Development of Radiofluorinated Nicotinamide/Picolinamide Derivatives as Diagnostic Probes for the Detection of Melanoma

Regarding the increased incidence and high mortality rate of malignant melanoma, practical early-detection methods are essential to improve patients’ clinical outcomes. In this study, we successfully prepared novel picolinamide–benzamide (18F-FPABZA) and nicotinamide–benzamide (18F-FNABZA) conjugates and determined their biological characteristics. The radiochemical yields of 18F-FPABZA and 18F-FNABZA were 26 ± 5% and 1 ± 0.5%, respectively. 18F-FPABZA was more lipophilic (log P = 1.48) than 18F-FNABZA (log P = 0.68). The cellular uptake of 18F-FPABZA in melanotic B16F10 cells was relatively higher than that of 18F-FNABZA at 15 min post-incubation. However, both radiotracers did not retain in amelanotic A375 cells. The tumor-to-muscle ratios of 18F-FPABZA-injected B16F10 tumor-bearing mice increased from 7.6 ± 0.4 at 15 min post-injection (p.i.) to 27.5 ± 16.6 at 3 h p.i., while those administered with 18F-FNABZA did not show a similarly dramatic increase throughout the experimental period. The results obtained from biodistribution studies were consistent with those derived from microPET imaging. This study demonstrated that 18F-FPABZA is a promising melanin-targeting positron emission tomography (PET) probe for melanotic melanoma.


Introduction
Malignant melanoma is the most aggressive skin cancer and accounts for approximately 60% of skin cancer deaths [1,2]. The incidence has continuously increased in recent years, especially in North America, Australia, and New Zealand, by 3-8% annually [3]. Clinically, 18 F-FDG is a standard PET probe to diagnose tumors; however, it could only detect late-stage malignant melanomas due to its non-specificity for melanin. In addition, 18 F-FDG PET imaging demonstrates relatively high false-positive findings caused by inflammation, pneumonia, infectious processes, and other organs with increased glucose metabolism [4]. The 5-year survival rate for patients with early-stage malignant melanoma receiving surgery (≈92%) was significantly higher than those having distant metastases (≈22.5%) [2]; thus, developing a reliable diagnostic probe for early and accurate detection is crucial for the following treatment.

The Preparation of 18 F-FPABZA and 18 F-FNABZA
The F-18 was labeled to compounds 2a and 2b by the nucleophilic substitution (SN2) reaction. The radiolabeling efficiency of 18 F-FPABZA and 18 F-FNABZA was 65 ± 5% and 10 ± 5%, respectively (Figure 2A,B). After purification, both compounds' radiochemical purities were greater than 98% ( Figure 2C,D). The radiochemical yields of 18 F-FPABZA and 18 F-FNABZA were 25 ± 5% and less than 2%, respectively. We previously demonstrated that the in vivo accumulation of 131 I-labeled nicotinamidebenzamide derivative, 131 I-iochlonicotinamide ( 131 I-ICNA), in B16F10 melanoma and liver was 13.48 ± 1.77 and 4.17 ± 0.76 %ID/g at 1 h after administration, respectively [13], suggesting that the conjugation with a heterocyclic compound can effectively increase the clearance of benzamide from the normal tissues to cause high tumor-to-background contrast. We also reported that both 131 I-iodofluoropicolinamide benzamide ( 131 I-IFPABZA, Figure 1) and 131 I-iodofluoronicotiamide benzamide ( 131 I-IFNABZA, Figure 1) exhibit superior in vivo targeting ability against melanotic B16F10 melanoma [14]. To have a better imaging resolution, this study aims to prepare novel 18 F-labeled picolinamide-benzamide ( 18 F-FPABZA) and nicotinamide-benzamide ( 18 F-FNABZA) derivatives and to determine their feasibility as a diagnostic agent for melanin-expressed melanoma.

Figure 2.
Radio-TLC analysis of (A,B) the crude products and (C,D) the purified 18 F-FPABZA and 18 F-FNABZA. The radiolabeling efficiency and the radiochemical purity were determined by radio-TLC using MeOH/CH 2 Cl 2 = 1:3 (v/v) as the eluting agent.

Partition Coefficient Determination and In Vitro Stability of 18 F-FPABZA or 18 F-FNABZA
The Log P values of 18 F-FPABZA and 18 F-FNABZA were 1.48 ± 0.04 and 0.78 ± 0.05, respectively, indicating that 18 F-FPABZA was more hydrophobic than 18 F-FNABZA. After incubation in either PBS or FBS, the percentage of intact 18 F-FPABZA remained greater than 98% until 4 h (Figure 3), demonstrating that no defluorination or decomposition occurred.

In Vitro Binding of 18 F-FPABZA and 18 F-FNABZA to Melanin
At a concentration of 0.05 mg/mL, the bound ratios of 18 F-FPABZA and 18 F-FNABZA rapidly reached a plateau at the initial 30 min post-incubation ( Figure 4A). The binding affinity of 18 F-FPABZA and 18 F-FNABZA showed a melanin concentration-dependent manner ( Figure 4B). The bound ratios of both radiotracers increased from approximately 20% at melanin concentration of 0.1 mg/mL to higher than 80% at a concentration of 0.2 mg/mL after 1 h incubation. However, the bound ratio of 18 F-FPABZA was higher than that of 18 F-FNABZA at each time point, suggesting that 18 F-FPABZA exhibited a superior affinity to melanin when compared to 18 F-FNABZA.

Partition Coefficient Determination and In Vitro Stability of 18 F-FPABZA or 18 F-FNABZA
The Log P values of 18 F-FPABZA and 18 F-FNABZA were 1.48 ± 0.04 and 0.78 ± 0.05, respectively, indicating that 18 F-FPABZA was more hydrophobic than 18 F-FNABZA. After incubation in either PBS or FBS, the percentage of intact 18 F-FPABZA remained greater than 98% until 4 h (Figure 3), demonstrating that no defluorination or decomposition occurred.

In Vitro Binding of 18 F-FPABZA and 18 F-FNABZA to Melanin
At a concentration of 0.05 mg/mL, the bound ratios of 18 F-FPABZA and 18 F-FNABZA rapidly reached a plateau at the initial 30 min post-incubation ( Figure 4A). The binding affinity of 18 F-FPABZA and 18 F-FNABZA showed a melanin concentration-dependent manner ( Figure 4B). The bound ratios of both radiotracers increased from approximately 20% at melanin concentration of 0.1 mg/mL to higher than 80% at a concentration of 0.2 mg/mL after 1 h incubation. However, the bound ratio of 18 F-FPABZA was higher than that of 18 F-FNABZA at each time point, suggesting that 18 F-FPABZA exhibited a superior affinity to melanin when compared to 18 F-FNABZA.

In Vitro Binding of 18 F-FPABZA and 18 F-FNABZA to Melanin
At a concentration of 0.05 mg/mL, the bound ratios of 18 F-FPABZA and 18 F-FNABZA rapidly reached a plateau at the initial 30 min post-incubation ( Figure 4A). The binding affinity of 18 F-FPABZA and 18 F-FNABZA showed a melanin concentration-dependent manner ( Figure 4B). The bound ratios of both radiotracers increased from approximately 20% at melanin concentration of 0.1 mg/mL to higher than 80% at a concentration of 0.2 mg/mL after 1 h incubation. However, the bound ratio of 18 F-FPABZA was higher than that of 18 F-FNABZA at each time point, suggesting that 18 F-FPABZA exhibited a superior affinity to melanin when compared to 18 F-FNABZA.

In Vitro Cellular Uptake Study of 18 F-FPABZA and 18 F-FNABZA in Melanoma Cells
The melanotic B16F10 murine melanoma cells and amelanotic A375 human melanoma cells were used to investigate the specific binding of 18 F-FPABZA and 18 F-FNABZA to melanin ( Figure 5A,B). The cellular uptake (expressed as %AD/10 6 cells) of 18 F-FPABZA in B16F10 cells was 7.57 ± 0.77 at 15 min after incubation and then decreased to 5.00 ± 0.29 at 4 h post-incubation, while that of 18 F-FNABZA increased from 4.81 ± 0.23 at 15 min postincubation to 7.82 ± 0.52 after 4 h incubation, which was comparable to the peak uptake of 18 F-FPABZA (7.94 ± 0.86 at 60 min post-incubation). However, the radioactivity of both radiotracers retained in A375 cells was significantly lower than that in melanotic B16F10 cells at all time points. The maximum uptake of 18 F-FPABZA and 18 F-FNABZA in A375 cells was 1.43 ± 0.08 and 3.03 ± 0.07, respectively. The highest melanotic-to-amelanotic cells ratio of 18 F-FPABZA and 18 F-FNABZA was 7.94 ± 0.86 at 30 min and 7.82 ± 0.52 at 4 h post-incubation, respectively ( Figure 5C).

MicroPET Imaging
The images of B16F10 melanoma-bearing mice injected with 18 F-FPABZA displayed a noted tumor uptake. The T/M of 18 F-FPABZA-injected B16F10 melanoma-bearing mice increased from 7.6 ± 0.4 at 15 min p.i. to 27.5 ± 16.6 at 3 h p.i. ( Figure 6A). In contrast to B16F10 melanoma, only limited 18 F-FPABZA radioactivity was retained in the A375 tumor at all imaging points ( Figure 6A and Supplementary Materials Figure S1). In both animal models, intense lung accumulation was noticed at the initial time points and then continuously washed out. In fact, the lung uptake was close to the background at although the B16F10 tumor uptake of 18 F-FNABZA slightly increased with time, it was not comparable to that of 18 F-FPABZA. The T/M raised from 2.8 ± 0.7 at 15 min p.i. to 5.1 ± 2.9 at 3 h p.i. ( Figure 6B). Except for melanotic tumor, the eye was also a high melanin-abundant tissue in C57BL6 mouse, so the accumulation of radiotracers in eyes can be considered an index of specific binding affinity toward melanin. Both 18 F-FPABZA and 18 F-FNABZA can clearly delineate the contour of eyes of C57BL6 mice, but the eye uptake of nude mice injected with 18 F-FPABZA was not observed ( Figure S1).
B16F10 melanoma, only limited 18 F-FPABZA radioactivity was retained in the A375 tumor at all imaging points ( Figure 6A and Supplementary Materials Figure S1). In both animal models, intense lung accumulation was noticed at the initial time points and then continuously washed out. In fact, the lung uptake was close to the background at 3 h p.i. in the B16F10 melanoma-bearing mice ( Figure 6A). The T/M of A375 xenograft-bearing mice administered with 18 F-FPABZA was around 1 throughout the experiment period, suggesting that no specific uptake occurred in the amelanotic tumor. Interestingly, although the B16F10 tumor uptake of 18 F-FNABZA slightly increased with time, it was not comparable to that of 18 F-FPABZA. The T/M raised from 2.8 ± 0.7 at 15 min p.i. to 5.1 ± 2.9 at 3 h p.i. ( Figure 6B). Except for melanotic tumor, the eye was also a high melanin-abundant tissue in C57BL6 mouse, so the accumulation of radiotracers in eyes can be considered an index of specific binding affinity toward melanin. Both 18 F-FPABZA and 18 F-FNABZA can clearly delineate the contour of eyes of C57BL6 mice, but the eye uptake of nude mice injected with 18 F-FPABZA was not observed ( Figure S1).

Biodistribution Study
Based on the results obtained from microPET imaging, only the biodistribution studies of 18 F-FPABZA in B16F10 melanoma-bearing mice were performed (Table 1). At 15 min p.i., noted radioactivity was retained in the lung (8.22 ± 1.87 %ID/g), spleen (11.73 ± 1.37 %ID/g), pancreas (12.80 ± 5.60 %ID/g), and kidney (15.57 ± 1.71 %ID/g), which was accompanied with relatively low blood radioactivity (3.7 ± 0.8 %ID/g), suggesting that the majority of 18 F-FPABZA rapidly distributed to the whole body from blood. Most of the radioactivity in the normal organ was continuously washed out, indicating that these organs' accumulation originated from non-specific retention rather than specific uptake. The high radioactivity level in kidney and small intestine implied 18 F-FPABZA was excreted through both the urinary and intestinal routes. The tumor uptake of 18 F-FPABZA was 12.32 ± 4.13 %ID/g at 0.25 h p.i., reached the maximum at 1 h p.i. (20.57 ± 2.22 %ID/g), and remained high until 2 h p.i. (16.89 ± 2.32 %ID/g). As a result of prolonged tumor retention, the tumor-to-muscle and tumor-to-blood ratios increased with time and were 6.97 ± 1.96 and 4.40 ± 1.22, 6.66 ± 3.16 and 5.18 ± 1.66, 26.47 ± 3.11 and 19.89 ± 1.19, and 86.57 ± 55.39 and 51.93 ± 18.05, at 0.25, 0.5, 1, and 2 h p.i., respectively. More importantly, the maximum tumor-to-liver ratio was 23.61 ± 9.04 at 2 h p.i., demonstrating that the conjugation with picolinamide may be a solution to the high liver non-specific uptake of benzamide analog. The black eyeballs of C57BL/6 mice also displayed high radioactivity uptake throughout the experiment period. The bony accumulation of 18 F-FPABZA was 1.84 ± 0.39 %ID/g at 0.25 h p.i. and steadily decreased to 1.40 ± 0.17 %ID/g at 2 h p.i., indicating that no obvious defluorination effect occurred   Results were expressed as the percentage of injected dose per gram of organ/tissue (%ID/g). Each value represented mean ± SD (n = 3). Small int., small intestine; large int., large intestine.

Discussion
Regarding there being no appropriate "specific" agent for the detection of early-stage melanoma until now, several benzamide analogs with high sensitivity and specificity to melanin have been developed, and their biological characteristics have been determined [6,7,9,10,13]. However, the high lipophilicity of these probes generally caused relatively low bioavailability and intense liver accumulation. Our previous works demonstrated that benzamide conjugation with picolinamide or nicotinamide would greatly enhance its hydrophilicity without significantly affecting melanin-targeting ability [13,14]. To improve the imaging resolution of 131 I-labeled radiotracers, we successfully prepared radiofluorinated fluoropicolinamide-benzamide ( 18 F-FPABZA) and fluoronicotinamidebenzamide derivatives ( 18 F-FNABZA) for PET imaging in the present study. The radiochemical yield of 18 F-FPABZA was higher than that of 18 F-5-FPN [11]. However, the labeling efficiency of 18 F-FNABZA was low despite using the identical radiofluorination method (≈12%, Figure 2B). This phenomenon may be explained by the different electron densities of pyridine rings between two tracers. Theoretically, the 3-position on the pyridine ring is more electron-deficient than the 2-position, causing a driving force for nucleophilic substitution reaction.
The melanin-binding assays indicated that the melanin-specific binding of 18 F-FPABZA and 18 F-FNABZA was higher than that of 125 I-IBZA (~80%) but lower than 131 I-IFPABZA and 131 I-IFNABZA at the corresponding concentration (0.2 mg/mL) (Figure 3) [6,14]. The cellular uptakes of 18 F-FPABZA and 18 F-FNABZA in melanotic B16F10 cells were 5.0-and 3.0-fold higher than in amelanotic A375 cells at 1 h after incubation, respectively, suggesting that melanin was the main target of these radiotracers (Figure 4). Although the accumulation of 18 F-FPABZA reached the maximum (7.94 ± 0.86 %AD/10 6 cells) at 0.5 h post-incubation, the value was significantly lower than the peak uptakes of 131 I-IFPABZA (67.8 ± 0.2) and 131 I-IFNABZA (62.6 ± 0.2) [14]. The possible explanation may be the difference in lipophilicity between these compounds. 131 I-IFPABZA and 131 I-IFNABZA were more hydrophobic than 18 F-FPABZA and 18 F-FNABZA, causing radioiodinated compounds more easily to enter into the cytoplasm via diffusion. Pham et al. indicated that radiolabeled probes, having a log P value > 1.4, owned superior accumulation in B16F10 tumor [15]. In our previous study, we also reported a positive correlation between the lipophilicity of radiotracers and their tumor uptakes [14]. The amounts of used precursor between radioiodination (≈30 µg) and radiofluorination (≈7 mg) may be another reason to cause different cellular uptakes between these radiotracers. The residual precursor would also bind to melanin and result in a competitive effect.
After the injection of 18 F-FPABZA into B16F10 melanoma-bearing mice, the prolonged retention of radioactivity in tumor and fast washout in normal tissue resulted in a high tumor-to-background ratio (TBR) and imaging contrast ( Figure 6A). To further assess the in vivo specific binding affinity of 18 F-FPABZA to melanin, the imaging of amelanotic A375 xenograft-bearing BALB/c nude mice was performed. As expected, the radioactivity of 18 F-FPABZA retained in amelanotic A375 tumors as well as eyes were not significant, suggesting that the main reason for 18 F-FPABZA long-term tumor accumulation was melanin binding ( Figure 6A). On the contrary, 18 F-FNABZA did not display an apparent tumor accumulation and TBR, unlike the results obtained from in vitro experiments ( Figure 6B). Regarding that eyes were still visible at 3 h p.i., 18 F-FNABZA should own moderate in vivo binding affinity to melanin, suggesting that the melanin affinity is not the primary factor to its limited tumor retention.
As a result of the great binding affinity of 18 F-FPABZA in imaging experiments, we only assessed the biodistribution study of B16F10 melanoma-bearing mice injected with 18 F-FPABZA in this study. The tumor uptake continuously elevated from 12.32 ± 4.13 %ID/g at 15 min p.i. to 20.57 ± 2.2 %ID/g at 1 h p.i. The tumor-to-liver ratio of 18 F-FPABZA was 23.61 ± 9.04 at 2 h p.i. due to insignificant liver accumulation after excretion. Considering that the late-stage melanoma often spreads to liver, these results indicated that 18 F-FPABZA would be an appropriate probe to detect the metastatic melanoma in liver when compared to other published benzamide analogs, such as 123 I-IBZA, 18 F-DAFBA, and 131 I-ICNA [6][7][8]13].
Uveal melanoma (UM), originating form melanocytes of the uveal, which is the middle layer of the eye and comprises the choroid, ciliary body, and iris, represents around 5% of all melanoma [16]. In fact, except for skin, the eye is the most likely site of melanoma throughout the body. UM can sometimes remain clinical silent for many years. Unfortunately, half of patients with UM develop distant metastases, especially in the liver, leading to poor prognosis [17]. Regarding the difficulties of biopsies to ocular melanoma, eye ultrasound, imaging of the blood vessels, and optical coherence tomography (OCT) have been applied to detect ocular melanoma. However, it is still difficult to differentiate the malignant melanoma from the benign one by using these modalities. In 1998, Bacin et al. found that a benzamide analogue, 123 I-BZA, was suitable for the diagnosis of ocular melanoma in a phase II clinical trial recruiting 48 patients with a suspicion of ocular melanoma [18]. As mentioned before, 18 F-FPABZA has a similar lipophilicity and relatively low liver uptake when compared with 123 I-BZA, suggesting it could be a potential probe to detect UM and the metastases in the liver. In addition, the genetic background is somewhat different from that of cutaneous melanoma despite a similar morphology between these two types of melanoma being observed [17,19,20]. For example, the expression of P16INK4a, which is related to the pigmentation, is a distinct feature in UM and can be a targeting mechanism for 18 F-FPABZA [17]. Further studies are warranted to determine if the PET imaging with appropriate probes can discriminate between the different types of melanoma.

Partition Coefficient Determination of 18 F-FPABZA or 18 F-FNABZA
The partition coefficients of 18 F-FPABZA and 18 F-FNABZA were determined by measuring their distribution between 1-octanol and phosphate buffer solution (PBS). Briefly, 2 µCi of either 18 F-FPABZA or 18 F-FNABZA was added to a mixture of 1-octanol (1 mL) and PBS (pH 7.4, 1 mL). After being vigorously shaken by vortex for 5 min and centrifuged at 3000 rpm for 5 min, aliquots of 1-octanol (0.5 mL) were taken out and then added to another tube containing 0.5 mL of 1-octanol and 1 mL of PBS. This process was repeated five times. The counting samples (100 µL) of each layer were aspirated, and their radioac-tivity was assessed by a gamma counter. The partition coefficient was expressed as Log P based on the following Equation (1).

In Vitro Stability of 18 F-FPABZA
To determine the in vitro stability, 100 µCi of 18 F-FPABZA was incubated in either 1 mL of PBS (at 4 • C and 37 • C) or 1 mL of fetal bovine serum (FBS) at 37 • C. At designated time points (0.25, 0.5, 1, 2, and 4 h), an aliquot of sample was aspirated for radio thin-layer chromatography (radio-TLC) measurement. The percentage of intact 18 F-FPABZA was considered as an index of its in vitro stability.

Binding Affinity to Melanin Assays
To investigate the binding affinity to melanin, 20 µCi of either 18 F-FPABZA or 18 F-FNABZA was added to each tube containing various concentrations of melanin (0.01 to 0.2 mg/mL). The mixture was incubated at 37 • C for 1 h. After incubation, the mixture was centrifuged at 20,000× g for 5 min. The supernatant was collected and then filtered by a 0.22 µm membrane filter to remove residual bound radiotracer. The radioactivity of the pellet and the filtrate was measured by a γ-counter. To evaluate the effect of incubation time on the melanin-binding affinity, 20 µCi of either 18 F-FPABZA or 18 F-FNABZA was incubated at 37 • C for different time durations (0.5, 1, 1.5, and 2 h). The binding affinity was calculated as follows (2). % of radiotracer bound = 1 − radioactivity of filtrate total administered radioactivity (2)

Cellular Uptake Studies
Melanotic B16F10 and amelanotic A375 melanoma cells were obtained from the Bioresource Collection and Research Center (Taiwan). Both cells were cultured in Dulbecco's modified Eagle high-glucose medium (Gibco Life Sciences) supplemented with 10% FBS at 37 • C in a humidified atmosphere of 5% CO 2 . Cells were seeded in a 6-well plate at a density of 1 × 10 6 cells/well and cultured in the incubator. After cells reaching 70% confluence, the culture medium was replaced by 3 mL of fresh medium containing 18 F-FPABZA or 18 F-FNABZA (1 µCi/mL). At 0.25, 0.5, 1, 2, 4, and 8 h post-incubation, the medium was aspirated, and the adhered cells were washed twice with 0.5 mL of PBS. The medium and washing PBS were collected and added to a counting tube. Trypsin (0.25%, 0.5 mL) was added to each well, and the plate was incubated at 37 • C for 5 min to detach cells. The cell suspension was taken out and added to another counting tube. The plate was rinsed with PBS (0.25 mL) twice. The washing PBS was also added to the counting tube containing cell suspension. The radioactivity of cells and medium was measured with a γ-counter and normalized to the number of viable cells. The cellular uptake of 18 F-FPABZA or 18 F-FNABZA was expressed as the percentage of administrated dose per one million cells (%AD/10 6 cells).

Establishment of Melanoma-Bearing Mouse Models
All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of National Yang-Ming University (IACUC no. 1060604). For the subcutaneous melanoma-bearing mouse models, B16F10 (5 × 10 5 ) and A375 (1 × 10 6 ) cells were inoculated in the right flank of male C57BL/6 mice and male BALB/c nude mice, respectively. Biological studies were conducted when the tumor burdens reached 100 ± 50 mm 3 .

MicroPET Scanning
MicroPET images were obtained by the Lab-PET system (FLEX Triumph) at Taipei Veterans General Hospital, Taipei, Taiwan. The tumor-bearing mouse was anesthetized with 2% isoflurane (in oxygen) using a vaporizer system in the prone position during the imaging process. The B16F10 melanoma-bearing mice were injected with approximately 300 µCi of 18 F-FPABZA or 18 F-FNABZA through the lateral tail vein, while the A375 melanoma-bearing mice were only injected with 18 F-FPABZA for 2 h dynamic imaging. For quantitative analysis, a region of interest (ROI) was placed on tumor and contralateral side of muscle. The tumor-to-muscle ratio (T/M) was calculated on the basis of counts per pixel in the ROIs.

MicroPET Scanning
B16F10 tumor-bearing mice were injected with 18 F-FPABZA (100 µCi/mice) through the tail vein. At 0.25, 0.5, 1, and 2 h post-injection (p.i.), the tissues/organs of interest were harvested, weighed, and subjected to radioactivity measurement using a γ-counter. The results were expressed as percentage of injected dose per gram of tissue/organ (%ID/g).

Conclusions
To our best knowledge, this study is the first to synthesize radiofluorinated picolinamidebenzamide and nicotinamide-benzamide derivatives and evaluate their biological characteristics. The results from in vivo experiments revealed that 18 F-FPABZA exhibited better melanoma-targeting ability than 18 F-FNABZA and relatively low liver accumulation compared to other existing benzamide analogs, suggesting that it could be a potential clinical probe for the detection of melanotic melanoma.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to ethical issues.