account for the interaction profile of the T*-dsON conjugate prepared by method 1 which is halfway between that of the fully hybridized T*-dsON, prepared by method 2, and that of T*-ssON.
The association of naked ONs with cells surface has now been studied for more than 50 years, with many cellsurface proteins proposed as receptors 4 . It has been documented that various types of ONs might bind to different receptors. As an example, toll-like receptors (TLRs), involved in the innate immune response, possess the ability to bind DNA molecules containing CpG motifs, dsRNAs as well as ssRNAs 30 . Cell-surface receptors of the scavenger receptors family, such as stabilin, have been reported to bind and internalize phosphorothioate-modified oligodeoxynucleotides 31 , despite some conflicting results having been published 32 . Furthermore, proteins of the . Association rate constants (k a ) of free 20mer, 37mer, and 74mer, single and double-stranded ONs (respectively ssON20*, ssON37*, ssON74*, and dsON20*, dsON37*, dsON74*) on SK-BR-3 (HER2 + cell line, red bars) and MDA-MB-231 (HER2 − cell line, blue bars) measured using the time-resolved analysis of ligand binding assay (Fig. S11-16). The symbol * indicates fluorescein-labelling. Error bars indicate the standard error of the fitted parameter k a .  Our results show that, when conjugated to an antibody, ONs are not simple linkers nor spectator payloads. Based on the present work and this body of literature, our group is currently investigating further the mechanism by which these interactions between AOCs and cell membranes operate at the molecular level and can be controlled.

Conclusion
As AOCs are developing into powerful tools in various applications 1 , investigations to get a better understanding of their interactions with cell surfaces appear to be stimulating a renewed interest 30 . Our previous 2 and present results show that ONs are a particular payload that may display weak but consistent interactions with cell surfaces, which can impact the binding properties of antibodies upon conjugation. We demonstrate that both the nature of the ON, single strand vs double strand, as well as the method used to prepare the dsON AOC have a clear impact on the non-specific interaction of the resulting conjugates. This phenomenon is likely to disturb the in vitro and in vivo behavior of AOCs and influence their fate beyond what can be extrapolated from the knowledge of classical protein conjugates. As such, it appears that both ON structure and preparation method should be taken into consideration when developing antibody-oligonucleotide conjugates for imaging, detection or therapeutic application.

Conjugates synthesis.
Oligonucleotide functionalization. BCN-PEG 6 -PFP (1) was synthesized as previously described 7 . In a 2 mL Eppendorf tube, 5′-amino-modified oligonucleotide (1 equiv., 50 µL, 1 mM in water) was combined with 1 (20 equiv., 50 µL, 20 mM in DMSO) and NaHCO 3 (100 equiv., 5 µL, 1 M in water). The mixture was incubated at 25 °C overnight. The mixture was then diluted with water to a final volume of 300 µL and added with acetone (900 µL) and LiClO 4 (20 µL, 3 M in water) in order to precipitate the oligonucleotide species. The sample was then centrifuged (15,000 G, 8 mn) and the supernatant was discarded. The precipitate was dissolved with water (300 µL) to repeat the precipitation and centrifugation procedure a second time. www.nature.com/scientificreports/ Oligonucleotide purification. The previously obtained precipitate was then dissolved with water (100 µL) and purified by HPLC (detection at 260 nm, mobile phase gradient A/B 9:1 to 6:4 in 30 mn). After lyophilization, the ON conjugate was dissolved in DPBS (1x, pH 7.4) and analyzed by absorption spectrophotometry (measured at 260 nm using a Nanodrop) to calculate the solution's concentration using Beer-Lambert's law.
Double stranded oligonucleotides and antibody-oligonucleotide conjugates synthesis. Free oligonucleotides were hybridized by stirring a solution of the complementary strands at an equimolar ratio (100 µM) in DPBS 1 × at 95 °C for 5 mn, and then allowing the solution to come back to room temperature. The hybridized species were then purified from the non-hybridized ones using AKTA Pure System (isocratic elution with DPBS (1x, pH 7.4), 0.5 mL/min).

Conjugates' characterization.
Protein-oligonucleotide conjugates concentration determination. The concentration of a protein in a given solution can usually be determined by measuring its absorption at 280 nm, and using Beer-Lambert's law. ON's absorbance at 280 nm thus makes it impossible to determine the concentration of protein-oligonucleotide conjugates through absorption spectrophotometry measurement. Protein-oligonucleotide conjugates' concentration was then determined using Pierce BCA protein assay kit (ThermoFisher ref 23225), following the manufacturer's protocol. This method allows quantification of the protein moiety's concentration, regardless of the presence of conjugated ONs. Concentrations were used to calculate the yields of conjugation (see Table S1). www.nature.com/scientificreports/ Additionally, we analyzed the deglycosylated azido-modified trastuzumab intermediate by native mass spectrometry (see figure S2). As observed in a previous work 7 , the mean DoC values obtained by integration of SDS-PAGE gel bands of the Ab-ssON conjugate (2.9) and native MS of the azido-modified intermediate (2.8) were closely correlated.

Protein-oligonucleotide conjugates DoC distribution determination by SDS
Proteins and protein-oligonucleotide conjugate degree of labelling (DoL) determination. After FITC-labelling, the fluorescein concentration of each protein and protein-ON conjugate was measured by absorption spectrophotometry using NanoDrop's "proteins and labels" mode.
The DoL of each compound was calculated using Eq. (2) (see Table S1).
For fluorescein-labelled proteins, the protein concentration was determined by absorption at 280 nm, while for fluorescein-labelled protein-ON conjugates, it was determined by BCA assay (see above). The day before the experiment on Ligand Tracer Green (Ridgeview Instruments), cells were seeded as 600 μL droplets with 8 × 105 cells/mL near the edge in 87 mm cell culture treated dishes (Greiner, Frickenhausen, Germany) and incubated at 37 °C overnight. Two droplets were prepared with SK-BR-3, one with MDA-MB-231, and one was left as a background reference (plastic).
Prior to kinetic measurements, the medium was carefully removed and 3 mL of fresh complete medium was added to the dish.
Time-resolved analysis of ligand binding assays. We measured the interactions of fluorescein-labelled proteins and oligonucleotides with living cells in real-time using LigandTracer Green (Ridgeview intruments).
The Petri dish on which cells were grown (see above) was placed on the inclined rotating support of the instrument, with the Blue/Green detector placed on its upper part.
First, a baseline signal was collected for 30 min. The fluorescein-labelled compound was then added in two steps at increasing concentrations (10 and 30 nM). Inclination of the Petri dish allows for the addition of the fluorescein-labelled compounds outside of the detection area. For each rotation of the Petri dish, the signal from the three areas containing cells (two spots for SK-BR-3, one for MDA-MB-231) and a background reference area (plastic) is recorded. Measurements last 30 s each, with 5 s in between each of them to allow the medium to sit in the lower part of the Petri dish. This results in three background-subtracted real-time binding curves, which represent the binding of the fluorescein-labelled compound to each cell-containing area.
For each concentration the incubation was performed until a sufficient curvature was obtained for the subsequent extraction of kinetic parameters. Dissociation of the ligand was recorded after replacing the incubation solution with 3 ml of fresh medium.
Signals from cell and reference areas are recorded during every rotation, resulting in a background-subtracted binding curve. Binding traces were analyzed with the evaluation software TraceDrawer 1.8.1 (Ridgeview Instruments) in order to determine k a according to the Langmuir, or "one-to-one", binding model.
Competition assay. After collection of the baseline signal, 100 equiv. of unlabelled ssON37 (relative to the total added amount of T*-ssON) were added to the medium. After 30 mn, the time-resolved analysis of T*-ssON binding was performed as previously described (Fig. S17).