Paired Agent Imaging to Evaluate the Improvement of Affinity-Matured Nimotuzumab K4 and K5 Variants as EGFR-Detecting Optical Imaging Agents

Abstract

Molecular imaging is a useful tool for the in vivo validation of antibody therapeutics. It shows the accumulation properties of antibodies in tumors and organs that cannot be predicted by using in vitro assays. Two issues with characterizing antibodies using in vivo imaging are: (i) the fluorescent signal from the antibody is sensitive to its depth in the tissue, and (ii) there are large biological variabilities observed in in vivo imaging studies. Paired agent imaging measures the fluorescence of a test antibody and a control antibody in the same animal, allowing the fluorescence of the test antibody to be normalized to that of the control antibody. We used paired agent imaging to compare the in vivo imaging properties of two affinity-matured variants of the anti-EGFR antibody nimotuzumab (K4 and K5) to the parental antibody. To perform paired agent imaging, we labeled the test antibodies (K4 and K5) with IRDye800CW and the control antibody (nimotuzumab) labeled with IRDye680RD, and coinjected them into mice bearing EGFR-positive xenografts. Near-infrared fluorescent imaging was used to quantitate the relative amount of each antibody present in tumors and organs. Paired agent imaging allowed us to detect differences in in vivo fluorescence between K4, K5, and nimotuzumab, where K5 had the highest accumulation in the tumor, followed by K4 and nimotuzumab.

Introduction

The epidermal growth factor receptor (EGFR) is overexpressed in many cancers, leading to cell proliferation and invasion. There are four approved anti-EGFR antibody therapeutics in the clinic, cetuximab, panitumumab, necitumumab, and nimotuzumab. Cetuximab, panitumumab, and necitumumab have high incidences of adverse events, such as acneiform skin rash, including grade ≥ 3 skin toxicity. , In contrast, nimotuzumab has an excellent safety profile, with minimal toxic skin reactions. This reduced skin toxicity is attributed to its lower affinity for EGFR.

Nimotuzumab has a dissociation constant (K_D ) between 4 and 10 nM when measured by flow cytometry and 67 nM when measured by SPR, which is 10- to 100-fold higher than cetuximab (0.39 nM), necitumumab (0.32 nM), and panitumumab (0.05 nM). Currently, it is not clear whether tighter antibody binding results in better efficacy, as it can also lead to increased off-target accumulation. A study by Zhou et al. showed that antibody dissociation constants (K_D s) between 1 and 10 nM are optimal for EGFR tumor targeting, accumulation, and retention, which is supported by mathematical modeling. In a previous study by Tundidor et al., affinity-matured variants of nimotuzumab were generated, referred to as K4 and K5, with K_D s of 22 nM and 19 nM when measured by SPR, respectively, which are ∼three times lower than nimotuzumab (67 nM). The ability of K4 and K5 to inhibit EGFR signaling in vitro correlated with their increased affinity; however, no in vivo efficacy was reported.

We previously characterized the near-infrared (NIR) imaging properties of IRDye800CW-nimotuzumab in mice with EGFR-positive xenografts. Due to the high variability between animals, it is difficult to compare nimotuzumab and its affinity-matured variants, K4 and K5, which differ in binding by ∼3-fold, without larger cohorts of animals to overcome biological variability.

Paired agent imaging (PAI) has been used to improve imaging quality by enhancing the contrast between targeted (disease) and untargeted (healthy) tissues. , PAI involves coinjecting, in the same animal, targeted and untargeted imaging probes that are labeled with different fluorescent dyes. The untargeted probe is used to measure the nonspecific signal in the tumor or tissue of interest, enabling a more accurate signal from the targeted probe to be obtained. Several different molecules have been used as untargeted probes, including free fluorescent dye or untargeted probes that are similar to the targeted probes, such as tetralactam macrocycles and affibodies.

Here, we compared the in vivo NIR fluorescent imaging properties of nimotuzumab, K4, and K5 to measure the effects of binding affinity on tumor accumulation. A previous study showed that cetuximab, which has a lower K_D than nimotuzumab, results in earlier tumor accumulation relative to nimotuzumab. We used a modification of PAI, using two targeted antibodies labeled with different fluorophores and coinjected them into the same animal. This strategy allows for a comparison of the fluorescent signal ratio between two probes rather than comparing the ratio of signal intensity in different animals.

Experimental Section

Protein Labeling and Characterization

Nimotuzumab, K4, and K5 IgGs were labeled with IRDye800CW NHS ester (LI-COR Biosciences), and nimotuzumab, and a controlan untargeted IgG that binds to maltose-binding protein (MBP) was labeled with IRDye680RD NHS ester (LI-COR Biosciences). IgGs were labeled at a 1:3 molar ratio of IgG to dye in PBS, according to the manufacturer’s instructions. Impurities, including free dye, were removed with a Zeba Spin Desalting Column 7 MWCO instrument (ThermoFisher Scientific). Protein concentration and labeling ratio were calculated as previously described.

The concentration and labeling ratio of labeled antibodies were determined by measuring the absorbance at 280 nm (A ₂₈₀) for protein concentration, 780 nm (A ₇₈₀) for IRDye800CW-labeled antibodies, and 680 (A ₆₈₀) for IRDye680RD-labeled antibodies. The concentration was corrected for the attached dye and was calculated using the following formula:

$$[\mathrm{antibody}](\mathrm{M})=\left(\frac{A_{280}-(A_{\mathrm{dye}}\times 0.03)}{{ϵ}_{\mathrm{protein}}}\right)$$

where 0.3 is the correction factor for the dye, which accounts for 3% absorbance of the dye in the 280 nm channel, ϵ _protein = 210,000 M^–1cm^–1 is the average extinction coefficient of an antibody at 280 nm, and A _dye is the absorbance at 780 nm for IRDye800CW or 680 nm for IRDye680RD. The labeling ratio was calculated using the following formula:

$$\mathrm{dyes}\mathrm{per}\mathrm{antibody}\left(\frac{\mathrm{dyes}}{\mathrm{antibody}}\right)=\left(\frac{A_{\mathrm{dye}}}{{ϵ}_{\mathrm{dye}}}\right)/[\mathrm{protein}]$$

where ϵ_dye = 240,000 M^–1cm^–1 for IRDye800CW and 170,000 M^–1cm^–1 for IRDye680RD. Labeled and unlabeled antibodies (2.5 μg) were analyzed for size and purity using an Agilent 2100 Bioanalyzer (Agilent Technologies) according to the manufacturer’s instructions.

Flow Cytometry

Flow cytometry was performed with the EGFR-positive breast cancer cell line MDA-MB-468. Cells were dissociated using TrypLE Express (ThermoFisher Scientific), washed, and suspended in PBSF (phosphate-buffered saline and 2% fetal bovine serum). IgGs were titrated against 1 × 10⁵ MDA-MB-468 cells with concentrations ranging from 0 to 2 μM by incubating for 30 min at room temperature and then 15 min on ice. Cells were washed with PBSF and incubated with Goat F­(ab’)­2 antihuman IgG FITC (Beckman Coulter) secondary for 30 min on ice. Cells were washed again, suspended in PBSF, and analyzed on a Beckman Coulter Gallios flow cytometer. FlowJo version 10.5.3 was used to determine the mean fluorescent intensity (MFI) values. MFI values were normalized and used to calculate the percentage of antibody bound. K_D apparent values were calculated with GraphPad Prism version 6 using a nonlinear regression curve fit.

Mouse Xenograft Models

Mouse experiments were performed with the University of Saskatchewan Animal Research Ethics Board approval (AUP20160112) under the guidelines of the University of Saskatchewan Animal Care Committee. CD-1 nude mice were purchased from Charles River Canada. Female mice, 6–8 weeks old, used for experiments, had ad libitum access to food and water. MDA-MB-468 and A-431 cells were purchased from ATCC and cultured in 90% Dulbecco’s Modified Eagle’s Medium (DMEM) or Roswell Park Memorial Institute (RPMI) medium, respectively, and 10% Fetal Bovine Serum (FBS) with 5% CO₂ in a humidified incubator. To establish xenografts, MDA-MB-468 and A-431 cells were washed, suspended in their respective serum-free media, and mixed 1:1 with Corning Matrigel Matrix (Corning) on ice. The right hind flank was injected with 100 μL (ten million cells) of the cell-matrigel suspension. Tumor volumes were measured by using external calipers.

NIR Fluorescence Imaging

When xenografts reached a size of 150–300 mm³, mice were coinjected with 0.25 nmol of IRDye680RD-labeled nimotuzumab and 0.25 nmol of IRDye800CW-labeled nimotuzumab, K4, or K5, or 0.5 nmol of IRDye800CW-labeled nimotuzumab. The untargeted antibody was labeled with IRDye680RD, and 0.25 nmol was coinjected with 0.25 nmol of IRDye800CW-labeled nimotuzumab. The mice were imaged over time using a LI-COR Pearl small animal imaging system. Accumulation of NIR fluorescent-labeled antibodies in the tumor and background was calculated by selecting three regions of interest (ROI) of equal size from the xenograft and contralateral side, respectively, and measuring the MFI in the 700 and 800 nm channels. The MFI of ROIs was averaged and normalized using the antibody labeling ratio. A normalization factor was calculated using the MFI from the ROIs of the 1:1 ratio of nimotuzumab labeled with IRDye680RD and IRDye800CW. This normalization factor was used to correct the ratios to 1 at each time point to account for any differences in the fluorescent dyes. This ratio was used to generate mouse images showing the ratio of dye in the mouse body using the R Project for statistical computing. This ratio was used to generate mouse images showing the ratio of dye in the mouse body using the R Project for statistical computing. A two-way analysis of variance (ANOVA) with multiple comparisons using GraphPad Prism was used to compare the normalized NIR fluorescence signals in the tumor and contralateral side. A two-way analysis of variance (ANOVA) with multiple comparisons using GraphPad Prism was used to compare the differences in the ratio of signal between K5:nimotuzumab and K4:nimotuzumab in the tumor compared to the contralateral and the tumor compared to the liver. All experiments were performed with at least three biological replicates (n ≥ 3) and represent the standard error of the mean (SEM).

A power analysis was done using the power.t.test function in R to determine the number of animals required per group to detect a change in fluorescence of 50% (delta = 0.5), a power of 80% (1 minus Type II error probability), and a significance level (Type I error probability) of 5%, using either a single NIR fluorescent channel or a ratio of two NIR fluorescent channels. The mean was fixed at 1, and the standard deviations were taken from the data as follows: the standard deviation from a single channel = 0.64, and standard deviation from two channels = 0.28.

Results

Characterization of Labeled Antibodies

Labeled antibodies were characterized for relative size to ensure that the size and purity did not change and/or create degradation or aggregation products after labeling (Figure ). All antibodies were within the predicted size range, with no change after labeling, and had purities of ≥90% (Figure ). The labeled antibodies were also tested and compared to their respective parent antibodies for their ability to bind to the EGFR-positive cell line MDA-MB-468. There was minimal change in binding when comparing the labeled vs unlabeled antibodies (Figure D–F). The control antibody did not show any binding (Figure F).

1.

Characterization of antibodies. (A–C) Labeled and unlabeled antibodies were analyzed by microcapillary electrophoresis for relative size and purity. (D–F) Labeled and unlabeled antibodies were assessed for binding to EGFR-positive cells. (G) K4, K5, and nimotuzumab were compared for binding to MDA-MB-468 cells. The percentage of antibody-bound cells (% Bound) was plotted against antibody concentration to obtain a K _Dapp. Error bars represent the standard deviation. (H) A representative flow cytometry histogram shows the binding of nimotuzumab to MDA-MB-468 cells.

Interaction of Nimotuzumab, K4, and K5 with the EGFR-Positive Cell Line MDA-MB-468

We analyzed the apparent K_D (K _Dapp) of nimotuzumab, K4, and K5 with the EGFR-positive cell line MDA-MB-468 by using flow cytometry. Nimotuzumab had a K _Dapp of 4.8 ± 1.1 nM, K4 had a K _Dapp of 3.1 ± 0.2 nM, and K5 had a K _Dapp of 1.9 ± 1.1 nM (Figure G, H); however, there was no significant difference between the K _Dapps.

Paired Agent Imaging

Paired agent imaging (PAI) can overcome some issues encountered with in vivo NIR fluorescence imaging, including variations due to the tissue depth of the tumors and animal variability, by using a control antibody to normalize these differences. To test PAI, we coinjected nimotuzumab-IRDye800CW (labeling ratio: 0.6) and nimotuzumab-IRDye680RD (labeling ratio: 2.5) at a 1:1 molar ratio in mice with A-431 xenograft tumors. Tumors had NIR fluorescent ratios at 800 and 700 nm close to 1. However, the ratio was not exactly 1 across all time points when the 800 and 700 nm signals were corrected for the labeling ratio (Figures A,D and 1A).

2.

Paired agent imaging (PAI). (A) Coinjection of nimotuzumab labeled with 800CW or 680RD NIR dyes in a 1:1 ratio into mice bearing A-431 xenografts. Mice are imaged in the 800 and 700 nm channels. Merge: Represents an overlay of the NIR fluorescence in the 800 and 700 nm channels. A ratio of the signal in the 800 nm:700 nm channel. S = stomach; T = tumor. (B) Coinjection of nimotuzumab-IRDye800CW and an untargeted control IgG-IRDye680RD administered at a 1:1 ratio into mice bearing A-431 xenografts. Mice are shown as a merge (overlay of the 800 and 700 nm channels) and a ratio of the 800 nm:700 nm fluorescence signal. (C) Coinjection of nimotuzumab labeled with 800CW or 680RD at a 2:1 ratio, respectively, into mice bearing A-431 xenografts. Mice are shown as a merge (overlay of the 800 and 700 nm channels) and a ratio of the 800 nm:700 nm fluorescence signal. (D) NIR fluorescent signal in the tumor is shown in MFI per area (mm²) versus time, and (E) the ratio of 800 nm:700 nm NIR fluorescence signal versus time for nimotuzumab-IRDye800CW and nimotuzumab-IRDye680RD coinjected at a 1:1 and 2:1 ratio, respectively, and nimotuzumab-IRDye800CW and control IgG-IRDye680RD at a 1:1 ratio. The 95% confidence interval is shown in gray shading in D and E. C = control, Nimo = nimotuzumab.

We tested whether the change in the IRDye800CW and IRDye680RD NIR fluorescent ratio (800 nm:700 nm) correlated to the ratio of nimotuzumab-IRDye800CW and nimotuzumab-IRDye680RD administered. We coinjected nimotuzumab-IRDye800CW and nimotuzumab-IRDye680RD at a 2:1 ratio, respectively (Figure B, D). The 800 nm:700 nm NIR fluorescence ratio was 2 ± 1 after correcting for the labeling ratio (Figure S1A). An additional correction factor was applied to the 1:1 ratio of nimotuzumab-IRDye800CW and nimotuzumab-IRDye680RD to normalize the 800 nm:700 nm NIR fluorescent ratio to 1, which improved the 2:1 800 nm:700 nm fluorescence ratio to 2 ± 0.5 (Figure S1B). This normalization factor accounts for differences in the fluorophores, penetration through tissue, or changes in distribution, et, and represents an important control for ensuring accurate comparison between antibodies.

We measured the 800 nm:700 nm NIR fluorescence ratio in tumors of mice with A-431 xenografts that were coinjected with nimotuzumab-IRDye800CW and control IgG-IRDye680RD at a 1:1 ratio (Figure B, D). Tumor accumulation of nimotuzumab-IRDye800CW was significantly higher than that of the control IgG-IRDye680RD. This experiment allowed us to obtain the tumor signal of nimotuzumab above the background signal due to the enhanced permeability and retention (EPR) effect. In this case, the 700 nm fluorescence signal due to EPR observed with IgG-IRDye680RD was relatively negligible (∼25-fold lower) compared to the 800 nm fluorescence signal observed with nimotuzumab-IRDye800CW. The same correction factor used for the nimotuzumab-IRDye800CW and nimotuzumab-IRDye680RD 1:1 ratio correction was applied to the 2:1 ratio study and to the nimotuzumab-IRDye800CW:control IgG-IRDye680RD 1:1 ratio study (Figure E).

Alfalfa-Free Diet Improves PAI

In the PAI experiments, the gastrointestinal (GI) tract was visible in the 700 nm channel (Figure ), which was due to alfalfa in the mouse diet that fluoresces at 700 nm (Figure S2A). To reduce 700 nm NIR fluorescence in the GI tract, mice were switched to the Picolab 75IF alfalfa-free diet (Lab Diet), which reduced NIR fluorescence in the 700 nm channel. The 700 nm channel signal reached its minimum after the mice were fed the alfalfa-free diet for 48 h (Figure S2B).

Comparison of K4 and K5 Xenograft Accumulation Relative to Nimotuzumab

The modestly improved K _Dapp values of K4 and K5 relative to nimotuzumab suggest that it will be difficult to detect differences in tumor accumulation of these antibodies. Such studies require many animals to overcome the biological differences and the inherent noise in NIR fluorescence imaging due to the absorption of fluorescence signals by tissues. To address these issues, we used PAI to compare imaging between nimotuzumab and its variants (K4 or K5) in the same animal.

To directly compare the xenograft accumulation of nimotuzumab relative to K4 or K5, we coinjected nimotuzumab-IRDye800CW (labeling ratio: 0.74), K4-IRDye800CW (labeling ratio: 1.43), or K5-IRDye800CW (labeling ratio: 0.83) at a 1:1 ratio with nimotuzumab-IRDye680RD. Mice were imaged over time on the 700 and 800 nm channels. MFIs of IRDye800CW-labeled K4, K5, and nimotuzumab were divided by the MFI of IRDye680RD-labeled nimotuzumab (Figure A). The 800 nm:700 nm NIR fluorescence ratio of mice coinjected with nimotuzumab-IRDye800CW and nimotuzumab-IRDye680RD was normalized to one, and this correction factor was applied to the 800 nm:700 nm NIR fluorescence ratio observed with K4-IRDye800CW and nimotuzumab-IRDye680RD imaging, as well as K5-IRDye800CW and nimotuzumab-IRDye680RD imaging (Figure B). K4 and K5 showed an average of 2- and 3-fold increases, respectively, in tumor accumulation relative to nimotuzumab (Figure B). There was a correlation of the measured K _Dapp for each antibody when plotted against the normalized 800 nm:700 nm NIR fluorescence ratio (Figure C). The difference in uptake in the tumor is clearly visible in the mouse imaged with the K5 antibody and modestly visible in the mouse imaged with the K4 antibody (Figure A).

3.

PAI of K4 or K5 relative to nimotuzumab in mice bearing MDA-MB-468 xenografts. (A) K4-IRDye800CW, K5-IRDye800CW, or nimotuzumab-IRDye800CW were coinjected with nimotuzumab-IRDye680RD into mice bearing MDA-MB-468 xenografts and imaged over time. The ratio of the fluorescent signal of K4-IRDye800CW:nimotuzumab-IRDye680RD (K4/Nimo), K5-IRDye800CW:nimotuzumab-IRDye680RD (K5/Nimo), and nimotuzumab-IRDye800CW:nimotuzumab-IRDye680RD (Nimo/Nimo) versus time is shown for each mouse. (B) The ratios of the MFI of K4/Nimo, K5/Nimo, and Nimo/Nimo in the A-431 xenograft tumor were normalized for labeling ratio and plotted versus time. The 95% confidence interval is shown in gray shading. (C) Correlation of K _Dapp with the K4-IRDye800CW:nimotuzumab-IRDye680RD, K5-IRDye800CW:nimotuzumab-IRDye680RD, and nimotuzumab-IRDye800CW:nimotuzumab-IRDye680RD ratios of 800 nm:700 nm at 24, 48, 72, and 96 h postinjection. The error is the standard deviation.

The ratio of the signal in the 700 and 800 nm channels for K4-IRDye800CW:nimotuzumab-IRDye680RD and K5-IRDye800CW:nimotuzumab-IRDye680RD was compared in the contralateral site and the liver. 800 nm:700 nm NIR fluorescence ratios in mice injected with nimotuzumab-IRDye800CW and nimotuzumab-IRDye680RD were normalized to 1. The same correction factor was applied to the K4-IRDye800CW:nimotuzumab-IRDye680RD- and K5-IRDye800CW:nimotuzumab-IRDye680RD-treated mice. The normalized NIR fluorescent signal for mice injected with K4-IRDye800CW:nimotuzumab-IRDye680RD was slightly higher in the tumor compared with the contralateral site (Figure A). The normalized NIR fluorescent signal for K5-IRDye800CW:nimotuzumab-IRDye680RD-injected mice was much higher in the tumor at later time points and had slower clearance compared to the contralateral site. Mice imaged with K4 and K5 antibodies had elevated levels of circulating antibody, as seen by the higher levels of background NIR fluorescence (Figure A) and the higher NIR 800 nm:700 nm fluorescence ratios in the contralateral site (Figure A).

4.

K4-IRDye800CW:nimotuzumab-IRDye680RD (K4:Nimo), K5-IRDye800CW:nimotuzumab-IRDye680RD (K5:Nimo), and nimotuzumab-IRDye800CW:nimotuzumab-IRDye680RD (Nimo:Nimo) ratios of 800 nm:700 nm in the contralateral site, liver, and tumor. (A) The ratio of the signal in the 800 nm:700 nm channel over time in the tumor and contralateral site. The difference in signal was significant in the K5:Nimo comparing tumor vs contralateral at 96 and 216–288 h after injection (*p ≤ 0.05) and at 120–192 h after injection (**p ≤ 0.01). (B) The ratio of the signal in the 800 nm/700 nm channel over time in the tumor and liver. The 95% confidence interval is shown in gray shading. The difference in signal was significant in the K5:Nimo comparing tumor vs liver at 72–96 and 216–288 h after injection (*p ≤ 0.05) and at 120–192 h after injection (**p ≤ 0.01).

Similarly, in the liver, K4-IRDye800CW and K5-IRDye800CW had higher normalized NIR fluorescent signals in the tumor over time compared to the liver (Figure B). K4 had a higher accumulation in the tumor compared to the liver, but the clearance was similar. K5 also had higher accumulation in the tumor compared to that in the liver and had slower clearance from the tumor (Figure B).

The coefficient of variance (CV) was calculated for K4, K5, and nimotuzumab, comparing variability in fluorescence from a single channel (700 or 800 nm) and the ratio of the 800 nm:700 nm channels (Figure A). The difference in variability was statistically significant (p < 0.0001) between the two analysis methods. The mean CV for a single fluorescence channel measurement was 64%, and the mean CV for the ratio of the 800 nm:700 nm channels was 30%, showing the ability of using the NIR fluorescence ratio from two channels to reduce the variability in the imaging data.

5.

Coefficient of Variance (CV), power analysis, and Cohen’s d (effect size) for NIR fluorescence imaging using a single 800 nm channel and the ratio of 800 nm/700 nm channels. (A) The violin plot shows the CV for the K4, K5, and nimotuzumab imaging data sets, comparing the variability using a single channel (700 or 800 nm) or the normalized signal using the ratio (800 nm/700 nm) from the two channels. (B) Power analysis using K4 and K5 imaging data sets determines the sample size required to detect a 50% change in tumor uptake using a single channel or the ratio of two channels. The dashed line represents 80% power. (C) Cohen’s dsingle channel compared to PAI (ratio) analysis, comparing all time points. (D) Cohen’s dsingle channel vs PAI (ratio) analysis, comparing K4 to K5 at 72 and 96 h postinjection.

A power analysis was done to determine the sample size required to detect a delta of 0.5, a power of 80%, and a significance of 0.05 by using either a single NIR fluorescent channel or a ratio of two NIR fluorescent channels (Figure B). A single NIR fluorescence channel measurement requires 15 animals to detect this difference, whereas using the ratio between the two NIR fluorescence channel measurements requires only five animals to detect a difference. This analysis shows that using two channels to normalize the NIR fluorescence signal significantly reduces the number of animals required to obtain statistically significant data.

We performed an analysis of the effect size at all imaging time points, as well as focusing on the 72–96 hpi time points, where the fluorescent tumor signal peaks. The effect size measures how K4 and K5 antibodies differ from nimotuzumab relative to the variability. In the single-channel analysis, differences between animals increase the variability, yielding small effect sizes (d ≈ 0.5 on average across all time points and probes). In contrast, PAI decreases animal-specific differences, producing larger effect sizes (d ≈ 1.8 on average across all time points and probes), indicating that the true difference is consistent and detectable using fewer animals (Figure C). At peak times of tumor accumulation (72–96 hpi), effect sizes based on the PAI ratio were large for both K4 and K5 (d ≈ 2.1 and d ≈ 2.7, respectively), whereas single-channel effect sizes based on single-channel measurements were lower for K4 and K5 (d ≈ 0.4 and d ≈ 0.8, respectively) (Figure D). These results demonstrate that PAI increases precision by removing variability between animals, revealing differences in tumor uptake that are less prevalent in conventional single-channel analyses.

Discussion

Several biological properties play a role in the tumor accumulation of antibodies, including tumor size, growth rate, blood vessel density, tumor type, location, and permeability. Antibody properties also influence tumor accumulation, including size and binding affinity. In this study, we compared the role of binding affinity on the tumor accumulation of nimotuzumab and two affinity-mature variants, K4 and K5. K4 and K5 antibodies have K_D s of 22.4 and 18.6 nM, respectively, which are lower than nimotuzumab’s monovalent Fab (67.4 nM), as measured by SPR. Using flow cytometry, we calculated the K _Dapp values of K4, K5, and nimotuzumab against the EGFR-positive cell line MDA-MB-468. The K _Dapp of K4 and K5 were 3 and 2 nM, respectively, which were modestly lower than nimotuzumab (5 nM). The K _Dapp against EGFR-positive cells, as measured by flow cytometry, are lower than the K_D s against recombinant EGFR, as measured by SPR, for K4, K5, and nimotuzumab. We compared how the K _Dapp values of nimotuzumab, K4, and K5 influenced tumor accumulation in a mouse xenograft model using NIR fluorescent imaging.

Given the many factors that affect tumor imaging in xenograft models, we employed PAI to minimize the biological variability in imaging signals. Using this approach, we observed a higher accumulation of K4 and K5 in the tumor relative to nimotuzumab. The 800 nm:700 nm ratio of the NIR fluorescent signal between nimotuzumab-IRDye800CW:nimotuzumab-IRDye680RD, K4-IRDye800CW:nimotuzumab-IRDye680RD, and K5-IRDye800CW:nimotuzumab-IRDye680RD in the tumor varied linearly with the ratio of the K_D s, showing that an increase in affinity results in an increase in tumor NIR fluorescence signal over the range of K_D s tested. The improved tumor targeting of affinity-matured K5 should translate to improvements in therapeutic and imaging applications.

Besides K_D , the residence time of the antibody in the blood can contribute to tumor accumulation. The NIR fluorescence imaging showed that K5 had higher background NIR fluorescence in mice and a higher NIR fluorescence peak at the contralateral site compared to K4 and nimotuzumab. This was likely a contributing factor to the increased tumor accumulation, since an increase in circulation time allows the antibody more time to bind and accumulate in the tumor. In addition to this, K4 has been shown to be more prone to aggregation, which may also account for the lower circulation time and tumor accumulation.

There are several ways to characterize antibody therapeutics in vitro, including binding affinity, internalization, cytotoxicity, and stability. These parameters do not always correlate with in vivo efficacy. Imaging is a good predictor of efficacy since it shows tumor accumulation longitudinally within the same animal, giving multiple measurements for assessment. Effective dosing can then be predicted using the antibody residence time in the tumor. Further, PAI allows multiple antibodies to be imaged simultaneously to more accurately assess the differences between variants. In addition, PAI allows the measurement of biodistribution and pharmacokinetics if done at low labeling ratios. PAI can also provide insight into toxicity. For example, you can image antibody residence time in organs over time to identify potential site(s) of toxicity. By coinjecting two antibodies, reductions in organ accumulation and the clearance of antibody variants can be monitored over time. We showed the accumulation and clearance time of antibodies in the liver and contralateral site (as a surrogate for nonspecific body clearance) using PAI, which showed that the antibodies cleared from both organs more quickly than the tumor. PAI is useful for comparing lead antibody candidates that are very similar. By comparing them in the same animal, small differences can be identified without the effect of biological differences, complicating the data.

In our experiments, we observed a limitation in using PAI, where antibodies conjugated to different fluorescent dyes may clear differently; however, this can be normalized by using a correction factor. We showed that the coefficient of variation was significantly higher when assessing tumor accumulation between antibodies when analyzing them in different animals versus the same animal. A power analysis showed that to get a power of 0.8 for tumor accumulation between K4 and K5 with nimotuzumab, using each probe in a different animal requires three times the number of mice compared to analyzing the probes in the same animal.

Typically, antibodies are compared in imaging studies using an average sample size of 3–5 animals to achieve statistical significance. However, sometimes this is not achievable due to differences in antibodies and biological variation between animals and their tumors. By coinjecting antibodies, you can reduce the number of animals while improving the results. PAI can be used to reduce the number of mice required without compromising the integrity of the data collected.

In general, we recommend using an untargeted control antibody in PAI because it avoids competition with the targeted antibody, enables accurate background correction, and supports validation of the imaging specificity. Although the control antibody is essential for normalization, it does not provide a biologically relevant signal and must be carefully matched in isotype, size, and labeling to ensure valid comparisons. In other contexts, particularly when evaluating affinity-matured antibodies, the parent antibody is more appropriate for assessing biologically meaningful uptake and distribution. The parent antibody offers insight into target expression and localization, though its signal may be influenced by off-target interactions or nonspecific binding, which should be accounted for during analysis.

PAI has been used with en face sectioning in mice by injecting probes a few hours prior to surgical resection. The excised tissues were examined to distinguish between tumor and normal tissues to assess tumor margins with very high accuracy. PAI and face sectioning can be used as screening tools for margin evaluation after tumor resection.

Conclusions

PAI has many advantages over single-agent imaging. It can normalize biological variability in animals due to blood flow or tissue permeability, which can affect molecule delivery and retention, thereby improving accuracy. PAI can also be incorporated into preclinical applications to compare lead antibodies by allowing longitudinal studies. Using PAI in longitudinal animal studies, compared to traditional biodistribution and pharmacokinetics studies, is an effective way to directly assess molecules in the same biological background while reducing animal use, reagents, time, and cost. In addition, the statistical power is increased with fewer animals.

Glossary

Abbreviations

EGFR

epidermal growth factor receptor

K_D

dissociation constant

NIR

near-infrared

SPR

surface plasmon resonance

PAI

paired agent imaging

MBP

maltose-binding protein

PBS

phosphate-buffered saline

MFI

mean fluorescent intensity

FITC

fluorescein

DMEM

Dulbecco’s Modified Eagle’s Medium

RPMI

Roswell Park Memorial Institute

FBS

fetal bovine serum

ROI

regions of interest

ANOVA

analysis of variance

SEM

standard error of the mean

EPR

enhanced permeability and retention

GI

gastrointestinal

CV

coefficient of variance

Data will be made available upon request.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.molpharmaceut.5c01261.

• Ratio of tumor signal 800/700 in mouse xenografts injected with nimotuzumab, showing the difference between the raw and normalized data. Imaging of regular diet compared to purified diet (PDF)

W.B.: Conceptualization, writingoriginal draft, review and editing, data curation, formal analysis, investigation, methodology, project administration, validation, visualization. K.B.: Review and editing, formal analysis, methodology, visualization. Y.T.: Methodology, validation. D.T.: Methodology, investigation. K.A.J.: Investigation. G.R. : Writingreview and editing, resources. C.R.G.: Conceptualization, resources, supervision, project administration, writingreview and editing

This work was supported, in part, by Grant Name (No. 123456).

The authors declare no competing financial interest.