Leveraging Open Science Drug Development for PET: Preliminary Neuroimaging of 11C-Labeled ALK2 Inhibitors.

Mutations in the gene encoding activin receptor-like kinase 2 (ALK2) are implicated in the pathophysiology of a pediatric brainstem cancer, diffuse intrinsic pontine glioma (DIPG). Inhibitors of ALK2 that cross the blood-brain barrier have been proposed as a method of treatment for DIPG. As part of an open science approach to radiopharmaceutical and drug discovery, we developed 11C-labeled radiotracers from potent and selective lead ALK2 inhibitors to investigate their brain permeability through positron emission tomography (PET) neuroimaging. Four radiotracers were synthesized by 11C-methylation and assessed by dynamic PET imaging in healthy Sprague-Dawley rats. One of the compounds, [ 11 C]M4K2127, showed high initial brain uptake (SUV ∼ 2), including in the region of interest (pons). This data supports the use of this chemotype as a brain penetrant ALK2 inhibitor that permeates evenly into the pons with potential application for the treatment of DIPG.

Activin receptor-like kinase 2 (ALK2) is a kinase receptor encoded by the ACVR1 gene. Mutations in ACVR1 causing a gain-of-function in ALK2 signaling have been reported in approximately 25–30% of children with diffuse intrinsic pontine glioma (DIPG), a rare pediatric brainstem cancer with a poor prognosis.[1−3] DIPG tumors are aggressive and arise from the brain’s glial tissue in the lowest stem-like portion of the brain, the pons. Although the pathology of the relationship is not yet fully known, ALK2 has emerged as a potential therapeutic target for the treatment of DIPG. Kinase inhibitors have notoriously poor selectivity, often showing off-target affinity to other kinases within a family because of the high degree of structural homology in ATP binding domains. In the case of ALK2, off-target binding to the closely related TGFβ pathway receptor ALK5 is of specific concern due to potential cardiotoxicity.[4] In addition, the discovery of kinase inhibitors that are sufficiently able to cross the blood–brain barrier (BBB) has been lacking[5] and there is no report of work aiming to measure the degree of penetration of DIPG-directed therapeutics into the pons.[6−8] We recently reported on leveraging an open science drug discovery model[9] to synthesize a series of 3,5-diphenylpyridine ALK2-selective inhibitors that cross the BBB for the treatment of DIPG.[10,11] The lead compounds from this effort show brain penetration in rodent models with total brain to plasma (B/P) ratios of 0.8–1.4, although distribution into the therapeutic region of interest in the brain (pons) was unknown. Central nervous system (CNS) drug discovery can be facilitated with position emission tomography (PET) imaging by measuring the BBB permeability of isotopically labeled drug candidates. In addition, PET imaging can be used to visualize distribution of the radiotracer into specific regions of the brain such as the pons. In addition to our recent work on ALK2 drug discovery, the goal of the present study was to apply an open science approach for PET radiotracer discovery, to label four lead potent and ALK2-selective inhibitors with carbon-11 (half-life = 20.4 min), and to assess the BBB permeability of these radiotracers by PET imaging in rodent models. Assessment of the brain permeability of these compounds as PET radiotracers identified a candidate radioligand that can be used for further drug development of ALK2 inhibitors of this chemotype, facilitating receptor occupancy, target engagement, and dosing regimen studies.

Four lead compounds with IC50 values for ALK2 between 9–19 nM were selected for radiolabeling: M4K2163, M4K2009, M4K2117, and M4K2127.[11] The structural position of the 11C radionuclide was chosen based on availability of the desmethyl precursors and ease of synthesis (Figure ). While there is no consensus for the ideal physiochemical properties of small molecules that are BBB penetrant,[12] CNS Multiparameter Optimization (MPO) scores offer a guideline for compound prioritization.[13] Prior to radiolabeling, CNS MPO desirability scores were calculated for all of the lead compounds, giving scores of 4.34–5.04 (maximum score is 6.0), suggesting that all compounds are rational leads as brain-penetrant radioligands. Calculated logD7.4 values were also within the range for most brain-penetrant PET radioligands[14,15] (2.03–3.31), with limited off-target binding. As such, none of the initially identified compounds were excluded due to physiochemical characteristics that would suggest low brain permeability (Table ).

Figure 1

Structures of our lead 11C-radiotracers for ALK2.

Table 1

Physiochemical Properties of the Lead Compounds

	ALK2 IC50 (nM)a	B/Pb	pKac	clogPc	clogD7.4c	logD7.4d	CNS MPOe
M4K2009	13	0.9*	8.5	3.20	2.03	2.76 ± 0.13	5.04
M4K2117	7	0.82	9.5	4.56	2.46	2.64 ± 0.05	4.34
M4K2127	9		7.6	3.74	3.31	3.37 ± 0.12	4.34
M4K2163	19	1.41	8.5	4.05	2.88	2.45 ± 0.04	4.34

Cellular IC50.[11]

Brain/plasma ratio in female SCID mice (4 h) or *male SD-1 mice (2 h) @ 10 mg/kg PO.[11]

Properties retrieved from ACD/I-Lab 2.0.

Determined experimentally by shake-flask method with 11C-labeled compounds.

Scored via an algorithm using a weighted scoring function for clogP, clogD, molecular weight, topological polar surface area, HBDs, and pKa.

All radiosyntheses were performed on a commercial automated carbon-11 synthesis unit for methylation reactions (GE FX2 C). [C]M4K2163 and [C]M4K2009 were synthesized by 11C-O-methylation with the “loop method”[16,17] using [11C]CH3I and the respective phenolic precursors (M4K2297 and M4K2298; see the Supporting Information for their preparation) with <1 equiv of potassium tert-butoxide (Scheme A). The radioligands were synthesized, isolated, and formulated in radiochemical yields of 6–10% ([C]M4K2163, n = 3) and 4–9% ([C]M4K2009, n = 2) relative to [11C]CO2. Both radiotracers were obtained in >95% radiochemical purities and high molar activities ([C]M4K2163: 270–350 GBq/μmol), [C]M4K2009: 304 GBq/μmol). Small-scale vial-based radiosyntheses of [C]M4K2117 and [C]M4K2127 using [11C]CH3I and various bases (KOtBu, NaOH, or TBAOH) resulted in low radiochemical conversion (<5%) by analytical HPLC and were not further pursued. [C]M4K2117 and [C]M4K2127 were each synthesized by 11C-N-methylation and also with use of the “loop method”, although synthesis of these radioligands employed the more reactive [11C]CH3OTf and the respective N-desmethyl precursors (M4K2009 and M4K2043, Scheme B). For the efficient radiolabeling of [C]M4K2117, 1 equiv of aqueous NaOH was added to the reaction mixture. These radioligands were synthesized, isolated, and formulated in radiochemical yields of 5–6% ([C]M4K2117, n = 3) and 11–26% ([C]M4K2127, n = 5) relative to [11C]CO2. Both radiotracers were obtained in >99% radiochemical purities and high molar activities ([C]M4K2117: 269 GBq/μmol), [C]M4K2127: 132–664 GBq/μmol). All radiotracers were prepared within 30–40 min from end of bombardment.

Scheme 1

Radiosynthesis of 11C-Labeled Compounds

(A) O-methylations via [11C]CH3I, (B) N-methylations via [11C]CH3OTf.

Shake-flask logD7.4 values were experimentally determined using established procedures.[18] Aliquots of the formulated radiotracers were partitioned between octanol/phosphate buffer, and the radioactivity in each layer was measured on a γ counter. Experimental logD7.4 values were slightly higher than the calculated values and were in the range of 2.45–3.37 (Table ), consistent with most CNS PET tracers.[14,15]

The brain permeability of the radiotracers was then assessed in healthy Sprague–Dawley rats (Figure ). Rats were injected (i.v.; tail-vein) with 16–23 MBq of each radiotracer, and dynamic PET imaging was acquired over 90 min. Time-activity curves (TACs) were extracted for whole brain and for the pons as a region of interest using the atlas of Schwarz et al. 2006[19] implemented in VivoQuant (4.0patch1, Invicro). Three tracers, [C]M4K2163 (n = 2), [C]M4K2117 (n = 1), and [C]M4K2009 (n = 1) showed relatively low brain uptake, with peak SUV values of 1.1, 0.9, and 0.6, respectively, within 1 min after injection. Fortunately, [C]M4K2127(n = 3) showed good permeability into the brain, with a peak SUV of 1.8 ± 0.2 for initial brain uptake within 1 min after injection, followed by a moderate rate of wash out over 30 min. Importantly, we found that [C]M4K2127 has high uptake in the pons (peak SUV = 2.0 ± 0.3). Looking at the similar TACs for [C]M4K2127 in the pons and the whole brain, it is evident the compound permeates homogeneously into the pons as compared to the rest of the brain. This, importantly, constitutes the first example of an ALK2 PET radiotracer and provides evidence that compounds of this chemotype can access the pons, which is the specific region of interest for the treatment of DIPG tumors.

Figure 2

(A) Summed (0–90 min, iterative reconstruction) PET images in transverse plane for the four 11C-radiotracers in male SD rats. (B) Whole brain TACs for the radiotracers (0–90 min). (C) TACs for the pons for [C]M4K2127 (0–90 min, n = 3).

To investigate the properties that could be affecting the BBB permeability of these four compounds, we reviewed their Caco-2 permeability parameters as an indication of membrane permeability as well as to consider the compounds as potential substrates of efflux transporters. As anticipated from the PET imaging results, M4K2127 had the highest Caco-2 permeability, three times greater than that of M4K2009 (Papp AB = 16.6 vs 5.4 × 10–6 cm/s where Papp > 3 indicates high permeability). Additionally, the efflux ratio for M4K2127 was found to be almost 2–3-fold better than that of M4K2009 and M4K2163 (efflux ratio BA/AB = 0.9, 1.5, and 2.7, respectively; see the Supporting Information, Table S1). This could be attributed to the lower pKa of the N-methyl piperazine moiety in M4K2127 (Table ) along with the molecule having a reduced number of hydrogen bond donors (HBDs), a parameter often associated with a decrease in permeability glycoprotein (P-gp)-mediated efflux.[12] Among the four lead compounds, M4K2127 showed the lowest microsomal stability (see the Supporting Information, Table S1), which is a potential hurdle for further development as a drug candidate. However, when administered via i.v. injection as in our rodent PET imaging study, M4K2127 initially avoids major first-pass metabolism by the liver which increases the initial Cmax. These in vitro ADME results align well with the PET imaging data demonstrating that [C]M4K2127, having the best permeability in vitro, efficiently crosses the BBB.

Herein, we present [C]M4K2127 as an ALK2-targeting radioligand that satisfactorily and homogeneously permeates the brain. By employing an open science approach to this work, we were able to rapidly radiolabel isotopologs of our lead ALK2 inhibitors and apply PET imaging for the assessment of brain penetration. The present study sheds light on our previous work to establish B/P ratios in this new class of ALK2 inhibitors,[11] with implications both for the drug development process and direct analysis of DIPG-targeting drug candidates in a clinical setting. To our knowledge, this report constitutes the first disclosure of ALK2 PET radiotracers and efforts aimed at assessing the degree of penetration into the brain (including the pons) of a compound being pursued as a potential treatment for DIPG.[20]

While numerous compounds have entered the clinic for DIPG, no studies on their degree of penetration into the pons (using PET imaging, or any other method) have been reported. [C]M4K2127 will be used as a biomarker to further assist the drug development of 3,5-diphenylpyridine ALK2 inhibitors. Blocking studies with [C]M4K2127 to confirm specific binding in the brain of rodents and higher species, evaluation of radiometabolism, and effect of anesthesia are planned. [C]M4K2127 may also be used as a tool in receptor occupancy, target engagement, and dosing regimens for other ALK2 inhibitors as they progress toward the clinic.