Castillo, 1994; H.-S. Wang, Pan, & Wang, 2005). As depicted, the same electrodes used to assemble the sensing components are used for electrochemical detection of the antibodies. As such, there are many advantages associated with the coupling of electroassembly and electrochemical readouts (J. Li, Maniar, et al., 2019;J. Li, Wu, et al., 2019;Liu et al., 2010), a most simple example being that the size of the electrode can be used to vary the sensor binding capacity and signal strength. The methodology is simple, portable, inexpensive and can provide fast, quantifiable results for a variety of applications, including bioprocessing applications, noted here.
2.1 Materials. 4-arm PEG thiol was purchased from JenKam (Plano, TX). RCA120 and fluorescein labelled conjugate (FITC RCA120) were purchased from Vector Laboratories (Burlingame, CA). 2-Thioethyl β-Galactopyranoside (thiolated Galβ) was purchased from Sussex Research Laboratories (Ottawa, ON Canada). Immunoglobulin (IgG) from human serum, SILU Lite SigmaMAb, and protein G:HRP conjugate were purchased from Millipore Sigma (Burlington, MA). 1,1′-Ferrocenedimethanol (Fc) was purchased from Santa Cruz Biotechnology (Dallas, TX). Dulbecco's Modified Eagle medium, high glucose, HEPES, no phenol red (DMEM) was purchased from Thermo Fisher Scientific (Waltham, MA).
A mixture of Fc (5 mM) and PEG thiol (50 mg/mL) was first prepared in phosphate buffer (0.1 M, pH 7.0). The surface of a 2 mm diameter gold standard electrode (working electrode) was fully immersed in the solution along with a platinum wire (counter electrode) and an Ag/AgCl reference electrode. PEG electrodeposition occurred for 1 minute at a constant potential of 0.4 V. After PEG hydrogel formation, remaining sulfhydryl groups were "activated" to sulfenic acids by immersing the surface in a solution of Fc (5 mM). A constant voltage of 0.4 V was applied for 2 minutes to ensure maximal sulfenic acid group
were immersed in a phosphate buffered solution (0.1 M, pH 7.0) containing K 3 Fe(CN) 6 / K 4 Fe(CN) 6 (1 mM). CV scans were performed from 0 to 0.5 V at a scan rate of 0.1 V/s. EIS measurements were taken over a frequency range of 100 kHz to 1 Hz at a potential that was the average of the reduction and oxidation peak potentials. Fluorescence microscopy images were then taken with an upright microscope with exposures of 90 ms (titer detection interface) and 50 ms (galactosylation detection interface).
2.5 Evaluation of hydrogel and interface reproducibility. PEG hydrogels were electrodeposited and activated as in Section 2.2. Charge profiles were measured during hydrogel electrodeposition. After electrodeposition and "activation", CV scans were performed from 0 to 0.5 V at a scan rate of 0.5 V/s in a solution of Fc (0.5 mM, diluted in phosphate buffer). For interface reproducibility, titer and galactosylation interfaces were individually assembled with FITC cysteinylated protein G (250 μg/mL) and FITC RCA120 (500 μg/mL) as in Section 2.3.
Fluorescence microscopy images of the individual electrodes were taken with an upright microscope using exposure times of 20 ms (galactosylation detection) and 400 ms (titer detection). ImageJ was used to calculate the mean grey intensity values of the images.
response from the interfaces were performed with the 3-electrode set-up (same as above) using a CHI1040C electrochemical analyzer (CH Instruments). The electrodes were immersed in buffered solution (0.1 M phosphate, pH 7.0) containing Fc (0.5 mM) which was mechanically stirred at a constant rate with an applied potential of 0 V. To achieve a response from the interfaces, 25 µM H 2 O 2 was added to the mixture and the increase in current from the baseline was recorded. The lower limit of detection for the titer detection interface was calculated as previously described (Armbruster & Pry, 2008).
Evaluation with culture media. Titer and galactosylation detection interfaces were assembled as in Section 2.3. Titer detection interfaces were incubated in IgG from human serum
(0 -1 g/L) that had been diluted with 0.1 M PBS or with DMEM. Galactosylation detection interfaces were incubated in SigmaMAb (1 g/L) diluted with 0.1 M PBS or with DMEM. We chose to use this antibody standard as it has a known concentration of galactosylated IgG.
Interfaces were then incubated in protein G:HRP (0.1 g/L). All incubations occurred at room temperature for 1 hour. Between incubations, interfaces were rinsed with wash buffer 3 times.
Current measurements were taken as described above.
To evaluate regeneration, the detection interfaces were assembled with IgG from human serum (2 g/L for titer detection, 1 g/L for galactosylation detection) and protein G:HRP (0.1 g/L) and measurements of the current response were taken using the same conditions as above. The interfaces were then washed with 0.1 M acetic acid, 1 M sodium chloride, pH 2.8 (3 washes, 15 minutes each) in order to strip IgG and protein G:HRP from the surfaces, leaving only the detection proteins (i.e., protein G or RCA120) conjugated to PEG surface. Current measurements were taken to confirm that IgG and protein G:HRP were removed. The interfaces were once again re-assembled with IgG and protein G:HRP. Current measurements were taken to confirm the re-binding of IgG and protein G:HRP to the interface surfaces.
Protein G is a bacterial protein that possesses high binding affinity for the Fc domain of human immunoglobulin G (IgG) (Akerström & Björck, 1986;Björck & Kronvall, 1984;Guss et al., 1986). We used protein G for the titer detection interface as it has been well-established that it can bind the most commonly used IgG subclasses for therapeutics development, IgG1 and IgG2, in cell culture broth (Chames, Van Regenmortel, Weiss, & Baty, 2009;Lund et al., 2011). We have previously used this protein G:antibody methodology for assembly of enzymes and cells
onto various substrates (Fernandes, Roy, Wu, & Bentley, 2010;Hebert et al., 2010;H. C. Wu et al., 2017). To incorporate protein G into this assembly and detection scheme, we introduced 5 cysteine codons into the C-terminus of the structural gene and expressed and purified the protein from E. coli. In this way, because it is expressed in E. coli, the protein G was not glycosylated, which could otherwise confound subsequent antibody binding. The extra cysteine residues, in effect, become an activatable "pro-tag" and could potentially be electroassembled based on the formation of a sulfenic acid enabled disulfide bond between the PEG and the protein G (Lewandowski, Small, Chen, Payne, & Bentley, 2006). In our example, the sulfenic acid of the PEG and the sulfhydryl of the cysteine residues enable covalent coupling. In Figure S1, we show the plasmid vector (Figure S1a), the protocol for expression and purification from E. coli, and the corresponding western blot (Figure S1b). Briefly, we cultured cells in LB media, induced with IPTG at OD 600 ~ 0.4, and centrifuged the cells after they were grown overnight at room temperature. The cell pellets were then resuspended in lysis buffer, sonicated for 10 minutes, and then centrifuged to remove cell debris. The N-terminal His-tagged protein G was then purified using immobilized metal affinity chromatography (IMAC). The purified protein was confirmed by western blot and then used for assembly onto the thiolated PEG previously electroassembled onto gold electrodes.
For both titer and galactosylation detection, a PEG hydrogel serves as the initial layer of the interfaces. Figure 1a provides an illustration detailing the proposed mechanism for thiolated PEG electrodeposition. An oxidative potential (+0.4 V, 1 minute) is first applied to a bare gold electrode that is immersed in a mixture of Fc, a redox mediator, and 4-arm thiolated PEG. Fc near the surface of the electrode will become oxidized, diffuse away from the electrode surface, and oxidize the free thiol groups of PEG, converting them to sulfenic acid groups. Sulfenic acid groups will react with nearby thiol groups attached to PEG, thus establishing inter-molecular
bonds allowing a hydrogel to form (Jinyang Li et al., 2020;Raja, Thiruselvi, Mandal, & Gnanamani, 2015;Sun & Huang, 2016). The thiolated PEG directly assembles onto gold through the sulfur-gold interactions commonly used for templated assembly on gold. Mediated electroassembly using Fc is advantageous as it enables a level of control over the oxidation of the thiol groups of PEG, which allows for a significant level of consistency in the production of the hydrogels (Yi et al., 2005). That is, the electrodeposition process serves to "activate" the thiolated PEG in that the sulfenic acid groups will spontaneously bond with sulfhydryl groups forming covalent disulfide linkages (Kettenhofen & Wood, 2010;Jinyang Li et al., 2020). After PEG electrodeposition, the hydrogel is submerged in Fc solution and further oxidized for 2 minutes in order to "activate" the hydrogel (i.e., oxidize sulfhydryl groups to reactive sulfenic acid groups) to enable covalent bonding of the molecular recognition elements. For titer detection, the PEG modified electrode is then immersed in a solution of cysteinylated protein G enabling its assembly through disulfide bond formation between the cysteine tag of protein G and the sulfenic acid groups on the PEG hydrogel (upper path, Figure 1a). The subsequent topmost protein G layer serves as the detection interface for IgG. For galactosylation detection, the electroactivated thiolated PEG surface can be immersed in a solution of thiolated Galβ, which will covalently bond with the available sulfenic acid groups on the surface of the hydrogel (lower path, Figure 1a). Subsequently, immersion in a solution containing RCA120, a Galβ binding lectin, allows for bio-specific binding between RCA120 and thiolated Galβ, thus forming the galactose-detection surface of the interface.
In Figure 1b, fluorescence microscopy was used to qualitatively characterize layer formation for both interfaces. FITC cysteinylated protein G and FITC RCA120 were used to visualize the binding of the proteins to the interfaces. For the titer detection interface, the presence of a PEG film was needed to establish fluorescence upon protein G capture. That is, there was no fluorescence in the absence of PEG (layering the assemblies onto bare gold). For the This article is protected by copyright. All rights reserved.
galactosylation detection interface, (iii) a strong fluorescence signal was observed only when all the interface layers were assembled. Controls, which included: (iv) bare gold; (v) an interface with no thiolated Galβ; (vi) and an interface with no thiolated PEG, were all weakly fluorescent, as expected. That is, when a component of the completely functional interface was not present (i.e., PEG or thiolated Galβ), there was essentially no FITC RCA120 binding. It is important to note that for all interfaces, the electrodeposition of PEG is localized on the pattern of the electrode, allowing for the interface to form only on the gold surface. This is an important attribute of our process as electrochemical outputs (later) will be confined to activity that is localized directly onto the electrodes. Again, this is an feature of electroassembling functional components such as proteins, cells, and other molecules that are otherwise difficult to array with great spatial resolution (Cheng et al., 2011;H. C. Wu et al., 2009). Because of this, electrochemical systems can be designed, and outputs quantified based on electrode area. This makes for well-controlled microfluidic application of the methodology (Betz et al., 2013;Lewandowski et al., 2008;Shang et al., 2018).
Importantly, our data support the conclusion that the interfaces were successfully assembled when layers were sequentially formed, and all necessary components were present. Moreover, these data show that the thiolated PEG serves an important function as it provides a great number of binding sites (i.e., thiols) for protein assembly, enabling the interfaces' detection capabilities.
In other work, we showed that the electroactivated hydrogel properties could be extended away from the electrode (normal to the surface) the longer the voltage was applied, demonstrating programmable control of hydrogel chemistry (Jinyang Li et al., 2020). While not shown here, we expect the number of sulfenic acid residues available for coupling could be electronically programmed.
Next, cyclic voltammetry (CV) and electrochemical impedance spectroscopy (EIS) were used as an orthogonal means to provide physical evidence of interface assembly using redox probe,
Fe(CN) 6 3-/4-. For CV measurements, bare gold electrodes for both interfaces had well-defined, reversible oxidation and reduction peaks, as shown in Figure 2a-b. When the surfaces were electrodeposited with PEG, there was a clear reduction in both peak currents and an increase in the oxidation to reduction peak separation, as evidenced in the CV profiles and corresponding bar graphs. This indicates that electron transfer at the electrode surface was significantly decreased as a result of PEG hydrogel formation (Patolsky, Zayats, Katz, & Willner, 1999;Zhang, Thomas, Kim, & Payne, 2012). For both interfaces, once the subsequent protein layer was assembled (i.e., cysteinylated protein G or RCA120), the peak current responses decreased further, and the peakto-peak separation became larger, indicating that both proteins were successfully bound to the PEG-modified surfaces. EIS was also performed as it can sensitively detect changes in impedance through alterations in the surface charge of the assembled interface layers. The Nyquist plots of the titer and galactosylation detection interfaces are shown in Figure 2c-d, respectively. At higher frequencies, the spectra have a semi-circular area that is related to charge transfer resistance (R CT )
and at lower frequencies, they contain a linear portion which is related to diffusion processes. For the bare gold electrodes, the plots are essentially linear, indicating that there is minimal resistance to charge transfer (Zhang et al., 2012). The diameter of the semi-circular area of the spectra becomes larger as the subsequent protein layers are formed on the surfaces (illustrated by the dotted arrows), indicating that R CT has increased due to further impedance of electron transfer (Chang & Park, 2010;J. Li, Maniar, et al., 2019;Patolsky et al., 1999). In both cases, CV and EIS measurements reinforce our conclusion that the interface layers were successfully assembled onto the electrode surfaces.
We next wanted to characterize the uniformity and reproducibility of hydrogel electrodeposition and interface assembly. We first monitored the charge profile over time throughout individual hydrogel electrodepositions for 3 separate electrodes. The results are shown
This article is protected by copyright. All rights reserved.
in Figure 3a, where the charge is directly correlated to the number of sulfhydryl groups of PEG that are oxidized to sulfenic acid groups (i.e. electrons exchanged) (Nishiumi, Abdul, & Aoki, 2005). Across all 3 hydrogel electrodepositions, the charge profiles were nearly identical (as depicted in the bar graph insert, the average final charge transfer was 1.1 mC +/-0.004). This data indicates that the number of sulfenic acid groups formed during deposition was identical and after considering the substantial reactivity of sulfenic acid (Gupta, Paritala, & Carroll, 2016; Kettenhofen & Wood, 2010) it suggests that the degree of crosslinking among the hydrogels was also similar.
We also wanted to test whether the current measurements across different hydrogel preparations were consistent. To evaluate, the individually electrodeposited and "activated" hydrogels were immersed in a solution of 0.5 mM Fc and CV scans were performed. Results are shown in Figure 3b. Here, the CV scans of the 3 electrodes appear nearly superimposed, indicating that the current measurements are consistent for individually prepared hydrogels (i.e., electron exchange between Fc and the gold electrode was not impacted). To provide quantitative measure, we calculated the average oxidative (Q Ox ) and reductive (Q Red ) charge transferred (Jinyang Li et al., 2020) from the CV profiles and found the values to be consistent (Q Ox = 2247 μC +/-42; Q Red = 1327 μC +/-48).
Lastly, we used fluorescence microscopy to gauge whether binding of the molecular recognition elements to individually assembled titer and galactosylation interfaces was also consistent. This is important in order to ensure that the interfaces provide for reproducible measurements. Electrodes were individually electrodeposited and "activated", as in Methods. The hydrogels were then incubated in either FITC-labeled cysteinylated protein G (250 μg/mL) or thiolated Galβ (50 μg/mL) and FITC-labeled RCA120 (500 μg/mL). Fluorescence microscopy images were taken, and the fluorescence intensity of the images was analyzed using ImageJ.
Results are shown in Figure 3c. As indicated at the far right, we analyzed both the full circular
area and 5 windows within each circular electrode. The bar graphs represent the average intensities of the electrodes (T1 to T3 or G1 to G3) for either the full electrode area or smaller areas from within the electrodes (A-E). For both interfaces, the variation in the intensity across the individual electrodes (error bars depict the standard deviation) was minimal, indicating that binding of molecular recognition elements to the hydrogels was remarkably consistent across separate electrode preparations.
We next evaluated the ability of the cysteinylated protein G interface to detect antibody. The experimental process and set-up is illustrated in Figure 4a. The assembled cysteinylated protein G interfaces were incubated with serially diluted concentrations of human IgG, washed with buffer containing 0.05% Tween-20 to remove any non-specific binding, and then immersed in protein G:HRP to form a sandwich configuration. It is important to note that IgG has 2 protein G binding sites on either side of its Fc portion (located in the hinge region that connects the C H 2 and C H 3 domains)(Sauer-Eriksson, Kleywegt, Uhlén, & Jones, 1995). This allows protein G to serve as both the recognition (cysteinylated protein G) and detection element (protein G:HRP). After many experiments, we found that a 0.1 g/L concentration of protein G:HRP (diluted in PBS) was best to maximize the electrochemical responses of the assembled 3.1 mm 2 interfaces while minimizing responses from non-specific binding. The cysteinylated protein G-IgG-protein G:HRP interfaces were then submerged in a stirred solution containing 0.5 mM Fc (diluted in phosphate buffer) and with an applied potential of 0 V. The solutions were equilibrated for 200 seconds and then 25 µM H 2 O 2 was spiked into the solutions to achieve the current responses from the cysteinylated protein G-IgG-protein G:HRP interfaces, shown in Figure 4b. In an analogous manner to our studies with protein G:HRP, we had earlier performed studies with varied amounts of H 2 O 2 ; we found that 25 uM was optimal for the conditions tested. Our results similarly indicate that response time is very rapid. After the initial transient, the current at all IgG concentrations
This article is protected by copyright. All rights reserved.
reached a steady value within 2 sec, and importantly, reached a very steady value that was maintained for an additional 2 minutes. The current appeared to be related to the concentration of IgG in the incubation buffer. In Figure 4c, we indeed found that when the interfaces were incubated with IgG concentrations ranging from 15.6 -1000 µg/mL, the average of the peak current response linearly increased with the concentration of IgG (R 2 =0.98). Importantly, this linear range for the interface is appropriate for many antibody therapeutics produced in mammalian cell lines such as Chinese hamster ovary (CHO) cells (Huang et al., 2010;Rader & Langer, 2014;Shukla, Wolfe, Mostafa, & Norman, 2017). That is, the response was observed to saturate above a concentration of 1000 µg/mL of IgG (note the 1000 μg/mL vs. the 2000 μg/mL responses). Using 3 times the standard deviation of the negative control, we found that the lower limit of detection was 88 µg/mL. We note that these results are from separate measurements across separate batches of individually assembled interfaces and their linearity demonstrates that the method is highly reproducible even when using different gold electrodes. The responses from the negative controls further show that non-specific binding from protein G:HRP was minimal (Figure S2). As noted above, a low level of non-specific binding is likely attributed to the antifouling properties of PEG (Jeon et al., 1991). Overall, these data support the interface's applicability for use in monitoring titer in support of antibody production and separation operations.
The selectivity of the galactosylation detection interface for Galβ terminating oligosaccharides was then evaluated. Figure 5a depicts the experimental process and set-up, which was performed in an identical manner as the titer studies above. Solutions containing dilutions of 1 g/L of each respective antibody glycoform (i.e., antibodies with a defined N-linked glycan pattern) were incubated with the interfaces. The antibody glycoforms we used to interrogate the selectivity of the interface had glycan patterns of G0, G0F, G2, G2F, and S2G2, which are described in Table 1 Accepted Article and depicted in Figure 5b. These represent many of the common glycan structures of antibody therapeutics, which are of the complex biantennary type that contain variable additions of fucose, galactose, and sialic acid (T. Li et al., 2017;Majewska, Tejada, Betenbaugh, & Agarwal, 2020).
These antibody glycoforms were developed in our labs (Giddens, Lomino, DiLillo, Ravetch, & Wang, 2018;Yang & Wang, 2017). We chose Galβ as a model terminal glycan group for interface detection as it is one of the most predominant terminal N-linked glycan modifications (Raju, 2003). That is, several interfaces were examined with the test case being RCA120 assembled onto PEG with thiolated Galβ. Controls with different interface assemblies (no thiolated PEG, no thiolated Galβ) and different IgG (altered glycan patterns, including sialic acid capped galactose) were also evaluated. We chose to use the lectin RCA120 for galactosylation detection as: (i) it has been shown to have specific binding to Galβ and none of the other glycans and, (ii) it contains two subunits that have binding affinity for Galβ terminating oligosaccharides, making it ideal for interface assembly as one binding site is already occupied by the thiolated Galβ (Green, Brodbeck, & Baenziger, 1987;Itakura et al., 2007;A. M. Wu et al., 2006). In order to maintain consistency, concentrations of 1 g/L IgG, 0.1 g/L protein G:HRP, and 25 µM H 2 O 2 were selected as these conditions helped to maintain linearity for the titer detection interface.
Results are shown in Figure 5b. As expected, the lectin-containing interface shows the highest current for the antibody glycoforms that contain terminal Galβ (i.e., G2 and G2F). For antibody glycoforms not containing Galβ (i.e., G0 and G0F) or when capped by a different glycan, such as sialic acid (i.e., S2G2), the current responses were significantly lower and relatively similar to one another, indicating that the interface selectively recognizes glycan structures that contain terminal Galβ. That is, measurements for galactosylated antibodies, G2 and G2F, were statistically different than the non-galactosylated antibodies, G0, G0F, and S2G2 (p < 0.005).
Similar to the titer detection interface, there was minimal response for the controls with no assembled IgG. This confirms that there was minimal contribution to the response from non-
specific binding of protein G:HRP. We tested this using other thiolated sugars, RCA120, and galactose-capped IgG and there was essentially no signal (not shown here). As this methodology relies upon a thiolated sugar and its corresponding lectin, our results suggest that the functionalized thiolated PEG interface could serve as a platform for other glycan specific, lectinbased detection interfaces for glycoprofiling purposes. Multiple interfaces that detect different glycan groups might then be used in parallel to give a rapid, high-throughput analysis of the glycan structures that are present on the antibodies. Lectin arrays using different assembly methodologies are commercially available, often employing fluorescence outputs on spotted membranes or microscope slides. Further work needs to be completed to determine the limits of detection of the interface and the selectivity of the interface for antibody glycoforms that contain varying numbers of galactose (i.e., G1, G1F) and sialic acid moieties (i.e., S1G1, S1G2).
To expand on the applicability of the interfaces for upstream bioprocessing, we wanted to characterize the interfaces with a representative sample from an upstream setting. We chose to evaluate the interfaces with IgG spiked into mammalian cell culture media (i.e., Dulbecco's Modified Eagle medium, DMEM) to determine if media components effected the performance of the interfaces. Incubation and current measurement conditions using the interfaces were identical to those in previous sections, except that the interfaces were immersed in solutions of IgG that were spiked into either PBS with 0.05% Tween-20 or DMEM. Results are shown in Figure 6.
Importantly, in Figure 6a, linearity was re-established for the IgG spiked in PBS with 0.05% Tween-20, consistent with results in Figure 4. However, the average current values in Figure 6 were uniformly higher than those in Figure 4 even though identical experimental conditions were used. Interestingly, as film assembly is robust and reproduceable, this suggests these differences were due to batch-to-batch differences in the preparations of cysteinylated protein G and/or protein G:HRP. Both of these are added to the interfaces as mg quantities and their activity might
vary based on batch. We also obtained a strong linear response for the IgG spiked DMEM samples, but with a lower slope. We suggest that the apparent lower currents coincide with media components lowering the number of binding events between the protein G and the IgG. We next superimposed all data for antibody titer in Figure 6b by normalizing each dataset by the maximum current obtained for the 1 g/L incubations; importantly linearity was maintained.
Interestingly, we found negligible differences between measurements of IgG in PBS or DMEM for the galactosylation detection (Figure 6b). As current responses from both samples were nearly identical, DMEM did not seem to impact the binding between RCA120 and galactosylated IgG. In discussion below, we provide a hypothesis for these observations. Overall, these data indicate that both of the interfaces can be used to evaluate IgG in samples of mammalian cell culture media.
Finally, we wanted to investigate whether the interfaces could withstand surface regeneration with the intent being to test whether they could be used for more than a single measurement.
After several experiments and in concordance with the literature (Firer, 2001;Goode, Rushworth, & Millner, 2015;Sheth et al., 2014), we found that 0.1M acetic acid and 1M sodium chloride, pH 2.8 could be used as a regeneration buffer as low pH conditions are often used to effectively decouple many protein-antibody interactions. We used this regeneration buffer to remove IgG while retaining the cysteinylated protein G (for the titer detection interface) and the thiolated sugar and hopefully, bound lectin (for the galactosylation detection interface). In this way, we retain the oxidized disulfide bonds and the functionalized hydrogel, but release product IgG.
As depicted in the schematics (top) in Figure 7a-b, after initial measurements of samples containing IgG (2 mg/mL IgG for titer detection, 1 mg/mL IgG for galactosylation detection) and 0.1 g/L protein G:HRP, we washed the surfaces of the assembled titer and galactosylation
detection interfaces 3 times (15 minutes each) with the low pH regeneration buffer. Then, after the regeneration washes, we made additional control measurements (added H 2 O 2 ) to ensure that the current responses were removed and that the surfaces had been successfully stripped of IgG and protein G:HRP. As seen in representative current responses (left) shown in Figure 7a-b, the interface responses were entirely diminished suggesting that complete regeneration was achieved.
Then, the interfaces were sequentially incubated with solutions containing IgG and protein G:HRP and repeat measurements were taken. After re-incubation using the same IgG concentrations, the responses returned to nearly identical signal strengths as the initial measurements. Interestingly, the dynamic responses were also unchanged. In all cases, the peroxide-mediated signal transfer was rapid (within seconds) and very stable. These data demonstrate that the interfaces could replicate the initial responses without incurring significant loss of activity. Both bar graphs (right) show that the absolute differences in response between the initial and re-incubation measurements for both the titer and galactosylation detection interfaces were 6.8% and 12%, respectively. We further note that the variations observed were both positive and negative in direction, suggesting there was no systematic error involved. Instead, deviations were likely due to random errors associated with the re-use protocol developed.
While these results were quite satisfactory, we decided to delve into the regeneration process to evaluate the apparent discord between maximum currents obtained in DMEM and PBS noted above. In Figure S3, we regenerated the interfaces after an initial measurement and then, unlike in Figure 7, we incubated the surfaces with protein G:HRP only to confirm that bound IgG had been removed. The low current obtained confirmed its effective removal from the interfaces. That said, we noted that during the regeneration and reuse of the galactosylation interface, subsequent addition of IgG indicated that the RCA120 lectin had been retained and was apparently not rate limiting in the analysis (Figure 7b). This was unexpected given that the manner of binding of
fluorescence microscopy with FITC RCA120 after regenerating the surface (Figure S4a). A strong fluorescent signal was retained (although, as expected, there was an observable loss compared to pre-wash) suggesting that a significant quantity of RCA120 remained bound to the interface. We re-evaluated the surfaces with both CV and EIS. Results shown in Figure S4b-c suggest a potential explanation; RCA120 was nonspecifically bound to the hydrogel through direct interactions with the thiolated PEG and thus was still attached to the surface after regeneration. That is, we ran the CV (Figure S4b) and EIS (Figure S4c) profiles without the thiolated Galβ and there was a degree of binding of RCA120 to the hydrogel. We suggest that these results might provide insight regarding the retention of galactose binding activity after regeneration, as noted above in Figure 6. We had found that the addition of DMEM did not decrease detection of galactosylation, while the titer measurement had dropped in half at the highest concentrations tested. We believe media components in DMEM might have lowered the number of binding events between protein G and the IgG, but since there were likely many RCA120 sites beyond those of the Galβ, this attenuation was not observed. Certainly, more studies would be needed to substantiate this hypothesis, but we note further that the enhanced stability of the lectin assembly to the oxidized thiolated PEG was a welcome surprise.
That is, our exploratory binding studies reinforced that the thiolated and functionalized PEG interfaces were robust and have the potential for multiple antibody-based measurements, potentially providing significant benefits to antibody development and production processes.
Moreover, we had not anticipated RCA120 binding to the thiolated PEG, but this is a welcome finding. Further work to characterize the lifetime and the robustness of the interfaces after multiple regeneration cycles is underway, along with the integration of these interfaces into a microfluidic device.