Exploring carbohydrate binding module fusions and Fab fragments in a cellulose-based lateral flow immunoassay for detection of cystatin C

The material and molecular parts of a conventional LFA cartridge were redesigned to explore the affinity of CBM for cellulose as a means of anchoring Cys-C capture antibodies. On the material side, a cellulose zone was incorporated on standard NC strips by layering cellulose over the test and control lines. On the molecular side, ZZ-CBM3 fusions were used at the cellulose zone of the LFA strip. This fusion, which combines a double Z domain from the staphyloccocal protein A with a type A, family III CBM (CBM3) from the cellulosomal-scaffolding protein A from Clostridium thermocellum, has been used successfully to anchor antibodies to paper^33,34, cellulose microparticles³⁵ and cellulose hydrogels³⁶. The ZZ domain, which is an engineered variant of the domain B of the staphyloccocal protein A, is expected to capture the anti-Cys-C antibodies via their Fc region^33,34,35,36. The CBM3 module, which displays a distinctive planar linear strip of polar and aromatic residues, located in one of the faces of its 9-strand β-sandwich jelly roll structure⁴⁸, is expected to bind strongly and specifically to the cellulose fibers^34,35,36. An initial set of experiments was designed to evaluate the feasibility of using cellulose coatings and CBMs in LFAs, which involved dispensing different biomolecules in the test lines of NC and NC + cellulose strips, addition of Alexa-labeled biomolecules to conjugate pads, cartridge assembly and running of samples containing adequate biomolecules.

Coating NC with cellulose fibers

Cellulose was incorporated on standard NC strips by dispensing repeatedly (five times) solutions of dissolved cellulose in NMMO over the same position of the test and control zones (see supplementary material S2). During the subsequent process of drying, cellulose fibers form in-situ via the bottom-up self-assembly of the dissolved cellulose chains. The final longitudinal width of the cellulose coat on the final NC strip varied between 2.7 and 2.9 mm (see Fig. S2b). The layer of regenerated cellulose fibers over NC was imaged with SEM by analyzing a strip over which cellulose was dispensed (Fig. 2). Images were obtained of the top layer (Fig. 2a) and of cross sections of the NC strip with cellulose (Fig. 2b).

The contrast between the cellulose layer and NC is quite evident in the image of the top view, which captures the boundary region between the two zones (Fig. 2a). The NC region presents the characteristic three-dimensional open pore structure, whereas the cellulose region is covered with a mesh of fibers. These fibers are relatively heterogenous, displaying diameters in the 30–50 µm range and lengths that seldom exceed 300 µm. In some cases, fibers with a hollow interior are visible. A connective layer between some fibers is also apparent in some zones. The cross-section image of the strip further shows that the cellulose layer sits on top of the NC membrane and that the inside structure of NC remains unaltered. Based on these images, we expect that fluid wicking through this region will split between the underlying NC membrane and the top cellulose layer. In other words, part of the fluid by-passes the top cellulose layer. The preparation of regenerated cellulose materials (fibers, films, membranes, hydrogels, etc.) from solutions of cellulose dissolved in NMMO has been described extensively^49,50. Fiber formation is usually induced by spinning, a process that involves the extrusion of the dissolved cellulose solution through an orifice spinneret and into an air gap, and then regeneration into a coagulation bath^50,51. Here the fibers regenerate during the process of drying that follows the dispensing, creating a layer of cellulose for affinity interaction with CBM fusions to take place.

A set of experiments was performed next using the control system (biotin-BSA:Alexa-streptavidin) to evaluate the impact of the layer of cellulose on the fluorescence of signals generated at the strips. Lines of biotin-BSA were dispensed on NC and NC + cellulose strips and LFA cartridges were assembled. Next, buffer samples containing either 1 ng/mL or 2 ng/mL of Alexa-labeled streptavidin were run. The flowing buffer carried the streptavidin-Alexa conjugates towards the test lines. The pixel volume of the fluorescent lines generated upon capture of the conjugates by the immobilized biotin-BSA was obtained with the ImageQuant analyzer (see details about the calculation of pixel volume of lines in “Materials and methods” section) and normalized relatively to the highest pixel volume (Fig. 3a).

Results show that when cellulose is used as a coating, the fluorescence lines are in general thicker, more intense (see Fig. 3b) and displayed a larger pixel volume (Fig. 3a). The increased thickness of the lines can be partly attributed to an increased lateral diffusion of the biotin-BSA solution on the top cellulose layer upon dispensing. We further suggest that this cellulose layer can adsorb a larger number of biotin-BSA molecules per unit area compared to the plain NC. As a result, the number of fluorescent complexes close to the surface increases, producing more intense signals. Finally, preliminary experiments confirmed that the coat of cellulose effectively constitutes an anchor point for ZZ-CBM3 fusions, as expected (see supplementary material S3).

Anchoring antibodies in NC strips with a cellulose coat

A first experiment was performed to check if ZZ-CBM3 fusions remain functional after being dispensed on the test line of NC and NC + cellulose strips. For comparative purposes, LFAs were also prepared by dispensing protein A alone on the test lines. The concentration of protein A in the solution used to produce the lines (2 mg/mL) was double the value of the concentration of ZZ-CBM3 solution (1 mg/mL). Since the molecular weight of protein A (64 kDa) is double the molecular weight of ZZ-CBM3 (32 kDa), the lines in the two LFA types had the same molar amounts of either protein. Once assembled, the cartridges were tested with samples containing either 2.5 or 5 ng/mL of Alexa-labeled anti-Cys-C antibodies. The pixel volume of test lines, V_T, was normalized relatively to the highest pixel volume obtained in the experiments (Fig. 4a). Fluorescence images of the test line region were also captured by the ImageQuant camera (Fig. 4b).

Results show that the immobilized ZZ-CBM3 was able to capture the flowing antibody, whether the fusion was simply adsorbed to a plain NC strip or anchored on the cellulose layer via biomolecular interaction. Moreover, thicker and more intense fluorescence lines were always obtained when the fusion was deposited over the cellulose layer (Fig. 4). This could indicate that more antibodies were captured and/or that there is an increase in sensitivity due to the presence of the cellulose coat, as seen above (Fig. 3) and before²⁶. Furthermore, the ZZ part of the fusion that is responsible for the antibody capture is likely to be less hindered by the surface and have a more favorable orientation when it is anchored on cellulose via the CBM3 part, as compared to the situation where it is simply adsorbed to NC.

Control experiments performed with test lines prepared with protein A show that the lines obtained were less intense and narrower when compared with the corresponding lines obtained with ZZ-CBM3 lines (Fig. 4). These results should be interpreted with caution. For once, even though the same molar amounts of protein A and ZZ-CBM3 were used, it should be kept in mind that protein A has in effect five antibody-binding domains (A, B, C, D, and E) compared to the two Z domains of the fusion⁵². Since the affinity towards antibodies is known to vary across these domains, it is virtually impossible to guarantee that the number of antibody capture sites is the same in the two experiments. Additionally, since the D and E domains of protein A have an affinity toward the Fab region of antibodies, the orientation of the capture antibody bound to protein A could be different from that of the antibody bound to CBM-ZZ⁵². Still, we think that the higher intensity observed when ZZ-CBM3 fusions were used to capture antibodies relatively to protein A in NC + cellulose strips can be justified in part by the more favorable orientation of the ZZ domain that is afforded by the fusion (Fig. 1b).

Interference of detection antibodies with capture antibodies anchored with ZZ-CBM fusions

An experiment was designed next with two specific objectives. Firstly, we wanted to confirm if detection antibodies labeled with Alexa Fluor do compete with capture antibodies when these are anchored on the test line of analytical strips via ZZ-CBM3 fusions (Fig. 1c), as anticipated and also seen by Yang²⁷. Next, we wanted to check if this interference issue could be resolved by using Fab fragments of the detection antibody instead of the full-length antibody (Fig. 1d). For this purpose, LFA were assembled with test lines containing conjugates of ZZ-CBM3 and anti-Cys-C capture antibody in the test line, which were then run with samples containing either Alexa-labeled, full-length detection antibodies, or Alexa-labeled Fab fragments of the detection antibodies. As before, tests were made with NC and NC + cellulose strips and the pixel volume of lines in the test region, V_T, were normalized relatively to the highest pixel volume (Fig. 5).

The results confirmed that the ZZ-CBM3 fusions, albeit being conjugated with the capture antibodies, do indeed capture a fraction of the full-length detection antibodies that flow by, as judged by the appearance of fluorescence in the test lines (see inset images in Fig. 5). This interference is more significant when the ZZ-CBM3:antibody conjugates are anchored on NC + cellulose strips, which further attests to the beneficial role of the cellulose layer seen above. When the full-length detection antibodies were replaced by the corresponding Fab fragments, however, no fluorescence could be observed in the test lines of either the NC and NC + cellulose strips. This result confirms that the use of Fab fragments for detection resolves the issue of interference as anticipated (Fig. 1c and d). Furthermore, additional experiments were performed with LFAs prepared by conventional adsorption of the anti-Cys-C capture antibody in the test line of analytical strips (NC, NC + cellulose) to check if the process used to generate the Fab fragments did not affect their ability to recognize Cys-C. Results confirmed that the Fab fragments were able to bind to Cys-C and generate signals comparable to those generated by full length detection antibodies (see supplementary material S4).

Detection of cystatin C with capture antibodies anchored with ZZ-CBM3 fusions and Fab detection

A set of experiments was designed next to evaluate the ability of LFAs based on anti-Cys-C capture antibodies anchored with ZZ-CBM3 fusions, and on Alexa-labeled Fab fragments, to detect Cys-C. These experiments were performed using NC strips with a cellulose layer on the test lines. Since we further wanted to check the effect of the amount of capture antibodies in the intensity of signals generated, strips were prepared by dispensing 1:1, 1:2 and 1:4 dilutions of the conjugates of ZZ-CBM3 and anti-Cys-C capture antibody (molar ratio of 1:1.5, 0.5 mg/mL antibody). Conjugate pads were prepared with Alexa-labeled Fab fragments. Following assembly of the LFA cartridges, 10 ng/mL Cys-C samples were run.

The results confirm that the use of conjugates of ZZ-CBM3 fusions with the capture antibodies in the test lines generates fluorescence signals that are superior to those obtained when adsorbed capture antibodies are used (Fig. 6). As expected, the use of larger amounts of capture antibodies led to an increase in the intensity of the fluorescence signals (Fig. 6). This is also seen by direct observation of the fluorescence images captured by the ImageQuant camera. The images show that lines obtained with ZZ-CBM3 fusions have a thickness that increases with the use of larger amounts of capture antibodies (see inset in Fig. 6).

Calibration curves

Sets of experiments were performed by running Cys-C standards with various concentrations in the range 0–10 ng/mL in LFA cartridges with the new architecture—analytical strip made of NC with layered cellulose, anti-Cys-C capture antibodies anchored via ZZ-CBM3 and detection with Alexa-labeled Fab fragments (see Fig. S5a in supplementary material S5). An intermediate architecture was also tested that featured analytical strips made of NC alone, anti-Cys-C capture antibodies anchored via ZZ-CBM3 and detection with Alexa-labeled Fab fragments (see Fig. S5b in supplementary material S5). For comparison purposes, the same Cys-C standards were run in LFA cartridges with the conventional LFA architecture—analytical strip made of NC, anti-Cys-C capture antibodies adsorbed on test line and detection with Alexa-labeled full-length antibodies (see Fig. S5c in supplementary material S5). Experiments were performed in triplicate. Representative fluorescence images of the analytical strips in LFA cartridges are shown in supplementary material S6. Well defined test and control fluorescent lines were obtained in all strips. As seen before (see Figs. 3 and 4), the lines were always thicker when a cellulose layer was included in the test and control zones.

The pixel volume of test (V_T) and control (V_C) lines was obtained with the ImageQuant instrument and the response of the LFA devices was measured by calculating the pixel volume ratio, V_R (V_T/V_C) (see “Materials and methods”). This triplicate V_R data was further used to compute the individual CoV, which were then averaged to yield the intra-assay CoV. The values of average CoV of 0.72%, 1.05% and 1.44% were obtained for the new, intermediate and conventional LFA architecture, respectively. This provides a good indication that measurements of Cys-C concentration in the devices are reliable and consistent. Calibration curves were constructed next by plotting the replicate V_R data as a function of Cys-C concentration for the new and standard LFA architectures (Fig. 7). A linear behavior of the concentration–response relationship was observed in both cases (see regression statistics data in Fig. 7). The working range obtained for the three LFA is clearly compatible with the clinical diagnostic range for Cys-C (5–120 ng/mL in healthy patients, > 250 ng/mL in patients with kidney disease)³⁹, if proper dilutions are made.

Detection of cystatin C in mock urine samples

The quantitative range for Cys-C concentration afforded by the LFA (up to 10 ng/mL) indicates that urine samples must be diluted when testing healthy (5–120 ng/mL) and diseased individuals (> 250 ng/mL). Experiments were thus designed to detect Cys-C in mock urine samples prepared in artificial urine that are representative of normal patients (100 ng/mL) and kidney tubular disease patients (4000 ng/mL). As a control, and to see if urine components somehow interfere with detection, standard samples with the same concentration were also prepared in buffer. The samples with normal and abnormal Cys-C concentration were diluted with buffer 1:20 and 1:800, respectively, to bring the Cys-C concentration down to 5 ng/mL, which falls within the testing range (up to 10 ng/mL). These diluted samples were then run in LFA cartridges with the new architecture—analytical strip made of NC with layered cellulose, anti-Cys-C capture antibodies anchored via ZZ-CBM3 and detection with Alexa-labeled Fab fragments. Fluorescence images of the corresponding analytical strips are shown in supplementary material S7.

The intensity of the fluorescent signals generated at the test line was slightly lower when samples were prepared in artificial urine as compared to buffer (Fig. 8). This could indicate that some components of urine may interfere with the signals generated. If this is in fact the case, the dilution required can be seen as advantageous given that it will dilute out possible interferents. The intensity of fluorescent signals obtained with samples containing abnormal values of Cys-C were equivalent to those obtained with samples representative of normal patients. This was expected since the dilution factors were selected to produce samples with the exact same Cys-C concentration.