A real-time in vitro assay as a potential predictor of in vivo tumor imaging properties

Nuclear Medicine and Biology 43 (2016) 12–18

Contents lists available at ScienceDirect

Nuclear Medicine and Biology journal homepage: www.elsevier.com/locate/nucmedbio

A real-time in vitro assay as a potential predictor of in vivo tumor imaging properties Diana Spiegelberg a,⁎, Jonas Stenberg a,b, Anna-Karin Haylock a,c, Marika Nestor a,c a b c

Department of Immunology, Genetics and Pathology, Uppsala University, Uppsala, Sweden Ridgeview Instruments AB, Vänge, Sweden Unit of Otolaryngology and Head & Neck Surgery, Department of Surgical Sciences, Uppsala University, Uppsala, Sweden

a r t i c l e

i n f o

Article history: Received 1 July 2015 Received in revised form 4 September 2015 Accepted 11 September 2015 Keywords: Radio-immunotargeting Radio-immunodiagnostics Molecular imaging AbD15179 CD44v6 HNSCC

a b s t r a c t Introduction: Selective tumor targeting strategies based on cell surface molecules enable new personalized diagnosis and treatments, potentially lowering adverse effects and increasing efficacy. Radio-immunotargeting generally relies on a molecule binding to a cancer-specific target. It is therefore important to understand the properties of molecular interactions in their working environment and how to translate these properties measured in vitro into the in vivo molecular imaging situation. Methods: Time resolved interaction analysis in vitro was compared with a corresponding in vivo xenograft mouse model. The antibody fragment AbD15179 was labeled with 125I or 111In, and analyzed on cell lines with differing CD44v6 expression in vitro, and in a dual tumor xenograft model derived from the same cell lines. In vitro LigandTracer measurements were analyzed with TraceDrawer and Interaction Map. Conjugate sensitivity, kinetics, and signal-to-background ratios were assessed for both tumor cells in vitro and xenograft tumors in vivo. Results: In vitro results revealed a general biphasic appearance of a high- and a low-affinity interaction event. The 111 In-labeled fragment displayed the largest proportion of the high-affinity interaction with increased sensitivity and retention compared to 125I-Fab. In vivo results were in agreement with in vitro data, with increased retention, higher sensitivity and better contrast for the 111In-labeled fragment compared to 125I. Conclusions: Time resolved binding characteristics measured in vitro largely matched the in vivo performance for the conjugates, which is promising for future studies. In vitro time-resolved LigandTracer assays are efficient, rapid, and in this study shown to be able to predict in vivo outcomes. Advances in Knowledge and Implications for Patient Care: Further studies are needed to confirm these findings, but the method is promising considering the ethical need to reduce the use of laboratory animals, as well as reducing costs for the development of tumor targeting compounds in the future. © 2015 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction There is currently an ongoing paradigm shift in drug discovery and development, and it is becoming increasingly important to develop new in vitro models for studying pharmacokinetics and predicting the in vivo behavior of drugs [1]. Such methodology, if properly applied and with data of adequate predictive power, could help to substantially reduce pain and discomfort as well as reduce the number of animals required in preclinical research [2]. The challenge is, however, to ensure that the results obtained by such in vitro methods correlate with the in vivo situation. Many in vitro models used today remain poor estimators of in vivo behavior, potentially leading to inaccurate predictions of in vivo drug interactions [2–5]. This discrepancy may prevent an otherwise suitable drug candidate from advancing through the development ⁎ Corresponding author at: The Rudbeck Laboratory, SE-751 85 Uppsala, Sweden. Tel.: +46 184713868; fax: +46 184713432. E-mail address: [email protected] (D. Spiegelberg).

process, or allow a less suitable candidate to advance and thus waste development resources. Radio-immunotargeting generally relies on radiolabeled tracers binding to specific target antigens, for example receptors overexpressed on tumor cells. Not surprisingly, physical interaction characteristics such as binding strength, retention time and selectivity of target binding versus background are absolutely critical parameters for the diagnostic performance. Hence, methods for quantifying and comparing strengths of molecular interactions are fundamental in the development of new targeting compounds. Such interactions are highly dependent on the nature of the environment in which they take place, which in molecular imaging applications often are highly complex cell surfaces. Thus, the choice of interaction assay is restricted by the capability of the assay to mimic the in vivo situation. In order to develop safe and effective drugs, it is important to accurately estimate pharmacokinetic parameters as early as possible, and traditionally the golden standard has been in vivo assays. To mimic the human environment the vast majority of in vivo pharmacokinetic

http://dx.doi.org/10.1016/j.nucmedbio.2015.09.004 0969-8051/© 2015 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

D. Spiegelberg et al. / Nuclear Medicine and Biology 43 (2016) 12–18

studies today are performed in rodents, especially in mice. For anticancer agents, tumor-xenografts are most commonly used, where human cancer cells are injected subcutaneously in the flank of immunodeficient nude mice. However, not only animal research is costly and timeconsuming, but it also suffers from inherent species related differences and strict limitations on sample size due to ethical considerations. Considering these difficulties, it is apparent why reliable in vitro model systems that accurately predict the pharmacokinetics of drugs are of increasing need in cancer research [6,7]. LigandTracer (Ridgeview Instruments AB (RIAB), Vänge, Sweden) is a novel in vitro method that measures how labeled proteins interact with target antigens on living cells over time, generating binding traces that describe kinetic properties of the interaction. This technology and its validation have previously been described in detail [8,9]. Real-time interaction data from LigandTracer measurements can be analyzed with Interaction Map® (RIAB) [10–12]. This mathematical analysis method translates binding curves to a sum of interactions, each describing one ligand binding to one target. Results are depicted in a modified on–off map and provided as kinetic parameters such as the on-rate (ka), off-rate (kd) and the affinity (KD), as well as the contribution or weight of each interaction. In this paper we address for the first time how accurately this in vitro time-resolved cell binding assay corresponds to outcomes in mouse xenografts when investigating new tracers for tumor targeting. Such in vitro evaluations can provide an initial characterization of binding (e.g., on-rate, off-rate, affinity), quality of labeled conjugate (e.g., occurrence/increase of low-affinity populations), and a chance to assess interaction differences between standard labeling methods before moving on to an in vivo model. Since the binding characterizations are made on living cells instead of using fixed antigens on a chip, we believe that this is closer to an in vivo setting than for example surface plasmon resonance (SPR) techniques. Furthermore, the characterizations are made in real-time and not, as many standard radioligand assays, by using endpoint methods such as saturation or displacement binding experiments, where binding is in equilibrium at the time of measurement [13–15]. The cell surface antigen CD44v6 was used as a cancer-target model, since it is a non-internalizing antigen commonly overexpressed in several cancer types, including head and neck squamous cell carcinoma (HNSCC) [16,17]. Well performing targeting molecules that bind to such antigens may be utilized as tools for the challenging tasks of targeted treatments of small residual disease, and are therefore of great importance. Early detection with a tumor specific medical imaging method such as radio-immunodiagnostics, combining the high sensitivity and resolution of a PET (positron emission tomography) or SPECT (single-photon emission computed tomography) camera with antibody tumor specificity, could potentially improve the management of patients with head and neck malignancies. The novel CD44v6 targeting recombinant Fab-fragment AbD15179 was chosen as a tracer due to its previously proven performance [18]. Even though interaction analysis of time-resolved binding data obtained with LigandTracer has been used to study several radiotracers in vitro, no previous study has assessed to what extent this in vitro assay can be used as a potential predictor of in vivo tumor imaging properties. If such an assay could reduce the need for animal screening of multiple lead radiotracers, this could significantly facilitate in vitro to in vivo translation in radio-immunodiagnostics. Thus, in this study, time resolved interaction analysis in vitro was compared with results from an in vivo xenograft mouse model. AbD15179 was labeled with either 125I or 111In, and conjugates from the same labeling batch were analyzed on two cell lines with differing CD44v6 expression in vitro, and in a dual tumor xenograft model derived from the same cell lines in vivo. Evaluation of factors such as conjugate sensitivity, kinetics, and signal-to-background ratios was then assessed for both cultured tumor cells in vitro and xenograft tumors in vivo.

13

2. Materials and methods 2.1. Antibody fragments The anti-CD44v6 Fab-fragment AbD15179 was supplied from AbD Serotec. Selection and production of the antibody has been described previously [19]. The Fab fragment was supplied in 3 × PBS and stored in − 80 °C. Fragments for 111In-labeling were first separated by sizeexclusion chromatography on a NAP-5 column (Amersham Biosciences, Uppsala, Sweden) preequilibrated with purified (MilliQ) water. It was then freeze-dried and stored at −20 °C before use in order to facilitate buffer exchange and concentration adjustments. 2.2. Cell lines The HNSCC human cell line H314 (obtained from European Collection of Cell Cultures) was cultured in a 1:1 mixture of Ham's F12 and Dulbecco's modified Eagle medium (DMEM), supplemented with 10% fetal calf serum, 2 mM L-glutamine, and antibiotics (100 IU penicillin and 100 μg/ml streptomycin). H314 has an average antigen density of approximately 650.000 CD44v6 receptors per cell [10]. The SCC (squamous cell carcinoma) cell line A431 (obtained from American Type Culture Collection) was cultured in Ham's F10, with the same supplements. A431 has an average antigen density of approximately 3,000,000 CD44v6 receptors per cell [10]. When used in LigandTracer studies, cells were seeded on a local part of a 10-cm cell dish (#150350; Nunc) and allowed to adhere firmly to the plastics prior to the addition of fresh medium. Experiments were conducted 2–3 days after seeding. 2.3. Labeling of the Fab fragments Antibody fragments were labeled with 125I (PerkinElmer, Waltham, MA, USA) using direct chloramine T labeling [20] (Sigma Aldrich) as described previously [19] In brief, a stock solution of 10 MBq 125I was mixed with 100 μl PBS and 14 μM of the Fab-fragment (140 μl, 0.72 mg/ml in PBS) was added. The reaction was initiated by addition of 18 mM chloramine T (Sigma Aldrich) in water (40 μl, 4 mg/ml) and was quenched after 5-min incubation on ice by addition of 21 mM sodium metabisulphite (Merck) in water (80 μl, 4 mg/ml). Each labeled protein was separated from non-reacted 125I and low-molecular-weight reaction components on a NAP-5 column pre-equilibrated with PBS. 125 I-labeled AbD15179 is referred to as 125I-Fab in this paper. Labeling of AbD15179 with 111In (Mallinckrodt Medical B.V.) was done using CHX-A″-DTPA labeling as described previously [10,21]. In short, to 97 μM AbD15179 solution (15 μl, 5 mg/ml in borate buffer, pH 9), 4.1 μl of CHX-A″-DTPA solution (1 mg/ml in borate buffer, 0.07 M, pH 9) was added, corresponding to a chelator-to-AbD15179 molar ratio of 5:1. The reaction mixture was incubated overnight at room temperature, and the conjugated antibody was then separated from free CHX-A″-DTPA using a NAP-5 column equilibrated with acetate buffer (0.2 M, pH 5.5). Approximately 10 MBq of 111In in acetate buffer was then added to the conjugated AbD15179, and the reaction mixture was allowed to incubate for 30 min at room temperature. Labeled Fab fragment was then separated from non-reacted radionuclide and low-molecular-weight reaction components by using a NAP-5 column pre-equilibrated with PBS. 111In-labeled AbD15179 is referred to as 111In-Fab in this paper. To determine the purity and stability of the labeled antibody fragments over time, instant thin-layer chromatography (ITLC) was performed on 125I and 111In-labeled conjugates. Samples taken immediately, as well as stored in PBS at 4 °C, or in serum at 37 °C, 24 h after the labeling procedure was analyzed. Approximately 1 μl of the conjugate was placed on a chromatography strip (Biodex) and put into a “running buffer” (70 % acetone or 5 nM citric acid for 125I and 111 In respectively), followed by measurements on a Cyclone Storage

14

D. Spiegelberg et al. / Nuclear Medicine and Biology 43 (2016) 12–18

Phosphor System. Data were analyzed using the OptiQuant image analysis software. 2.4. Real-time binding measurements on SCC cells in vitro Binding and retention measurements of the radiolabeled conjugates on H314 and A431 cells were performed in room temperature in vitro with LigandTracer according to a previously described protocol [10]. No internalization of these conjugates in vitro was observed in a simultaneous study, even after 24 h at 37 °C [18], hence the kinetics for surface receptor binding alone could be detected. Experiments were run simultaneously to allow the use of the same 125I-Fab and 111In-Fab batches for both in vitro and in vivo experiments. 125I-Fab was measured in LigandTracer Grey and 111In-Fab in LigandTracer Yellow. Binding traces in room temperature using three subsequent concentrations (10, 30 and 90 nM) of Fab fragments were obtained for at least 1 h per concentration, followed by a dissociation measurement for at least 13 h. Interaction data were then evaluated using TraceDrawer and Interaction Map software. For calculations on conjugate retention, conjugate sensitivity and cell to background uptake ratios, data on signal intensities (counts per second) from each run were exported and aligned in Microsoft Excel (Microsoft Office 2011, Redmond, CA, USA). For estimations on conjugate retention at selected time points, five subsequent data points at the chosen times during the dissociation phase were collected from each measurement, normalized to the start of the dissociation phase. Conjugate sensitivity was calculated by dividing obtained A431 cell signals (background corrected) with H314 cell signals for each corresponding time point. Cell to background uptake ratios were calculated by collecting the last five data points from each binding measurement phase (10 nM, 30 nM, 90 nM) where binding was observed to be in equilibrium, and at the end of the dissociation phase (approximately 13 h after start of dissociation), and dividing the signal from the cell area to the corresponding background area for each dish. See section on statistics. 2.5. Biodistribution Tumor xenografts were formed by injection of A431 and H314 cells in 20 balb/c (nu/nu) female mice. The mice were housed in a controlled environment and fed ad libitum. Experiments complied with current Swedish law and were performed with permission granted by the Uppsala Committee of Animal Research Ethics. All mice received two injections, one in the right posterior leg and one in the left posterior leg. Twenty mice were injected with 8 × 10 6 A431 cells, resuspended in matrigel, in the right posterior leg and 10 × 10 6 H314 cells, resuspended in complete medium, in the left posterior leg. Three weeks after tumor cell injection they received an intravenous injection (i.v.) via the tail vein with 125I-Fab (100 kBq, specific activity of 20 kBq/μg (28 Ci/ mmol)) and 111In-Fab (100 kBq, specific activity of 20 kBq/μg (28 Ci/ mmol)), totally 5 μg Fab conjugates in 200 μl PBS (483 nM). At time points 6 (N = 5), 24 (N = 5), 48 (N = 5) and 72 (N = 5) hours the animals were euthanized with a mixture of ketamine and xylazin followed by heart puncture. Tumors and blood were excised, weighed and measured in a gamma well counter (1480 WIZARD; Wallace Oy). Three injection standards were measured for each group, and background was measured for 111In. Radioactivity uptake in organs was calculated as percent of injected activity per gram of tissue (%ID/g). Tumor to organ ratio was calculated as activity/gtumor divided by activity/gorgan. For assessment of conjugate sensitivity, ratios for A431 tumor uptake were divided with H314 tumor uptake in each individual mouse. See section on statistics. 2.6. Statistical analyses Kinetic data obtained from LigandTracer studies were analyzed in TraceDrawer 1.6.1 (Ridgeview Instruments AB, Vänge, Sweden) and

using the Interaction Map kinetic analysis method, as described previously [11,12,19,22]. Statistical data comparisons were performed using GraphPad Prism Version 5 (Graphpad Software Inc., La Jolla, CA, USA). Comparisons of kinetic parameters (high affinity interaction vs low affinity interaction), as well as comparisons between 125I-Fab and 111InFab regarding retention and A431/H314 ratios in vitro and in vivo were performed using a two-tailed t-test and were considered statistically significant if p b 0.05. Comparisons between several groups ( 111In-Fab on A431 vs. 111In-Fab on H314, 125I-Fab on A431, and 125I-Fab on H314 cells) regarding cells to background (in vitro) or tumor to blood uptake (in vivo) ratios were performed using one-way ANOVA followed by Tukey's multiple comparison test, and were considered statistically significant if p b 0.05. Unless otherwise stated, data are presented as the mean ± standard deviation (SD). N = number of replicates. 3. Results 3.1. Radiolabeling of AbD15179 The labeling yield for 125I-labeling of AbD15179 was 74%, and the purity of labeled protein was N99.9% according to ITLC analysis. For AbD15179 111In-labeling the yield and purity were 63% and N96%, respectively. The purity of labeled conjugates stored at 4 °C in PBS, or in serum at 37 °C, during the experimental period was re-analyzed after 48 h, and was unchanged according to the ITLC analysis. 3.2. In vitro kinetic analysis of labeled conjugates Real-time in vitro measurements of conjugate–cell interactions were initially evaluated for binding kinetics using TraceDrawer and Interaction Map calculations. An on/off rate plot of the interactions can be seen in Fig. 1A, and detailed kinetic parameters are displayed in Fig. 1B. In line with previous studies [10], binding of radiolabeled AbD15179 to cells mainly consisted of two interactions; one highaffinity (KD = 11.6 ± 4.96 nM, N = 4) and one low-affinity (KD = 218 ± 94.5 nM, N = 4) binding event (depicted as a red and a black cluster respectively in Fig. 1A). Both 125I-Fab and 111In-Fab produced this biphasic binding behavior in both cell lines. The kinetic parameters ka kd and KD differed significantly (p b 0.05) between the high and low interactions (p b 0.05). Furthermore, in line with previous studies [10] the proportion of high-affinity binders was considerably higher for 111 In-Fab (91% and 93% on H314 and A431 cells respectively) than for 125 I-Fab (62% and 63% on H314 and A431 cells respectively) for the used concentration span (Fig. 1B). 3.3. Dissociation of conjugates in vitro vs. in vivo The contrast of signal in H314 and A431 cells was assessed for 125Iand 111In-labeled conjugates both in vitro (LigandTracer) and in vivo (mouse xenografts). LigandTracer dissociation curves for Fab-fragment AbD15179 labeled with 125I and 111In in H314 and A431 cells can be seen in Fig. 2A and B respectively. In vitro, 125I-Fab and 111In-Fab follow the same pattern in both cell lines, with an initially faster off-rate for 125 I-Fab than for 111In-Fab, resulting in a superior retention for 111InFab. For example, one hour after start of dissociation measurements, 90% ± 1.5% (H314 cells) and 94% ± 1.3% (A431 cells) of 111In-Fab were still retained on the cells, whereas only 74% ± 2.8 % (H314 cells) and 79% ± 4.3% (A431 cells) of 125I-Fab were retained. Ten hours after start of dissociation measurements, 70% ± 0.5% (H314 cells) and 79% ± 0.4% (A431 cells) were still retained of 111In-Fab, whereas 51% ± 1.1% (H314 cells) and 61% ± 0.8% (A431 cells) were retained of 125 I-Fab. Retention of 111In-Fab was significantly higher than 125I-Fab at all assessed time points (1, 5 and 10 h after dissociation start, p b 0.0001, N = 5). A similar pattern could be seen in vivo (Fig. 2C and D). 125I-Fab displayed a more rapid initial dissociation in both investigated tumor types compared to 111In-Fab, which demonstrated a greater retention

D. Spiegelberg et al. / Nuclear Medicine and Biology 43 (2016) 12–18

15

Fig. 1. Kinetics of conjugates binding to CD44v6 on A431 and H314 cells, as measured in vitro with LigandTracer. Binding of radiolabeled AbD15179 to cells mainly consisted of two interactions; one high-affinity and one low-affinity binding event. (A) Logarithmic on/off rate plot of the association rate constants ka and dissociation rate constants kd of 125I-Fab and 111In-Fab binding to A431 and H314 cells. Red dots represent high-affinity and black dots correspond to low-affinity binding populations. C1/C2 = 125I-Fab–A431, D1/D2 = 111In-Fab–A431, E1/ E2 = 125I-Fab–H314 and F1/F2 = 111In-Fab–H314 interactions. (B) Results from kinetic fits using the One-To-Two model in TraceDrawer and Interaction Map calculations. Two distinct binding populations were present for all conjugates; one high-affinity (KD = 11.6 ± 4.96 nM) and one low-affinity (KD = 218 ± 94.5 nM) binding event. 111In-Fab comprised a larger proportion of high-affinity binders than 125I-Fab as determined by Interaction Map weight calculations.

of conjugates. 111In-Fab levels remained high in tumors for all time points, while 125I-Fab uptake was rapidly reduced with time. At 24 h post-injection, 90% ± 24% of the 6 h value of 111In-Fab was still present in H314 tumors, compared to 45% ± 7.5% for 125I-Fab. In A431 tumors, 72% ± 15% and 39% ± 5.8% were retained of 111In-Fab and 125I-Fab respectively. Even 72 h post-injection, 111In-Fab levels were unchanged in H314 tumors (100% ± 38%) while 125I-Fab had decreased to 28% ± 12%. In A431 tumors, 75% ± 19% of the initial 111In-Fab levels were retained, compared to only 14% ± 2% of the initial 125I-Fab levels at 72 h post-injection. Retention of 111In-Fab was significantly higher than 125I-Fab at all assessed time points (24, 48 and 72 h p.i., p b 0.05, N = 5).

A

3.4. Sensitivity of conjugates in vitro vs. in vivo Differences in conjugate sensitivity of the in vitro and in vivo studies are shown in Fig. 3, displayed as A431/H314 signal ratios. In vitro measurements demonstrated significantly higher ratios for 111In-Fab compared to 125I-Fab (p b 0.0001) (Fig. 3A). During binding measurements using 10 and 30 nM conjugate, 111In-Fab A431/H314 ratios were around two, whereas incubation with the highest concentration (90 nM) increased the ratios further, and remained around three throughout the dissociation measurement. 125I-Fab displayed a ratio around one throughout the analysis, and it was not possible to distinguish between A431 and H314 cells at the assessed concentrations. A comparable

B

A431 cells

% of max uptake

% of max uptake

H314 cells

time (h)

C

time (h)

D

H314 tumors

A431 tumors 150

* 100

**

**

50

% of max uptake

% of max uptake

150

100

* ****

** 50

0

0 6

12 18 24 30 36 42 48 54 60 66 72

time p.i. (h)

6

12 18 24 30 36 42 48 54 60 66 72

time p.i. (h)

Fig. 2. Dissociation of 111In-Fab (orange) and 125I-Fab (black) conjugates in vitro (A-B) and in vivo (C-D). Both in vitro and in vivo studies revealed a more rapid release of 125I-Fab than of 111 In-Fab from tumor cells. Top: In vitro results from LigandTracer retention measurements after withdrawal of ligand and replacement with complete medium, preceded by 6 h of uptake measurements, on (A) H314 cells and (B) A431 cells. Bottom: In vivo analysis of the conjugates in mice xenografts bearing (C) H314 tumors and (D) A431 tumors in the same subject, demonstrating significantly faster dissociation of 125I-Fab compared to 111In-Fab, analyzed using two-tailed t test. Error bars = SD, N = 5.

16

D. Spiegelberg et al. / Nuclear Medicine and Biology 43 (2016) 12–18

A

Ratio A431/H314

**** 2

in vivo 3

****

**** Ratio A431/H314

B

in vitro 3

****

1

* 2

**

1

0

0 0

5

10

15

20

0

20

40

time (h)

60

80

time (h)

Fig. 3. In vitro (A) and in vivo (B) analysis on conjugate sensitivity for 111In-Fab (orange) and 125I-Fab (black) by calculating the A431/H314 uptake ratios. Both in vitro and in vivo studies demonstrated a clearer difference in signal intensity between A431 and H314 cells for the 111In-Fab fragment. (A) Data from LigandTracer measurements on H314 and A431 cells, where binding traces using three subsequent concentrations (10, 30, 90 nM) were obtained for at least 1 h per concentration (marked by dotted lines), followed by a dissociation measurement (0 nM) for at least 15 h. (B) Data from in vivo analysis of the conjugates in mice xenografts bearing H314 tumors and A431 tumors. At 6 and 72 h p.i., 111In-Fab displayed a significantly higher A431/H314 ratio than 125I-Fab in vivo, analyzed using two-tailed t test. Error bars = SD, N = 5.

outcome was seen in vivo, where differences in 111In-Fab levels between A431 and H314 tumors were more apparent than for 125I-Fab, with significantly higher A431/H314 ratios at 6 h (p = 0.016) and 72 h (p = 0.0016) p.i. (Fig. 3B). The difference in tumor uptake for the 125I-Fab fragment was diminished at later time points (A431/H314 ratio 0.98 ± 0.10 at 72 h p.i., N = 5), whereas 111In-Fab retained a clear difference in tumor uptake, ranging between 1.5 and 2 times higher for the high CD44v6-expressing A431 tumors throughout the study (A431/H314 ratio 1.5 ± 0.2 at 72 h p.i., N = 5). Characteristics of in vivo and in vitro measurements of the 125I-Fab and 111In-Fab were also evaluated by calculating the tumor cell to background ratios in vitro and tumor to blood ratios in vivo. As shown in Fig. 4, in vitro and in vivo measurements both demonstrated the highest ratio for the 111In-Fab A431 model. In vitro, the 111In-Fab A431 model was significantly higher compared to the other models at all assessed concentrations (p b 0.001) (Fig. 4A). Cell to background ratios were highest during the dissociation phase (measured approximately 15 h after start of dissociation), where the low amount of unbound conjugate present in the incubation medium is expected to generate lower background. In vivo, the 111In-Fab A431 model demonstrated a significantly higher ratio compared to the other models at 24 h (p b 0.05), 48 h (p b 0.01) and 72 h (p b 0.05) p.i., with the exception of 125I-Fab in A431 tumors 24 h p.i., and 111In-Fab in H314 tumors at 72 h p.i. (Fig. 4B). At 72 h p.i., ratio was 13 ± 3.2 for the 111In-Fab A431 model,

A

compared to 8.5 ± 1.2 for 111In-Fab H314, 4.3 ± 0.8 for and 4.6 ± 1.2 for 125I-Fab H314.

I-Fab A431

4. Discussion The main purpose of this study was to address whether time resolved assays using LigandTracer can complement and possibly replace some animal studies in pre-clinical development of new tumor targeting compounds for radio-immunotargeting. To our knowledge, little has so far been done to systematically validate this in vitro model and demonstrate the extent to which it mimics in vivo kinetics. The possible in vitro–in vivo correlation was investigated by assessing the kinetics of radiolabeled AbD15179 in two squamous carcinoma cell lines of different CD44v6 expression. Conjugates from the same labeling batch were simultaneously characterized both in vitro with LigandTracer, and in vivo using tumor bearing mice, with tumors originating from the same cell lines used for in vitro assessment. Simultaneous characterization of the same conjugates both in vitro and in vivo reduced potential bias from, e.g., batch variability, specific activity or immunoreactivity. Moreover, by simultaneously injecting the two differently labeled tracers into mice carrying double SCC tumors with different CD44v6 expression, biodistribution, tumor uptake, contrast, as well as specificity and sensitivity of the tracers could be evaluated while eliminating inter-subject variability in vivo.

B

in vitro

125

in vivo **

15

*

**

tumor to blood ratio

tumor cell to background ratio

8

6

4

***

***

***

2

0

10

**

5 *

*

0 10

30

90

added concentration (nM)

0

6

48

24

72

time (h)

Fig. 4. Comparison of tumor cell to background in vitro (A), and tumor to blood ratios in vivo (B) using 111In-Fab (orange) and 125I-Fab (black) in A431 (filled bars) and H314 (striped bars) models. Both in vitro and in vivo measurements demonstrated the highest ratio for the 111In-Fab A431 model. Error bars = SD, N = 5. Asterisks mark if 111In-Fab on A431 cells (A) or tumors (B) was significantly higher than other groups, as calculated by one-way ANOVA with *p b 0.05, **p b 0.01, and ***p b 0.001. When only the 111In-Fab on A431 bar is marked, ratio was significantly different from all groups at that concentration or time point.

D. Spiegelberg et al. / Nuclear Medicine and Biology 43 (2016) 12–18

In the present study, the dissociation of both conjugates was first investigated in vitro and in vivo. LigandTracer results demonstrated clear differences between 125I-Fab and 111In-Fab in vitro, with a slower release of conjugates for 111In-Fab than for 125I-Fab in both H314 and A431 cells (Fig. 2). This can be attributed to the considerably larger population of high-affinity binders observed for 111In-Fab than for 125I-Fab (91%–93% and 62%–63% respectively, see Fig. 1) and as previously shown to differ significantly between 111In-Fab and 125I-Fab [10], resulting in higher retention of 111In-Fab compared to 125I-Fab on SCC cells. The larger population of low-affinity 125I-Fab conjugates may be explained by the relatively harsh conditions of direct 125I-labeling with chloramine T as shown previously [10]. The slow dissociation of 111 In-Fab was also in line with the in vivo data, where the 111In-Fab resulted in a high and stable tumor signal, significantly higher compared to 125I-Fab in both tumor models. However, it cannot be fully excluded that some of the retention for 111In-Fab seen in vivo could be due to slow non-receptor mediated internalization of the conjugates by the tumor cells as the cellular membrane renews itself, trapping this residualizing radiolabel. In contrast, endocytosis of directly labeled 125I-Fab could lead to deiodination and excretion of iodotyrosine or free iodine [25]. However, previous in vitro studies have shown that CD44v6 in general is not an internalizing receptor [23,24], and that AbD15179 in particular is not internalized into either H314 or A431 tumor cells [18]. Thus, we hypothesize that the major cause of the initial differences in tumor retention in vivo was due to the larger population of high-affinity binders for 111In-Fab than for 125I-Fab. Next, sensitivity of the tracers was compared by assessing the ability of the conjugates to discern moderately CD44v6-expressing H314 cells from highly CD44v6-expressing A431 cells. The in vitro and in vivo results were in agreement, where the 111In labeled Fab fragment showed significant differences in signal intensity between high- and medium CD44v6-expressing cells and tumors, whereas the 125I-labeled Fab fragment could not sufficiently distinguish between the two cell lines at the concentrations used (Fig. 3). For molecular imaging, this can be an important parameter for assessment of biomarker density, for tumor characterization, diagnosis, dosing and patient stratification, and it is encouraging that the in vitro results correlated so closely to the in vivo results. Furthermore, tumor cell to background in vitro and tumor to blood ratios in vivo were compared (Fig. 4). Our hypothesis was that the in vitro situation, where tumor cells are first exposed to unbound conjugates present in the incubation medium, followed by withdrawal of ligand and replacement with medium alone, to some extent resembles the situation of tumors in vivo, where tumor cells are first exposed to large amounts of unbound conjugates in blood, followed by blood clearance. The in vitro results demonstrated that the highest tumor cell to background ratio was obtained for 111In-Fab in the A431 model. In vivo, 111In-Fab again displayed a significant tumor delivery advantage compared to 125I-Fab, with the highest tumor to blood ratio obtained in the A431 tumors. Even though the in vitro method does not take factors such as tumor vascularization, permeability and blood clearance into account, it is promising that in vitro and in vivo results were in agreement. However, more studies are needed to further explore to what extent in vitro tumor cell to background ratios can be used to predict in vivo tumor to blood ratios. When evaluating results as a whole, the comparison between in vitro time-resolved data and the in vivo mouse xenograft model clearly showed a strong in vitro–in vivo correlation for the assessed parameters. Due to experimental necessities, measurement times and signal magnitudes were different between in vivo and in vitro assays, however the general shape of the data and performance of each conjugate correlated strongly. Had the in vivo data not been available, we believe that a reasonable estimation of the comparative tumor targeting performance of the different labeled conjugates could have been formed based on the in vitro data alone. It is important to note however, that many aspects of in vivo experiments simply currently cannot be addressed in cell

17

based in vitro studies, no matter how complex the model system is. Today, we are still far away from the point where all necessary information required for introducing a new therapeutic drug or imaging agent into the clinic in terms of safety and efficacy could be gained without the use of laboratory animals [26]. Moreover, in terms of interactions on the level of organs and general physiology; in vivo stability of the conjugate, renal clearance rates, toxicity etc, whole-organism models remain absolutely essential. However, with a careful choice of analysis method we can achieve an important informative base on what will occur in vivo, especially concerning antigen binding kinetics; this can enable us to design smaller and yet more informative animal experiments, and exclude unsuitable conjugates already at the in vitro level. With that in mind, we believe that assessment of new binders for radio-immunotargeting is likely a good candidate for replacing select animal studies with extended in vitro assays. The performance of a radiolabeled tracer is dependent on parameters relatively easy to measure, but generally not on the biological response from the ligand binding to the target. This is considerably more complicated to examine in in vitro assays; both because of non-linearities of signaling pathways and because of the complex tissue and organ-level interactions which cell assays do not address. In radio-immunodiagnostics, such biological effects are of less importance if you have a good understanding of the ligand interaction parameters (e.g., affinity, Bmax and dissociation), ligand biodistribution, and tumor specific parameters such as target expression levels. This provides a solid base to predict in vivo performance of the tracer in e.g., SPECT or PET applications. Through time-resolved binding analysis methods, data can be collected which even go beyond the capability of in vivo models. Information on the total number of binding events and interaction kinetics can enable the precise fine-tuning of binding characteristics at an early stage. This can be very important, since changing compound specifics and formulation become more complicated the further you reach in the development pipeline. To conclude, this study is one of few addressing the problem of in vitro to in vivo translation in radio-immunodiagnostics. In our case, time-resolved in vitro experiments could adequately predict in vivo tumor targeting performance of radiolabeled binders. In some cases the information from the in vitro assay was of higher quality and resolution than the in vivo data, with in-depth kinetic analysis of results. However, today it is still not feasible to identify an ideal radiotracer based on the in vitro results alone. We believe that the improvement and development of new in vitro techniques and research tools will be critical to accelerate development of the next generation of tumor targeting diagnostic compounds. Furthermore, such a development has the potential of leading to a reduction in the usage of laboratory animals. Such a reduction is naturally of ethical interest, but also essential for reducing the prohibitively high drug development costs; in turn helping to reduce the price, and thereby increasing the availability of modern cancer treatments. More cost-efficient development of new improved diagnostic and therapeutic tools can only serve to benefit society in general and cancer patients in particular.

Conflict of interest J. Stenberg is employed by and a shareholder of Ridgeview Instruments AB.

Acknowledgments and funding The authors would like to acknowledge the Swedish Cancer Society (CAN2012/399), the Swedish Research Council (2013-30876104113-30), and the Swedish Society for Medical Research for kind support.

18

D. Spiegelberg et al. / Nuclear Medicine and Biology 43 (2016) 12–18

References [1] Couch RD, Mott BT. Personalized medicine: changing the paradigm of drug development. Methods Mol Biol 2012;823:367–78. [2] Doke SK, Dhawale SC. Alternatives to animal testing: a review. Saudi Pharm J 2015; 23:223–9. [3] Balls M. Future improvements: replacement in vitro methods. ILAR J 2002;43: S69–73. [4] McKim Jr JM. Building a tiered approach to in vitro predictive toxicity screening: a focus on assays with in vivo relevance. Comb Chem High Throughput Screen 2010;13:188–206. [5] Wienkers LC, Heath TG. Predicting in vivo drug interactions from in vitro drug discovery data. Nature reviews. Drug Discov 2005;4:825–33. [6] Davila JC, Rodriguez RJ, Melchert RB, Acosta Jr D. Predictive value of in vitro model systems in toxicology. Annu Rev Pharmacol Toxicol 1998;38:63–96. [7] Ito K, Iwatsubo T, Kanamitsu S, Nakajima Y, Sugiyama Y. Quantitative prediction of in vivo drug clearance and drug interactions from in vitro data on metabolism, together with binding and transport. Annu Rev Pharmacol Toxicol 1998;38:461–99. [8] Bjorke H, Andersson K. Measuring the affinity of a radioligand with its receptor using a rotating cell dish with in situ reference area. Appl Radiat Isot 2006;64:32–7. [9] Bjorkelund H, Gedda L, Andersson K. Comparing the epidermal growth factor interaction with four different cell lines: intriguing effects imply strong dependency of cellular context. PLoS One 2011;6:e16536. [10] Stenberg J, Spiegelberg D, Karlsson H, Nestor M. Choice of labeling and cell line influences interactions between the Fab fragment AbD15179 and its target antigen CD44v6. Nucl Med Biol 2014;41:140–7. [11] Altschuh D, Bjorkelund H, Strandgard J, Choulier L, Malmqvist M, Andersson K. Deciphering complex protein interaction kinetics using Interaction Map. Biochem Biophys Res Commun 2012;428:74–9. [12] Bjorkelund H, Gedda L, Barta P, Malmqvist M, Andersson K. Gefitinib induces epidermal growth factor receptor dimers which alters the interaction characteristics with (1)(2)(5)I-EGF. PLoS One 2011;6:e24739. [13] Kudo M, Vera DR, Trudeau WL, Stadalnik RC. Validation of in vivo receptor measurements via in vitro radioassay: technetium-99m-galactosyl-neoglycoalbumin as prototype model. J Nucl Med 1991;32:1177–82.

[14] Patel S, Hamill T, Hostetler E, Burns HD, Gibson RE. An in vitro assay for predicting successful imaging radiotracers. Mol Imaging Biol 2003;5:65–71. [15] Motulsky H, Neubig R. Analyzing radioligand binding data. Curr Protoc Protein Sci 2001 21:3H:A.3H.1–A.3H.55. [16] Spiegelberg D, Kuku G, Selvaraju R, Nestor M. Characterization of CD44 variant expression in head and neck squamous cell carcinomas. Tumour Biol 2014;35: 2053–62. [17] Mack B, Gires O. CD44s and CD44v6 expression in head and neck epithelia. PLoS One 2008;3:e3360. [18] Haylock AK, Spiegelberg D, Nilvebrant J, Sandstrom K, Nestor M. In vivo characterization of the novel CD44v6-targeting Fab fragment AbD15179 for molecular imaging of squamous cell carcinoma: a dual-isotope study. EJNMMI Res 2014;4:11. [19] Nilvebrant J, Kuku G, Bjorkelund H, Nestor M. Selection and in vitro characterization of human CD44v6-binding antibody fragments. Biotechnol Appl Biochem 2012;59:367–80. [20] Hunter WM, Greenwood FC. Preparation of iodine-131 labelled human growth hormone of high specific activity. Nature 1962;194:495–6. [21] Nestor M, Andersson K, Lundqvist H. Characterization of (111)In and (177)Lu-labeled antibodies binding to CD44v6 using a novel automated radioimmunoassay. J Mol Recognit 2008;21:179–83. [22] Svitel J, Balbo A, Mariuzza RA, Gonzales NR, Schuck P. Combined affinity and rate constant distributions of ligand populations from experimental surface binding kinetics and equilibria. Biophys J 2003;84:4062–77. [23] Nestor M, Persson M, van Dongen GA, Jensen HJ, Lundqvist H, Anniko M, et al. In vitro evaluation of the astatinated chimeric monoclonal antibody U36, a potential candidate for treatment of head and neck squamous cell carcinoma. Eur J Nucl Med Mol Imaging 2005;32:1296–304. [24] Nestor M, Sundstrom M, Anniko M, Tolmachev V. Effect of cetuximab in combination with alpha-radioimmunotherapy in cultured squamous cell carcinomas. Nucl Med Biol 2011;38:103–12. [25] Opresko L, Wiley HS, Wallace RA. Proteins iodinated by the chloramine-T method appear to be degraded at an abnormally rapid rate after endocytosis. Proc Natl Acad Sci U S A 1980;77:1556–60. [26] Workman P, Aboagye E, Balkwill F, Balmain A, Bruder G, Chaplin D, et al. Guidelines for the welfare and use of animals in cancer research. Br J Cancer 2010;102:1555–77.