Development of Novel Phosphonodifluoromethyl-Containing Phosphotyrosine Mimetics and a First-In-Class, Potent, Selective, and Bioavailable Inhibitor of Human CDC14 Phosphatases

Abstract

Together with protein tyrosine kinases, protein tyrosine phosphatases (PTPs) control protein tyrosine phosphorylation and regulate numerous cellular functions. Dysregulated PTP activity is associated with the onset of multiple human diseases. Nevertheless, understanding of the physiological function and disease biology of most PTPs remains limited, largely due to the lack of PTP-specific chemical probes. In this study, starting from a well-known nonhydrolyzable phosphotyrosine (pTyr) mimetic, phosphonodifluoromethyl phenylalanine (F2Pmp), we synthesized 7 novel phosphonodifluoromethyl-containing bicyclic/tricyclic aryl derivatives with improved cell permeability and potency toward various PTPs. Furthermore, with fragment- and structure-based design strategies, we advanced compound 9 to compound 15, a first-in-class, potent, selective, and bioavailable inhibitor of human CDC14A and B phosphatases. This study demonstrates the applicability of the fragment-based design strategy in creating potent, selective, and bioavailable PTP inhibitors and provides a valuable probe for interrogating the biological roles of hCDC14 phosphatases and assessing their potential for therapeutic interventions.

Introduction

Protein tyrosine phosphorylation is a critical component of cellular signaling that is tightly controlled by the opposing actions of protein tyrosine kinases (PTKs) and protein tyrosine phosphatases (PTPs) and is essential for the regulation of a variety of cellular functions¹. The roles of several PTPs, such as PTP1B^2–4, TC-PTP^3–7, SHP2^8,9, and LYP^10–12, have been clearly established in the etiology of several human diseases, highlighting the therapeutic potential of these PTPs. However, there are over 100 PTPs in the human genome¹³ and the roles of the vast majority of PTPs in the development of human diseases are still unexplored or incompletely understood. Although small molecule inhibitors are valuable chemical tools for interrogating the role of specific PTPs in the pathogenesis and validating their therapeutic potential under a more clinically relevant setting, the development of potent, selective, and bioavailable PTP inhibitors is a difficult endeavor. Traditionally, targeting the phosphotyrosine (pTyr) binding pocket (i.e. the PTP active site) with nonhydrolyzable pTyr mimetics is the most straight forward strategy to realize PTP inhibition. However, due to the positively charged and highly conserved nature of PTP active site^14–16, most nonhydrolyzable pTyr mimetics bear at least one negative charge and can indiscriminately bind to multiple PTPs, which hinders their development into potent, selective, and bioavailable inhibitors for the PTP of interest. Although allosteric regulatory mechanisms can be harnessed to bypass the aforementioned challenges for some PTPs^17–20, the uniqueness of each allosteric mechanism also limits broad application of allosteric inhibition to all PTPs. Moreover, many PTPs do not possess known allosteric pockets, and full-length structures are not available to facilitate detection of new allosteric sites. Therefore, a more general method is still warranted for the development of potent, selective and bioavailable inhibitors for in-depth interrogation of PTP biology and therapeutic translation.

Among the “untapped” PTPs, the cell division cycle 14 (CDC14) phosphatases are highly conserved members of the dual-specificity phosphatases (DUSPs) within the PTP superfamily that exist in most eukaryotes. Although DUSPs generally dephosphorylate both phosphotyrosine and phosphoserine (pSer)/phosphothreonine (pThr) with the same active site and employ a catalytic mechanism identical to that of pTyr-specific PTPs²¹, the CDC14 phosphatases evolved a unique substrate specificity. An ancient duplication of the DUSP fold, a signature feature of CDC14 enzymes²² created a novel peptide binding groove at the domain interface adjacent to the catalytic pocket^23,24. This interface renders the catalytic pocket highly selective for protein substrates containing pSer followed by the docking motif Pro-X-Lys/Arg²⁵. For example, the catalytic efficiency of CDC14 enzymes is typically >1,000-fold higher towards pSer-containing peptide substrates compared to identical pThr-containing sequences and >100-fold higher than identical pTyr-containing sequences^22,25. Consistent with this, most characterized biological targets of CDC14 are substrates of cyclin-dependent kinases, which share recognition of the Ser-Pro-X-Lys/Arg motif. We recently suggested that the unique active site selectivity and nearby binding pockets for the +1 Pro and +3 Lys/Arg may allow highly selective inhibitor development towards the CDC14 phosphatases²².

The human genome encodes two widely expressed CDC14 homologs (hCDC14A and B)²⁶ and a hominid-specific hCDC14C that is nearly identical to hCDC14B and expressed only in testis and brain²⁷. The biological function of hCDC14A/B and their connections with the etiology of diseases are an active area of investigation. While early studies with gene overexpression or RNAi knockdown suggested that both enzymes regulate essential processes in mitosis such as spindle formation^28–30, centrosome duplication³¹ and separation²⁹, chromosome segregation²³, nuclear organization³², mitotic exit³³, and cytokinesis²³, more recent work with double knockout mice and human cell lines indicate that hCDC14A and B are completely dispensable for cell division³⁴. Instead, hCDC14A and B are essential for differentiation of neuronal stem cells during mouse embryogenesis³⁵. In addition, hCDC14A, in particular, appears to play important roles in ciliogenesis and cilia maintenance, and human families with hCDC14A mutations suffer from hearing loss and male infertility associated with loss of ciliated cells in the inner ear and reproductive tract³⁶. The connection between hCDC14A/B and cancer is complicated. Existing studies suggest that both hCDC14A and hCDC14B are potential oncoproteins, for example promoting tumorigenesis via dephosphorylation of p53 at Ser315, which disrupts p53-Pin1 and p53-E2F1 interactions, reducing p53 stability and transactivation activity.^37–42 On the other hand, hCDC14A was reported to restrain cell migration associated with malignant tumor growth⁴³ and reduced hCDC14A and B expression is correlated with poor prognosis in certain cancers^44,45. Additionally, genetic knockout of hCDC14A or B also impaired the DNA damage response in certain tumor cells and reduced their viability after ionizing radiation^46–48, which indicated that both enzymes can be utilized as therapeutic targets that sensitize tumors to radiotherapy and overcome the radiation resistance^49,50. Potent and selective hCDC14A/B inhibitors would be invaluable chemical probes to help clarify contributions of hCDC14A/B to cancer and other physiological processes and test various therapeutic hypotheses.

To offer a general approach that can overcome the challenges associated with orthosteric PTP inhibitor development, our group previously advanced a fragment-based PTP inhibitor design strategy^1,51. It is well known that pTyr plays an essential role in PTP-substrate recognition and the amino acids flanking the pTyr residue increase the binding affinity and selectivity of pTyr-containing peptides^52–56. Accordingly, we hypothesized that molecules engaging both the conserved PTP active site and nearby distinct peripheral pockets, which can be generated by tethering a nonhydrolyzable pTyr mimetic with appropriate diversity fragments, will possess enhanced potency and selectivity due to the additivity of free energy of binding and the unique interaction involved with the less conserved peripheral sites^1,51. Additionally, owing to the uncharged nature of most peripheral pockets, many of the corresponding binding fragments are neutral and better at balancing inhibitor polarity and lipophilicity. With this strategy, a number of potent, selective, and bioavailable orthosteric PTP inhibitors have been developed^10,57–63 and recently, a PTP1B/TC-PTP dual targeting active site-directed competitive inhibitor ABBV-CLS-484 has successfully proceeded to clinical trials⁶⁴. Although peripheral site ligands undoubtedly play a crucial role in augmenting the potency, selectivity, and cell permeability of PTP inhibitors, the pTyr mimetic remains the cornerstone in the fragment-based design of competitive PTP inhibitors. Therefore, a pTyr mimetic possessing superior PTP binding affinity and cell permeability can greatly facilitate the development of potent, selective, and drug-like inhibitors for a variety of therapeutically relevant PTPs.

Among the reported nonhydrolyzable pTyr mimetics, phosphonodifluoromethyl phenylalanine (F2Pmp)⁶⁵ is widely utilized in PTP inhibitor development¹. By virtue of its structural resemblance to pTyr, F2Pmp retains the ability to bind the PTP active sites⁶⁶. The amino acid-like structure of F2Pmp also facilitates its further structural expansion and decoration. However, this structural characteristic also imposes restrictions on feasible approaches for its subsequent elaboration (i.e. limited to peptide incorporation). Moreover, the difluorophosphonic acid within F2Pmp is di-ionized at physiological pH, which generally causes poor cell membrane penetration and bioavailability. Lastly, the structural simplicity of the difluorobenzyl moiety of F2Pmp also limits its interactions with the PTP active site, leading to relatively low PTP binding affinity⁵⁶. Due to these limitations, most of the existing F2Pmp-based potent, selective, and cell-active PTP inhibitors are peptide-like molecules with high molecular weight.^{59,60,66–68}. As a result, the ligand efficiency and metabolic stability of these inhibitors were usually compromised. To create high affinity and cell permeable pTyr mimetics for further PTP inhibitor development, we designed and synthesized 7 novel phosphonodifluoromethyl-containing bicyclic/tricyclic aryl derivatives that were more cell permeable and exhibited significantly improved binding affinity toward one or multiple screened PTPs. Using fragment- and structure-based design, we developed the first potent and selective hCDC14A/B inhibitor 15, which interacts with both the active site and the nearby P+1 Pro binding pocket. In addition, compound 15 blocked hCDC14A/B mediated p53 phosphorylation and stabilized p53 protein in U2OS cells. Importantly, compound 15 exhibited good pharmacokinetic properties and was also orally bioavailable. Taken together, our study demonstrated that potent, selective, and bioavailable active site-directed PTP inhibitors can be obtained using a fragment-based design approach. The yielded hCDC14A/B inhibitor can serve as a chemical probe for further interrogation of the biological function and assessment of the therapeutic potential of these two enzymes. Furthermore, the reported novel pTyr mimetics can also be applied to the design of inhibitors for other therapeutically relevant PTPs.

Results and Discussion

Design, Synthesis, and Characterization of Novel Phosphonodifluoromethyl-Containing pTyr Mimetics

As depicted in Figure 1A, our work started by simplifying the F2Pmp or Ac-F2Pmp-OMe (compound 1, which better mimics the F2Pmp moiety in most reported PTP inhibitors^{59,60,67–69}) structure to P-(Difluorophenylmethyl) phosphonic acid (PhFP, compound 2), the segment crucial for the interaction between PTP active site and F2Pmp. This simplification was driven by the recognition that the amino acid-like structure present in F2Pmp constrains further elaboration and offers only a limited contribution to binding affinity. To further enhance PTP binding while reducing the polarity of PhFP, we retained the essential difluorophosphonic acid moiety within PhFP while replacing the benzene ring with bicyclic or tricyclic aromatic rings. This strategy is supported by earlier studies indicating that PTPs have a greater affinity for bicyclic/tricyclic aryl phosphate substrates and nonhydrolyzable pTyr mimetics compared to their single-ring counterparts^70–72. The basis for this approach is also grounded in the idea that the more lipophilic fused aromatic ring system can potentially better offset the pronounced polarity of the difluorophosphonic acid moiety and thereby enhance cell permeability.

Figure 1.

Design and synthesis of the novel phosphonodifluoromethyl-containing pTyr mimetics. (A) Design strategy for the novel pTyr Mimetics. (B) Structure of the 7 novel pTyr mimetics.

As shown in Figure 1B, 7 novel pTyr mimetics were synthesized with the methods described in the Chemistry section. Upon obtaining the 7 new pTyr mimetics, we set out to determine their PTP inhibitory activity. Since self-aggregation of small organic molecules can lead to the formation of colloidal aggregates and cause 85–95% of the false positives in enzymatic assays⁷³, we began by screening all of the pTyr mimetics with a previously described dynamic light scattering (DLS)-based small molecule aggregation detection protocol⁷⁴ to exclude small molecule aggregators, if any. All new pTyr mimetics were dissolved in 3,3-Dimethylglutaric acid (DMG) buffer (pH 7.0) (same as our enzymatic inhibition assay) at 1 mM concentration. As shown in Figure S1, no DLS signal decay was observed in any solution, indicating that no particle with >1.0 nm diameter was detected in these solutions, thus none of the compounds can form self-aggregates. In contrast, a known aggregator (Ketoconazole) gave strong DLS signal decay at concentrations as low as 15 μM, agreeing with previous reports using various detection methods⁷⁵. We next profiled the inhibitory activity of PhFP and all 7 new pTyr mimetics against a panel of 10 major mammalian PTPs with 6,8-Difluoro-4-Methylumbelliferyl Phosphate (DiFMUP) as substrate. As shown in Table 1, PhFP exhibited low millimolar half-inhibitory concentration values (IC₅₀s) toward most tested PTPs (except CD45, for which PhFP showed no inhibition at 10 mM concentration). Moreover, no significant inhibition selectivity was observed with this compound. In contrast, all of the new bicyclic or tricyclic pTyr mimetics (compounds 3-9) exhibited low- to mid- micromolar IC₅₀s for one or more PTPs with initial selectivity over other tested PTPs. These findings agreed with previous studies showing PTPs prefer bicyclic/tricyclic aryl phosphate substrates and pTyr mimetics compared to their single-ring counterparts^70–72. Interestingly, several of these novel pTyr mimetics demonstrated distinct patterns of inhibition selectivity across specific PTPs. For instance, compounds 3, 5, and 6 displayed modest inhibition selectivity towards SHP1, while compound 4 exhibited a predilection for binding with CD45. Additionally, compound 7 showed a preference for inhibiting both PTP1B and Laforin, and compound 9 demonstrated remarkable selectivity for hCDC14A.

Table 1.

IC₅₀s of PhFP (compound 2) and 7 new pTyr mimetics on 10 major mammalian PTPs.^a

Cmpd	IC50,μM
	CD45	Laforin	MKP5	SHP2	SHP1	hCDC14A	TC-PTP	PTP1B	LYP	STEP
2	>>10,000	8384 ± 759	8226 ± 626	7985 ± 672	5730 ± 477	3269 ± 303	4853 ± 196	2217 ± 126	3205 ± 169	2860 ± 119
3	234.8 ± 16.0	>500	>500	165.9 ± 12.4	57.0 ± 6.2	220.6 ± 26.3	>500	>500	>500	>500
4	21.2 ± 1.4	61.3 ± 24.1	267.9 ± 32.9	167.2 ± 2.3	64.7 ± 5.2	98.7 ± 2.9	402.3 ± 17.4	>500	41.9 ± 4.9	206.9 ± 20.6
5	42.9 ± 3.7	122.4 ± 9.0	214.7 ± 12.0	71.8 ± 6.5	22.7 ± 2.4	119.6 ± 14.6	397.2 ± 27.4	447.2 ± 23.9	37.5 ± 2.0	334.7 ± 28.5
6	107.1 ± 8.1	197.8 ± 18.4	431.3 ± 47.2	75.6 ± 16.4	36.6 ± 2.3	273.2 ± 18.8	>500	>500	139.4 ± 24.8	>500
7	339.0 ± 25.8	59.7 ± 3.2	332.2 ± 28.6	181.0 ± 13.2	138.7 ± 5.7	339 ± 19.9	89.1 ± 6.7	40.2 ± 1.4	98.4 ± 6.0	287.2 ± 21.0
8	477.65 ± 41.9	116.85 ± 13.7	101.7 ± 8.9	164.1 ± 17.9	167.3 ± 15.6	360 ± 28.8	427.45 ± 26.4	189.25 ± 12.7	173.7 ± 14.1	245.4 ± 29
9	>500	205.1 ± 23.5	>500	>500	>500	10.4 ± 0.6	>500	>500	>500	>500

^aDetails of the enzyme assays are provided under Experimental Section. IC₅₀ values are mean ± standard deviation (SD) of the results from 3 independent experiments.

After validating the PTP inhibition potency of the novel pTyr mimetics, their polarity and cell permeability were also characterized. As suggested by the calculated physicochemical properties including topological polar surface area (tPSA) and cLogP (Table 2), all newly obtained pTyr mimetics are less polar than Ac-F2Pmp-OMe (compound 1) and PhFP (compound 2), which confirmed our hypothesis that fused aromatic ring systems can better balance the high polarity of the difluorophosphonic acid. Expectedly, the decreased polarity of these pTyr mimetics led to a significantly improved cell permeability in our PAMPA assay. As shown in Table 2, Ac-F2Pmp-OMe and PhFP have permeability coefficient (P_e) values of 0.6 × 10⁻⁶ cm/s and 0.5 × 10⁻⁶ cm/s, respectively, indicating both are poorly cell permeable (compounds with P_e > 1.5 × 10⁻⁶ cm/s are defined as cell permeable^76,77). On the other hand, the cell permeability of the new pTyr mimetics was significantly improved with P_e values between 1.5 × 10⁻⁶ cm/s and 1.9 × 10⁻⁶ cm/s. Collectively, we showed that the new phosphonodifluoromethyl-containing bicyclic/tricyclic aryl derivatives are genuine PTP inhibitors with increased binding affinity and selectivity than their single-ring counterparts. They also possess improved cell permeability.

Table 2.

Calculated and determined physicochemical properties of the selected compounds

Compound	M.Wa	tPSAb	cLogPc	Pe(×10−6 cm/s)d
1	351.1	112.9	1.1	0.6
2	208.0	57.5	1.9	0.6
3	325.9	86.2	2.6	1.7
4	248.0	70.7	2.7	1.5
5	264.1	57.5	3.1	1.6
6	293.0	70.4	3.1	1.9
7	391.9	57.5	5.0	1.7
8	409.9	57.5	4.8	1.6
9	298.0	70.7	3.8	1.7
Verapamil	454.3	64.0	5.1	12.9
Furosemide	330.7	122.6	1.8	0.3

^aMolecular weight.

^bTopological PSA (Ų).

^cMolinspiration calculated Log P.

^dPermeability coefficient determined with PAMPA assay in pH 7.4 PBS buffer. The calcium channel blocker Verapamil and loop diuretic Furosemide were used as positive control and negative control, respectively.

Compound 9 Competitively and Selectively Inhibited Both hCDC14A and B

Among the newly obtained pTyr mimetics, compound 9 inhibited hCDC14A with an IC₅₀ value of 10.4 ± 0.6 μM. Further characterization revealed that compound 9 also inhibited hCDC14B with comparable potency (IC₅₀ = 11.2 ± 0.6 μM). Subsequently, we determined the mode of hCDC14A/B inhibition by compound 9 through steady-state kinetic analyses. As shown in Figures 2A and B, the Lineweaver–Burk plots showed that compound 9 is a classic competitive inhibitor that only affects the apparent K_m value, while the V_max remains unchanged. This observation is consistent with the expectation that compound 9 binds to the active site of hCDC14A/B due to its pTyr mimetic properties. Consistent with competitive inhibition, compound 9 displays inhibition constant (K_i) values of 5.8 ± 0.9 and 7.3 ± 0.9 μM for hCDC14A and B, respectively.

Figure 2.

Compound 9 competitively inhibited both hCDC14A and B with selectivity over the other 10 major mammalian PTPs. (A-B) Effect of compound 9 on hCDC14A (A) and hCDC14B (B) catalyzed DiFMUP hydrolysis. Compound 9 concentrations were 0 (●), 3 (■), 6 (▲), 9 (▼), and 12 (◆) μM, respectively. The Lineweaver-Burk plot displayed the characteristic intersecting line pattern, consistent with competitive inhibition. (C) Selectivity of compound 9 over a panel of 10 mammalian PTPs. IC₅₀ values of compound 9 tested against indicated PTPs. IC₅₀ values are mean ± standard deviation (SD) of the result from 3 independent experiments. (D) Putative binding pose of compound 9 (green stick) bound to hCDC14B (PDB 1OHC) colored by element with the nearby P+1 pocket shown in the transparent surface. Hydrogen bonds and ionic bonds are represented by yellow dashes. Cation-π interactions are shown with purple dashes. The 4-position of the dibenzofuran centroid points towards the P+1 pocket. (E) P-loop sequence alignment of hCDC14A, hCDC14B, and additional DUSPs (MKP5, Laforin, and VHZ). UniProt accession codes used for alignment are as follows: CDC14A – Q9UNH5; CDC14B – O60729; MKP5 – Q9Y6W6; Laforin – O95278; VHZ - Q9BVJ7. Key: red box with white text = identical residues; white box with black text = different residues; white box with red text = similar residues; text outlined with blue box = similar residues across the group. Figure was prepared using ESPript 3.0.

In addition to its notable hCDC14A/B inhibition potency, compound 9 also displays at least 20-fold selectivity over the other tested PTPs (Figure 2C), including both closely related dual-specificity phosphatases and more distantly related classic pTyr-specific PTPs. To gain insights into the molecular basis of selective hCDC14A/B inhibition by compound 9, and guide further modification, molecular docking studies were performed using Glide⁷⁸. The hCDC14A and B active sites are strongly conserved, and no hCDC14A crystal structure exists. In light of this, our molecular docking studies were performed on an existing hCDC14B crystal structure (PDB code 1OHC). As shown in Figure 2D, compound 9 makes extensive interactions with hCDC14B. Expectedly, the phosphonate group of compound 9 forms an ionic bond with the R320 side chain and several hydrogen bonds with the amide -NH backbones of K315, A316, G319, and R320. Further elevating compound 9’s binding affinity, the dibenzofuran core forms two cation-π interactions with the K315 side chain. Lastly, the dibenzofuran core also makes hydrophobic contact with the alkane side chains of A316 and I353. As shown in the sequence alignment (Figure 2E), the K315 residue in human CDC14A and B, which is involved in two cation-π interactions with compound 9, is absent in the other closely related DUSPs, which could largely contribute to the binding selectivity of compound 9 for hCDC14A and B. Moreover, despite the PTPs and DUSPs sharing the conserved signature catalytic motif HCXXGXXRS(T) and exhibiting similar mechanisms of catalysis²¹, the active site topology of these two classes of phosphatases is distinctive. Extensive structural biology studies have previously revealed that the active site in DUSPs is wide and shallow for the hydrolysis of phosphorylated serine, threonine, or tyrosine protein residues, whereas the pTyr-specific PTPs possess deeper active sites that restrict substrate specificity to only phosphotyrosine^79–81. As shown in Figure S2, we docked compound 9 into a classical nonreceptor PTP (PTP1B) and a receptor-like PTP (CD45). In both cases, compound 9 sterically clashes with multiple active-site residues as a result of poor steric complementarity between the ligand and these pTyr-specific active sites. Moreover, the planarity of the dibenzofuran core is distorted to allow binding in the obtained docking poses, further highlighting unfavorable binding to PTP1B and CD45. To gain additional insights into the selectivity for hCDC14 among the various tested DUSPs, we performed docking studies on three dual-specificity phosphatases including Laforin, VHZ, and MKP5. The docking studies revealed that, while the Laforin active site is expectedly broader than that of the pTyr-specific PTP1B and CD45 and better able to accommodate compound 9, it lacks the key lysine residue (Figure 2E), and the ligand clashes with a nearby active site tyrosine residue. Notably, docking studies for VHZ and MKP-5 failed to generate a putative pose due to the inability of compound 9 to fit into their active sites, which are significantly smaller than the large peptide-binding groove of hCDC14B. Computational docking studies show that this poor steric complementarity and clash is a result of the perpendicular 1-phosphonodifluoromethyl substitution of compound 9. In contrast, due to the more linear alignment of the phosphonodifluoromethyl head and the aromatic ring system, compounds 7 and 8 exhibited moderate inhibition potency against most of the tested PTPs. These results could help explain the selectivity of compound 9 over both the tested pTyr-specific PTPs and DUSPs.

Fragment-Based Elaboration of Compound 9 Led to Improved hCDC14A/B Inhibitors

To further improve the hCDC14A and B inhibition potency and amplify inhibition selectivity over other PTPs, fragment- and structure-based elaboration of compound 9 was deployed. According to the docking result (Figure 2D), the 4-position of compound 9 pointed toward the aromatic/hydrophobic P+1 pocket of hCDC14B (also exists in hCDC14A). This pocket is composed of aromatic residues F85, Y86, and Y170, all of which are located in the A-domain that is unique to the CDC14s²³. L318 and I353 also contribute to this hydrophobic pocket. Previous studies have shown that this P+1 pocket contributes to the substrate selectivity of CDC14s²³, therefore, we envisioned that introducing hydrophobic, aromatic functional groups at the 4-position of compound 9 could establish extra hydrophobic/π-π interactions with the P+1 pocket, thus improving hCDC14A/B binding potency and selectivity.

As shown in Table 3, with compound 10 as the starting material, a variety of aromatic rings were introduced to the 4-position of compound 9 and rapidly yielded compounds 11-17. Compared with the parent compound 9 and the 4-bromo substituted derivative compound 10, directly attaching a benzene ring (compound 11) increases the hCDC14A/B inhibition potency by about 5-fold. Substituting this benzene ring with either an electron donating group (compound 12) or electron withdrawing group (compound 13), as well as replacing it with a substituted heteroaromatic ring, 2-chlorothiophene (compound 14) has a very marginal impact on the inhibition potency. Considering the docking pose of compound 9 and the nearby hydrophobic P+1 pocket, these results implied that the directly attached aromatic rings only participated in hydrophobic contacts with the pocket but were not ideally situated for the desired π-π stacking interaction with the deeper posited aromatic residues in the P+1 pocket. In contrast, compound 15 has a 4-styrene substitution and inhibits hCDC14A and B with IC₅₀ values of 0.09 ± 0.01 μM and 0.22 ± 0.02 μM, respectively. While the 4-styrene moiety enjoys a 108-fold increase in binding affinity, attaching phenyl rings via an alkyne or alkene linker (compounds 16 and 17) likely precludes the phenyl functionality from engaging the P+1 pocket and forming π-π stacking interactions. As a result, compared to compound 15, these compounds exhibited decreased hCDC14A/B inhibition potency.

Table 3.

Structures and IC₅₀s of compound 9 and its derivatives on hCDC14A and B.^a

Compound	R Structure	IC50, μM		Compound	R Structure	IC50, μM
		hCDC14A	hCDC14B			hCDC14A	hCDC14B
9	-H	9.85 ± 0.30	12.78 ± 0.41	14		3.21 ± 0.25	5.85 ± 0.33
10	-Br	9.41 ± 0.41	11.88 ± 0.44	15		0.09 ± 0.01	0.22 ± 0.02
11		1.92 ± 0.16	2.17 ± 0.24	16		2.00 ± 0.42	4.19 ± 0.17
12		2.17 ± 0.24	5.21 ± 0.46	17		2.74 ± 0.27	3.77 ± 0.39
13		2.78 ± 0.44	3.76 ± 0.25

^aDetails of the enzyme assays are provided under Experimental Section. IC₅₀ values are mean ± standard deviation (SD) of the results from 3 independent experiments.

Compound 15 is the Most Potent and Selective hCDC14A/B Inhibitor

After obtaining the most potent compound 15, we next sought to confirm the mode of hCDC14A/B inhibition by this molecule. Expectedly, compound 15 behaved as a reversible and competitive inhibitor for both hCDC14A and B with K_i values of 57.4 ± 7.6 and 90.0 ± 8.8 nM, respectively (Figure 3A and B). To be qualified as a chemical probe that can be applied to interrogate the function and therapeutic potential of hCDC14A/B, the inhibition specificity of compound 15 is of utmost importance. To demonstrate this, the inhibition activity of compound 15 against a panel of 16 representative PTPs including the closely related DUSPs, more distantly related non-receptor PTPs, receptor-like PTPs, and low molecular weight PTP, was profiled (Figure 3C). To our delight, compound 15 showed at least 50-fold selectivity over all of the PTPs tested. Although the IC₅₀ values of compound 15 on all enzymes were measured with 0.01% Triton X-100 in the assay buffer, which is known to be capable of eliminating most of the false positive caused by the self-aggregation of organic molecules⁸², we still confirmed that compound 15 was not an aggregator with the DLS-based small molecule aggregation detection method. As shown in Figure 3D, no DLS signal decay was observed with 100 μM of compound 15 solution. This DLS testing concentration of compound 15 is higher than its IC₅₀s on all the tested PTPs, suggesting that compound 15 is a bona fide inhibitor. We further validated the target engagement of compound 15 with the Differential Scanning Fluorimetry (DSF) assay. As shown in Figure 3E, the melting temperature (T_m) for hCDC14A was 44.7 ± 0.1 °C. Compound 15 treatment increased the thermal stability of hCDC14A and resulted in a positive thermal shift with ΔT_m of 3.4 ± 0.3 °C, hence validating direct binding between compound 15 and the hCDC14A protein. Due to the limited hCDC14B protein solubility and the relatively high protein concentration required for the DSF assay, we did not run this assay with hCDC14B.

Figure 3.

Compound 15 is a potent and selective competitive inhibitor for human CDC14A and B. (A-B) Effect of compound 15 on hCDC14A (A) and hCDC14B (B) catalyzed DiFMUP hydrolysis. For CDC14A K_i measurement, compound 15 concentrations were 0 (●), 25 (■), 50 (▲), 75 (▼), and 100 (◆) nM, respectively. For CDC14B K_i measurement, compound 15 concentrations were 0 (●), 100 (■), 200 (▲), 300 (▼), and 400 (◆) nM, respectively. The Lineweaver-Burk plot displayed the characteristic intersecting line pattern, consistent with competitive inhibition. (C) Selectivity of compound 15 over a panel of 16 mammalian PTPs. IC₅₀ values are mean ± standard deviation (SD) of the result from 3 independent experiments. (D) DLS signal of 100 μM compound 15 dissolved in DMG assay buffer. (E) Thermal shift first-derivative curves of 5 μM hCDC14A (green) and 5 μM hCDC14A + 50 μM compound 15 (red).

To investigate the significant improvement in potency and selectivity conferred by the added styrene ring, we performed additional molecular docking studies on compound 15, and the result is shown in Figure 4A. Analogous to compound 9, the phosphonate group forms an ionic bond with the R320 side chain and several hydrogen bonds with the amide -NH backbones of K315, A316, G319, and R320. The dibenzofuran core forms two cation-π interactions with K315 while also making hydrophobic contacts with the alkane side chains of A316 and I353. According to our docking model, the added styrene moiety engages the hydrophobic P+1 pocket and is suitably oriented to form an edge-to-face π-π interaction with F85. Additional van der Waals interactions between the alkene and A168 and Y170 side chains further strengthen the binding affinity of compound 15. Notably, this P+1 pocket is only present in the CDC14s, thus explaining the over 50-fold selectivity observed for compound 15 (compared to over 20-fold for compound 9 which lacks the styrene ring). To assess the stability of our docking pose, we performed a 50 ns molecular dynamic (MD) simulation on the docked pose of compound 15 to hCDC14B. As shown in Figure S3, the ligand root mean square deviation (RMSD) (as aligned on the protein) reveals that the docked pose is stable and does not diffuse away during the simulation. Notably, the protein root mean square fluctuation (RMSF) during the simulation correlates well with the experimental B factor, a measure of protein flexibility, in hCDC14B. Alignment of our docking result and the cocrystal structure of hCDC14B complexed with a phosphoserine peptide that is known to engage the P+1 pocket (Figure 4B) show that the styrene ring binds deeper into the P+1 pocket compared to the prolyl ring of the peptide, thus making greater hydrophobic contacts. Unlike the prolyl ring side chain, styrene is also able to participate in the aforementioned π-π interaction, which contributes to the high potency and selectivity of the scaffold. Notably, our docking model does not predict this π-π stacking interaction to be formed by the scaffolds with a directly attached aromatic ring (compounds 11-14, Table 3) or those with an alkene (16) or alkyne (17) linker. These findings support the biochemical potencies reported in Table 3 and suggest that these other derivatives only gain additional hydrophobic contacts with the hCDC14 binding site.

Figure 4.

The docking pose of compound 15 bound to hCDC14B suggests engagement with the P+1 pocket. (A) Docking pose of compound 15 (cyan stick) with hCDC14B (PDB code 1OHC) colored by element. Hydrogen bonds and ionic bonds are depicted by yellow dashes, cation-π interactions are depicted by purple dashes, and the edge-to-face π-π stacking is shown with a red dash. (B) Docking pose of compound 15 (cyan stick) aligned to the cocrystal structure of hCDC14B (surface colored by element) and a phosphoserine peptide ligand (green stick) (PDB code 1OHE).

Mutagenesis of Key Binding Residues in the P+1 Pocket in hCDC14A and B Supports the Binding Mode of Compound 15

To ascertain the importance of the unique P+1 pocket for the binding of compound 15 and further corroborate our docking model, we carried out site-directed mutagenesis experiments. In particular, we individually substituted the Phe48 (hCDC14A)/Phe85 (hCDC14B) residues, which were anticipated to capture π-π stacking interactions with the benzene ring of the styrene group in compound 15, to alanine. We also replaced the Phe134 (hCDC14A)/Tyr170 (hCDC14B) residues, which were predicted to establish hydrophobic contacts with the ethylene of the styrene group in compound 15, with alanine residues. Additionally, as shown in Figure S4, residues of the P+1 pocket in hCDC14A and B are relatively conserved. Among the P+1 pocket residues that directly interact with compound 15, Phe134 (hCDC14A)/Tyr170 (hCDC14B) is the only pair of amino acids that is distinct between these two enzymes, which can potentially contribute to the slight (2.4-fold) inhibition preference of compound 15 for hCDC14A. Notably, this sequence difference results in a more lipophilic P+1 pocket for hCDC14A compared to hCDC14B due to the absence of the polar hydroxyl group. To determine the impact of these residues on the selectivity of compound 15, we also mutated the Phe134 residue in hCDC14A to tyrosine to mimic the structure of hCDC14B.

The wild-type and mutants of hCDC14A and B were expressed in E. coli and purified to homogeneity. As shown in Table S1, the kinetic parameters of hCDC14A F134A and F134Y mutants and hCDC14B Y170A mutant are comparable to those of the wild type hCDC14A and B, indicating that these mutations do not perturb the overall enzyme structure and function. On the other hand, the K_m values of hCDC14A F48A and hCDC14B F85A mutants are similar to the corresponding wild-type enzyme albeit the k_cat values for these mutants are lower than those of the wild-type enzymes. A similar observation was also documented in an earlier publication, wherein the hCDC14B F85A mutant displayed decreased k_cat when assessed with a phosphopeptide as substrate.²³ After characterizing these mutants, we determined the impact of altering the above-mentioned P+1 site residues on the IC₅₀ values of compounds 15. As shown in Table 4, compared to the wild-type enzymes, hCDC14A F48A and hCDC14B F85A mutants displayed reduced sensitivity towards compound 15, resulting in IC₅₀ values that were 27.3- and 15.6-fold higher, respectively. To ascertain that this diminished effectiveness of compound 15 could be attributed to the absence of the π-π interaction, we also assessed the IC₅₀ of vanadate (a known general phosphate mimicking PTP competitive inhibitor^83,84), PhFP, and compound 9 (both PhFP and compound 9 are not supposed to engage the P+1 pocket) for the hCDC14A F48A and hCDC14B F85A mutants. As anticipated, the IC₅₀ values of these compounds for the hCDC14A F48A and hCDC14B F85A mutants were comparable to those observed with the corresponding wild-type enzymes (data shown in Table S2). Additionally, the hCDC14A F134A and hCDC14B Y170A mutants, which disrupt the aforementioned hydrophobic interaction with the styrene alkene, experienced a 2 to 3-fold reduction in the inhibitory potency of compound 15 (Table 4). These data supported the proposed binding mode for compound 15, namely it interacts with both the active site and the nearby P+1 Pro binding pocket. Finally, the introduction of the F134Y mutation in hCDC14A resulted in an IC₅₀ value of 0.19 ± 0.02 μM, which is comparable to the IC₅₀ value of compound 15 observed for the wild-type hCDC14B (0.22 ± 0.02 μM). These results indicated that the Phe134 (hCDC14A)/Tyr170 (hCDC14B) residues indeed contribute to the slight (2.4-fold) hCDC14A binding preference of compound 15 for hCDC14A.

Table 4.

Comparison of IC₅₀ data of compound 15 against wild-type and mutant hCDC14A and B.^a

Enzyme	Mutation	IC50,μM
CDC14A	WT	0.09 ± 0.01
	F48A	2.46 ± 0.21
	F134A	0.26 ± 0.03
	F134Y	0.19 ± 0.02
CDC14B	WT	0.22 ± 0.02
	F85A	3.43 ± 0.28
	Y170A	0.51 ± 0.03

^aIC₅₀ values are mean ± standard deviation (SD) of the result from 3 independent experiments.

Compound 15 is Cell Active and Bioavailable

Previous studies have revealed that hCDC14 dephosphorylates Ser315 of p53³⁷. Genomic stresses such as DNA damage induce p53 Ser315 phosphorylation, which promotes the binding of Pin1, leading to a conformational change that facilitates additional modifications and influences its interaction with Mdm2, a regulatory partner of p53, resulting in full stabilization of p53.^39,40 To determine the cellular efficacy of our hCDC14A/B inhibitor compound 15, we treated the U2OS cells with the DNA damage reagent Mitomycin C along with various doses of compound 15 for 24 hours. As expected, Mitomycin C treatment causes p53 stabilization along with increased Ser315 phosphorylation (Figure 5A). compound 15 dose-dependently increased both p53 Ser315 phosphorylation and p53 protein levels with maximum 20- and 13-fold enhancement, respectively (Figure 5A). To increase confidence that the engagement of the hCDC14A/B is responsible for the observed p53 Ser315 phosphorylation enhancement and p53 stabilization, we designed and prepared a negative control compound 18 (structure shown in Figure 5A), which is closely related in structure to compound 15 but showed no hCDC14A/B inhibition potency up to 100 μM due to the lack of the PTP active site binding-essential difluorophosphonic acid moiety. Satisfactorily, compound 18 showed negligible impact on total p53 and p53 Ser315 phosphorylation levels at 20 μM concentration. Taken together, these results agreed with previous studies and indicated that hCDC14s mediated dephosphorylation of p53 on Ser315 was abolished upon compound 15 treatment.

Figure 5.

Compound 15 is a cell active and bioavailable hCDC14A and B inhibitor. (A) Immunoblots of whole cell lysates from U2OS cells with indicated treatment for 24 hours. Compound 15 was able to further amplify Mitomycin C-induced p53 ser315 phosphorylation and stabilize p53 protein, whereas no effect on the negative control (compound 18) treated cells were observed. (B) In vivo pharmacokinetics data based on mass spectrometry quantification at 0, 0.5, 1, 3, 6, and 24 h from the time of intraperitoneal injection (I.P.) or oral administration (P.O.) of compound 15 with 50 mg/kg dose. Values are mean ± standard deviation (SD), n=3.

Having obtained a first-in-class, potent, selective, and cell-active hCDC14A and B inhibitor, we next preliminarily assessed the safety and pharmacokinetics (PK) of compound 15 to enable and support future in-depth mechanistic studies. For safety assessment, the impact of compound 15 treatment on the proliferation of 4 human and murine cell lines (HEK293, U2OS, Jurkat, and Raw 264.7) was tested. As shown in Figure S5, following a 24-hour treatment period, compound 15 showed no considerable cytotoxic effects across all tested cells at a maximum concentration of 50 μM. To characterize the pharmacokinetics, compound 15 was dosed by intraperitoneal injection (I.P.) or oral administration (P.O.) in C57BL/6 mice. As shown in Figure 5B, a single I.P. injection of compound 15 at 50 mg/kg achieved a peak plasma concentration (C_max) of 21.9 ± 3.7 μM with a half-life of 2.2 ± 0.4 hours. Moreover, compound 15 was also orally bioavailable. A single oral administration of compound 15 at 50 mg/kg gave a C_max of 4.1 ± 0.6 μM with a half-life of 3.0 ± 1.2 hours (Figure 5B). Overall, compound 15 showed a good PK profile with acceptable exposure and elimination to enable its use as a chemical probe for future in vivo studies.

Chemistry

As depicted in Scheme 1, the P-(Difluorophenylmethyl) phosphonic acid (PhFP, compound 2) and compound 2–10 were synthesized via a CuI-promoted coupling reaction⁸⁵ followed by the hydrolysis of the phosphonates. Briefly, the cadmium-containing compound 19 was first prepared by stirring a DMA solution of diethyl bromodifluoromethylphosphonate with cadmium shots under argon. Subsequently, compound 19 was coupled to commercially available iodo-substituted aromatic rings in the presence of copper (I) iodide to yield intermediates 2a-10a. After 19 hours of reaction at 50 ^°C, the yields of indicated compounds were between 63% and 74% (determined by LC-MS). Next, compounds 2a-10a were dissolved in DCM and treated with trimethylsilyl iodide to furnish the final pTyr mimetics 2-10.

Scheme 1.

Synthesis of intermediates 2a-10a and pTyr mimetics 2–10^a

^aReagents and conditions: (a) cadmium shot, DMA, argon, rt, 3 h. (b) compound 19, CuBr, argon, DMA, rt to 50 °C, 15 h. 63–74%. (c) TMSI, DCM, 0 °C, 4 h. 41–64%.

The synthesis of Ac-F2Pmp-OMe (compound 1) was depicted in Scheme 2. Starting with the commercially available starting material N-[(1,1-dimethylethoxy)carbonyl]-4-iodo-L-phenylalanine methyl ester, the phosphonate intermediate 20 was obtained with the above-described CuI-promoted coupling reaction. Ac-F2Pmp-OMe (compound 1) was obtained by the sequence of Boc removal, acetylation of the N-terminal amine, and hydrolysis of the diethyl phosphonate.

Scheme 2.

Synthesis of Ac-F2Pmp-OMe (1) ^a

^aReagents and conditions: (a) compound 19, CuBr, argon, DMA, rt to 50 °C, 19 h. 77%. (b) 20% TFA/DCM, rt, 6h. (c) acetyl chloride, triethylamine, DCM 0 °C to rt, 4 h. 64% (b and c 2 steps). (c) TMSI, DCM, 0 °C, 4 h. 59%.

The synthesis of compound 11-18 was depicted in Scheme 3. Started with compound 10, the substitutions on the 4-position of the dibenzofuran rings were introduced through palladium-catalyzed Suzuki (compound 11–16 and compound 18) or Sonogashira cross-coupling (compound 17) reactions with the corresponding boronic acids or alkyne. Similarly, the negative control compound 18 was synthesized from 4-bromodibenzo[b,d]furan by the Suzuki coupling reaction.

Scheme 3.

Synthesis of compound 11–18 ^a

^aReagents and conditions: (a) Aryl boronic acid, K₂CO₃, Pd(PPh₃)₄, argon, DMF, H2O, 90°C, 16h, 39–49%. (b) ethynylbenzene, CuI, Pd(PPh₃)₄, triethylamine, argon, DMF, 90°C, 16h, 56%. (c) K₂CO_3, Pd(PPh₃)₄, argon, DMF, H₂O, 90°C, 16h, 71%.

Conclusions

PTPs are pivotal regulators of protein tyrosine phosphorylation, which plays indispensable roles in modulating cellular signaling. Abnormal perturbation of the level of protein tyrosine phosphorylation is intricately associated with the pathogenesis of various human diseases. Gaining an improved understanding of the involvement of PTPs in dysregulated signaling events provides valuable insights into the etiology of diseases and holds the potential to unveil novel therapeutic targets. In contrast to the extensively studied PTKs, research on PTPs is relatively underdeveloped, which is partially due to the limited availability of high-quality small molecule inhibitors that can serve as invaluable chemical probes to interrogate PTP function and evaluate therapeutic hypothesis. The challenge of identifying active site-directed, potent, selective and pharmacologically efficacious inhibitors for PTPs is well known in drug discovery. Although success in harnessing PTP allosteric regulatory mechanisms for PTP inhibitor development has been documented in a few cases^17–20, the discovery of allosteric inhibitors has been mostly serendipitous. Previous efforts aimed at the highly conserved and positively charged PTP active sites have often led to predominantly negatively charged inhibitors with limited cell permeability, which makes them unsuitable as chemical probes and therapeutics. Thus, an improved general approach targeting the PTP active site is highly desirable for the development of orthosteric PTP inhibitors as chemical probes to advance our understanding of PTPs in disease biology and target validation.

We have previously shown that despite the significant structural conservation among the PTP active sites, potent, selective and bioavailable inhibitors for PTPs can be generated by employing fragment-based approaches to capitalize on subtle sequence differences in the outer region of the active site.¹ To that end, we have developed a number of highly selective and efficacious orthosteric PTP inhibitors by tethering a negatively charged pTyr mimetics, such as F2Pmp, salicylic acids and α-sulfophenylacetic amide with appropriate diversity fragments to engage both the active site and nearby unique peripheral pockets.^1,10,57–63 Excitingly, an orally bioavailable acidic thiadiazolidinone dioxide moiety-containing PTP1B/TC-PTP active site inhibitor, ABBV-CLS-484, has entered clinical evaluation for cancer immunotherapy in human⁶⁴, which significantly boosts the druggability of targeting the PTP active site. To further enhance our ability to target the PTP active site, we sought to develop novel nonhydrolyzable pTyr mimetics with improved binding affinity and cell permeability.

Here, starting with a classic nonhydrolyzable pTyr mimetic, F2Pmp, we synthesized 7 novel phosphonodifluoromethyl-containing bicyclic/tricyclic aryl derivatives in order to better balance the lipophilicity and binding affinity. As expected, these new bicyclic/tricyclic pTyr mimetics display increased cell permeability and greater intrinsic PTP binding potency. Among these derivatives, compound 9 with a dibenzofuran scaffold exhibited promising hCDC14A and B inhibition potency and exceptional selectivity over 10 other PTPs. Subsequent molecular docking studies elucidated the binding mode between compound 9 and hCDC14B, explained the hCDC14A/B biased inhibition potency, and guided the following fragment-based optimization. Among the compound 9 derivatives, compound 15 showed a K_i of 57.4 ± 7.6 and 90.0 ± 8.8 nM for hCDC14A and B, respectively, with >50-fold selectivity over a panel of 16 major mammalian PTPs. Through kinetic and biophysical analyses, molecular docking, and site-directed mutagenesis, compound 15 was confirmed to be a genuine active site-directed competitive inhibitor of both hCDC14A and hCDC14B. Consistent with the design strategy, the extraordinarily high selectivity of compound 15 for the CDC14 phosphatases derives from its unique interactions with the P+1 pocket in addition to the active site. Moreover, despite its negative charges, compound 15 can engage its target and block CDC14-mediated p53 dephosphorylation in a cell-based assay with low μM efficacy. Lastly, compound 15 exhibits good pharmacokinetic properties and oral bioavailability. Collectively, the results qualify compound 15 as a chemical probe for the CDC14 phosphatases because it meets the specific criteria recommended for a small molecule probe^86,87. Given its potency, selectivity, and in vivo bioavailability, compound 15 holds great potential as a molecular probe for further investigation of the roles of hCDC14A and hCDC14B in normal physiological processes and diseases. Additionally, it could be utilized to validate the therapeutic potential of targeting these enzymes. Finally, we demonstrate that the newly developed phosphonodifluoromethyl-containing bi/tri-cyclic pTyr mimetics are sufficiently polar to bind the PTP active site yet remain capable of efficiently crossing cell membranes. Importantly, these novel nonhydrolyzable pTyr mimetics offer new starting points for the development of active site-directed inhibitors for other PTPs of therapeutic interest. Our work also furnishes another example of the acquisition of highly potent, selective, and efficacious active site-directed PTP inhibitors by exploiting the specific structural features of different PTP active sites and less conserved peripheral pockets.

Experimental Section

General Information

Unless otherwise noted, all reagents and starting materials were purchased from commercial suppliers and used without further purification. Thin-layer chromatography was performed using glass precoated Merck silica gel 60 F254 plates. Flash column chromatography was performed on Biotage prepacked columns using the automated flash chromatography system Biotage Isolera One. Organic solvents were evaporated using rotary evaporation at 50–55 °C. The 1H- and 13C NMR spectra were recorded on a Bruker AVANCE 500 MHz spectrometer using CDCl₃ or dimethyl sulfoxide (DMSO-d₆) as the solvent. Chemical shifts are expressed in ppm (δ scale) and referenced to the residual protonated solvent. Peak multiplicities are reported using the following abbreviations: s (singlet), d (doublet), t (triplet), q (quartet), m (multiplet), or br (broad singlet). Low-resolution mass spectra and purity data were obtained using an Agilent Technologies 6470 series, triple quadrupole LC–MS. The purity of all final tested compounds was determined to be >95% (UV, λ = 254 nm). Anti-p53 (catalog# sc-126), antiphospho-p53 (catalog# sc-135772), and anti-GAPDH (catalog# SC-59541) antibodies were purchased from Santa Cruz. 6,8-Difluoro-4-Methylumbelliferyl Phosphate (DiFMUP) was purchased from Invitrogen (catalog# D6567). bpV(Phen) was purchased from Sigma-Aldrich (catalog# SML0889). Mitomycin C was purchased from Thermo Scientific (catalog# J63193). Phosphatase inhibitor cocktail was purchased from Selleck Chemicals (catalog# B150001). Protease inhibitor mixture was purchased from Roche Applied Science (catalog# 4693116001)

Chemical Synthesis

Synthesis of cadmium-containing intermediate (19)

Cadmium metal (6.00 g, 53.4 mmol) was washed by stirring under argon with 1 N HCl (5 mL, 15 min), H₂O (5 mL, 3 × 1 min) and acetone (5 mL, 3 × 1 min). The metal was dried overnight under high vacuum while stirring until a metallic shine was observed. While warming the reaction vessel by hand, anhydrous DMF (10 mL) and diethyl bromodifluoromethylphosphonate (12.5 mL, 31.2 mmol) were added dropwise over 15 min to the metal under argon. The slightly exothermic reaction proceeded and was allowed to stir at r.t. for 3 h. The supernant of this reaction was directly used in the following step.

Synthesis of Methyl (S)-2-((tert-butoxycarbonyl)amino)-3-(4-((diethoxyphosphoryl)difluoromethyl)phenyl)propanoate (20)

N-[(1,1-Dimethylethoxy)carbonyl]-4-iodo-L-phenylalanine methyl ester (4.10 g, 10.0 mmol, 1.0 eq) and CuBr (2.87 g, 20.0 mmol, 2 equiv.) were suspended in anhydrous DMF (100 mL) under argon atmosphere, then the supernatant solution of the Cd reagent 19 (ca. 1.7 equiv.) was added in a dropwise manner. After 3 hours, further CuBr (1.43 g, 10.0 mmol, 1 equiv.) and the Cd reagent solution (ca. 1.7 equiv.) were added. The reaction was allowed to stir at r.t. for 19 hours in total. The progress of the reaction was monitored by LC-MS. Upon completion, the reaction mixture was diluted with EtOAc (500 mL), filtered through Celite, and extracted with aq NH₄Cl (2 × 500 mL) and brine (500 mL). The organic layer was dried over anhydrous Na₂SO₄ and evaporated in vacuo. The product was then purified by flash chromatography (EtOAc/Hexanes, 0%→40%). Yield 3.58 g (77%). ¹H NMR (500 MHz, CDCl₃) δ 7.52 (d, J = 8.0 Hz, 2H), 7.20 (d, J = 7.9 Hz, 2H), 5.01 (d, J = 8.3 Hz, 1H), 4.57 (q, J = 6.8 Hz, 1H), 4.20 – 4.07 (m, 4H), 3.68 (s, 3H), 3.20 – 2.99 (m, 2H), 1.39 (s, 9H), 1.27 (tt, J = 7.2, 1.6 Hz, 6H). MS (ESI): calcd for C₂₀H₃₀F₂NO₇P [M + H]⁺ 466.17; found, 466.16.

Synthesis of Methyl (S)-2-acetamido-3-(4-((diethoxyphosphoryl)difluoromethyl)phenyl)propanoate (21)

To a solution of Intermediate 20 (3.58 g, 7.7 mmol) in DCM (40 mL), trifluoroacetic acid (10 mL) was added and stirred for 6 hours at rt. Then the excess reagent and solvent were evaporated under reduced pressure to give the crude deprotected amine, which was used in the next step without further purification. Acetyl chloride (824 μL, 11.55 mmol,1.5 equiv.) was added at 0 °C to a stirred solution of the deprotected amine and triethylamine (1.6 mL, 11.55 mmol, 1.5 equiv.) in dry DCM (50 mL). The reaction mixture was then allowed to warm to rt and stirred for 4 hours. Upon completion, the reaction mixture was diluted with EtOAc and washed with brine. The organic layer was dried over anhydrous Na₂SO₄ and evaporated in vacuo. The product was then purified by flash chromatography (EtOAc/Hexanes, 0%→40%). Two steps yield 2.01 g (64%). ¹H NMR (500 MHz, CDCl₃) δ 7.51 (d, J = 7.4 Hz, 2H), 7.18 (d, J = 8.2 Hz, 2H), 6.24 (d, J = 7.8 Hz, 1H), 4.85 (dt, J = 7.7, 5.9 Hz, 1H), 4.21 – 4.05 (m, 4H), 3.68 (s, 3H), 3.20 – 3.04 (m, 2H), 1.95 (s, 3H), 1.28 (td, J = 7.1, 4.2 Hz, 6H). MS (ESI): calcd for C₁₇H₂₄F₂NO₆P [M + H]⁺ 408.13; found, 408.12.

Synthesis of (S)-((4-(2-acetamido-3-methoxy-3-oxopropyl)phenyl)difluoromethyl)phosphonic acid (Ac-F2Pmp-OMe, 1)

The solution of Intermediate 21 (2.01 g, 4.9 mmol) in anhydrous DCM (25 ml) was cooled to 0 °C and stirred vigorously. Then trimethylsilyl iodide (4.9 mL, 34.3 mmol, 7.0 eq) was added to the solution dropwise. The reaction was kept at 0 °C and monitored with LC-MS. Upon completion, the reaction mixture was added to a 50% MeCN/H₂O mixture dropwise and stirred at r.t. for 30 minutes. Then the water and organic solvent was evaporated in vacuo to give crude product which was further purified by prep HPLC. Yield 1.01 g (59%). ¹H NMR (500 MHz, DMSO-d₆) δ 8.37 (d, J = 7.7 Hz, 1H), 7.44 (d, J = 7.9 Hz, 2H), 7.31 (d, J = 8.0 Hz, 2H), 4.53 – 4.41 (m, 1H), 3.59 (s, 3H), 3.09 – 2.88 (m, 2H), 1.79 (s, 3H). MS (ESI): calcd for C₁₃H₁₆F₂NO₆P [M - H]⁺ 350.07; found, 350.08.

General procedure for the synthesis of intermediate 2a-10a

Iodo-substituted aromatic rings (0.2 mmol, 1.0 eq) and CuBr (57.4 mg, 0.4 mmol, 2.0 equiv.) were suspended in anhydrous DMA (2 mL) under argon atmosphere, then the supernatant solution of the Cd reagent 19 (ca. 1.7 equiv.) was added in a dropwise manner. After 3 hours, further CuBr (28.7 mg, 0.2 mmol, 1.0 equiv.) and the Cd reagent solution (ca. 1.7 equiv.) were added. The reaction was then allowed to stir at 50 °C for 12 hours. The progress of the reaction was monitored by LC-MS. Upon completion, the reaction mixture was diluted with EtOAc (50 mL), filtered through Celite, and extracted with aq NH₄Cl (2 × 50 mL) and brine (50 mL). The organic layer was dried over anhydrous Na₂SO₄ and evaporated in vacuo. The product was then purified by flash chromatography (EtOAc/Hexanes, 0%→40%).

Compound 2a. Yield 37.5 mg (71%). ¹H NMR (500 MHz, CDCl₃) δ 7.62 (m, 2 H), 7.46 (m, 3 H), 4.25–4.10 (m, 4 H), 1.29 (td, J = 7.1, 4.2 Hz, 6H). MS (ESI): calcd for C₁₁H₁₅F₂O₃P [M + H]⁺ 265.07; found, 265.08.

Compound 3a. Yield 52.9 mg (69%). ¹H NMR (500 MHz, CDCl₃) δ 10.21 (s, 1H), 8.25 (s, 1H), 7.83 (s, 1H), 7.53 (s, 1H), 4.34 – 4.20 (m, 4H), 1.36 (t, J = 7.1 Hz, 6H). (ESI): calcd for C₁₂H₁₄BrF₂O₃P [M + H]⁺ 382.99; found, 383.00.

Compound 4a. Yield 38.3 mg (63%). ¹H NMR (500 MHz, CDCl₃) δ 7.90 (s, 1H), 7.68 (d, J = 2.2 Hz, 1H), 7.56 (d, J = 1.2 Hz, 2H), 6.82 (d, J = 2.2 Hz, 1H), 4.25 – 4.09 (m, 4H), 1.33 – 1.28 (m, 6H). MS (ESI): calcd for C₁₃H₁₅F₂O₄P [M + H]⁺ 305.07; found, 305.09.

Compound 5a. Yield 44.8 mg (70%). ¹H NMR (500 MHz, CDCl₃) δ 8.09 (s, 1H), 7.95 (d, J = 8.5 Hz, 1H), 7.58 (d, J = 8.5 Hz, 1H), 7.52 (d, J = 5.5 Hz, 1H), 7.40 (d, J = 5.4 Hz, 1H), 4.26 – 4.10 (m, 4H), 1.31 (t, J = 7.1 Hz, 6H). MS (ESI): calcd for C₁₃H₁₅F₂O₃PS [M + H]⁺ 321.04; found, 321.06.

Compound 6a. Yield 45.5 mg (65%). ¹H NMR (500 MHz, CDCl₃) δ 9.13 (s, 1H), 8.29 (d, J = 8.3 Hz, 1H), 8.12 – 8.01 (m, 2H), 7.57 (dd, J = 8.4, 4.1 Hz, 1H), 4.30 – 4.18 (m, 4H), 1.34 (td, J = 7.1, 0.7 Hz, 6H). MS (ESI): calcd for C₁₄H₁₅ClF₂NO₃P [M + H]⁺ 350.04; found, 350.04.

Compound 7a. Yield 66.5 mg (74%). ¹H NMR (500 MHz, CDCl₃) δ 8.40 – 8.31 (m, 2H), 7.93 (d, J = 8.4 Hz, 1H), 7.77 – 7.70 (m, 2H), 7.59 (dd, J = 8.4, 1.9 Hz, 1H), 4.30 – 4.14 (m, 4H), 1.33 (td, J = 7.0, 0.6 Hz, 6H). MS (ESI): calcd for C₁₇H₁₆BrF₂O₃PS [M + H]⁺ 450.97; found, 450.99.

Compound 8a. Yield 61.7 mg (66%). ¹H NMR (500 MHz, CDCl₃) δ 7.84 (s, 1H), 7.80 – 7.75 (m, 2H), 7.66 – 7.62 (m, 2H), 7.49 (d, J = 8.1 Hz, 1H), 4.30 – 4.16 (m, 4H), 1.34 (t, J = 7.1 Hz, 6H). MS (ESI): calcd for C₁₈H₁₆BrF₄O₃P [M + H]⁺ 467.00; found, 467.03.

Compound 9a. Yield 51.0 mg (72%). ¹H NMR (500 MHz, CDCl₃) δ 8.31 (d, J = 7.2 Hz, 1H), 7.69 (d, J = 8.2 Hz, 1H), 7.63 – 7.44 (m, 4H), 7.36 (ddd, J = 8.2, 7.2, 1.2 Hz, 1H), 4.24 – 4.06 (m, 4H), 1.22 (td, J = 7.0, 0.7 Hz, 6H). MS (ESI): calcd for C₁₇H₁₇F₂O₄P [M + H]⁺ 355.08; found, 355.09.

Compound 10a. Yield 62.9 mg (71%). ¹H NMR (500 MHz, CDCl₃) δ 8.25 (d, J = 8.2, 1H), 7.65 (d, J = 8.3 Hz, 1H), 7.60 (d, J = 8.3 Hz, 1H), 7.50 – 7.40 (m, 2H), 7.34 (ddd, J = 8.2, 7.3, 1.1 Hz, 1H), 4.23 – 4.02 (m, 4H), 1.19 (t, J = 7.1 Hz, 6H). MS (ESI): calcd for C₁₇H₁₆BrF₂O₄P [M + H]⁺ 432.99; found, 432.98.

General procedure for the synthesis of compound 2–10

The solution of Intermediate 2a-10a (0.1 mmol) in anhydrous DCM (1 ml) was cooled to 0 °C and stirred vigorously. Then trimethylsilyl iodide (100 μL, 0.7 mmol, 7.0 eq) was added to the solution dropwise. The reaction was kept at 0 °C and monitored with LC-MS. Upon completion, the reaction mixture was added to a 50% MeCN/H₂O mixture dropwise and stirred at r.t. for 30 minutes. Then the water and organic solvent was evaporated in vacuo to give crude product which was further purified by prep HPLC.

Compound 2. Yield 13.3 mg (64%)¹H NMR (500 MHz, DMSO-d₆): δ = 7.47 (m, 5H); ¹³C NMR (126 MHz, DMSO-d₆): δ = 130.5, 129.2,126.7, 126.9, 126.5. MS (ESI): calcd for C₇H₇F₂O₃P [M - H]^- 207.01; found, 207.00.

Compound 3. Yield 17.9 mg (55%). ¹H NMR (500 MHz, DMSO-d₆) δ 8.11 (s, 1H), 7.71 (s, 1H), 7.41 (s, 1H). ¹³C NMR (126 MHz, DMSO-d₆) δ 153.72, 140.22, 133.57, 124.72, 121.11, 113.33, 108.67. MS (ESI): calcd for C₈H₆BrF₂N₂O₃P [M - H]^- 324.93; found, 324.91.

Compound 4 Yield 13.1 mg (53%). 1H NMR (500 MHz, DMSO-d₆) δ 8.06 (d, J = 2.2 Hz, 1H), 7.85 (s, 1H), 7.69 (d, J = 8.6 Hz, 1H), 7.48 (d, J = 8.6 Hz, 1H), 7.05 (d, J = 1.3 Hz, 1H). ¹³C NMR (126 MHz, DMSO-d₆) δ 141.20, 139.42, 129.27, 124.77, 122.95, 122.30, 121.88. MS (ESI): calcd for C₉H₇F₂O₄P [M - H]^- 247.01; found, 247.02.

Compound 5. Yield 16.1 mg (61%). ¹H NMR (500 MHz, DMSO-d₆) δ 8.09 (d, J = 8.4 Hz, 1H), 8.04 (s, 1H), 7.85 (d, J = 5.4 Hz, 1H), 7.57 (d, J = 5.4 Hz, 1H), 7.49 (d, J = 8.5 Hz, 1H). ¹³C NMR (126 MHz, DMSO-d₆) δ 155.46, 147.65, 129.22, 127.46, 122.93, 120.04, 119.71, 111.55, 107.48. MS (ESI): calcd for C₉H₇F₂O₃PS [M - H]^- 262.98; found, 262.99.

Compound 6. Yield 12.0 mg (41%). 1H NMR (500 MHz, DMSO-d₆) δ 9.09 (d, J = 3.2 Hz, 1H), 8.62 (d, J = 8.0 Hz, 1H), 8.19 (s, 1H), 7.96 (s, 1H), 7.72 (dd, J = 8.2, 4.1 Hz, 1H). ¹³C NMR (126 MHz, DMSO-d₆) δ 153.20, 144.29, 138.12, 133.09, 132.83, 128.91, 127.05, 126.19, 123.72. MS (ESI): calcd for C₁₀H₇ClF₂NO₃P [M - H]^- 291.98; found, 291.97.

Compound 7. Yield 20.1 mg (51%). ¹H NMR (500 MHz, DMSO-d₆) δ 8.74 (d, J = 1.9 Hz, 1H), 8.56 (s, 1H), 8.16 (d, J = 8.4 Hz, 1H), 8.04 (d, J = 8.5 Hz, 1H), 7.76 – 7.62 (m, 2H). ¹³C NMR (126 MHz, DMSO-d₆) δ 141.63, 138.48, 137.07, 134.07, 131.60, 130.56, 125.97, 125.60, 125.55, 123.52, 120.62, 120.35, 118.84. MS (ESI): calcd for C₁₃H₈BrF₂O₃PS [M - H]^- 390.91; found, 390.90.

Compound 8. Yield 23.4 mg (57%). ¹H NMR (500 MHz, DMSO-d₆) δ 8.03 – 7.99 (m, 2H), 7.89 (d, J = 8.1 Hz, 1H), 7.84 (dd, J = 8.1, 1.8 Hz, 1H), 7.77 – 7.72 (m, 2H). ¹³C NMR (126 MHz, DMSO-d₆) δ 140.21, 139.32, 137.58, 136.72, 136.46, 135.82, 131.45, 127.61, 124.53, 123.33, 122.59, 122.11, 121.87, 120.20. MS (ESI): calcd for C₁₄H₈BrF₄O₃P [M - H]^- 408.93; found, 408.92.

Compound 9. Yield 18.2 mg (61%). ¹H NMR (500 MHz, DMSO-d₆) δ 8.28 (d, J = 8.2 Hz, 1H), 7.86 (d, J = 8.2 Hz, 1H), 7.70 (d, J = 8.2 Hz, 1H), 7.63 (t, J = 8.0 Hz, 1H), 7.57 – 7.47 (m, 2H), 7.40–7.34 (m, 1H). ¹³C NMR (126 MHz, DMSO-d₆) δ 156.25, 156.20, 128.95, 128.51, 127.40, 126.13, 123.28, 122.57, 122.33, 121.85, 120.32, 114.35, 111.78. MS (ESI): calcd for C₁₃H₉F₂O₄P [M - H]^- 297.02; found, 297.01.

Compound 10. Yield 20.3 mg (54%). ¹H NMR (500 MHz, DMSO-d₆) δ 8.27 (d, J = 7.0 Hz, 1H), 7.89 (d, J = 8.3 Hz, 1H), 7.81 (d, J = 8.3 Hz, 1H), 7.63 – 7.57 (m, 1H), 7.48 – 7.39 (m, 2H). ¹³C NMR (126 MHz, DMSO-d₆) δ 156.11, 153.08, 130.10, 129.28, 128.48, 126.56, 123.93 (2C), 123.35, 122.44, 120.03, 112.13, 106.90. MS (ESI): calcd for C₁₃H₈BrF₂O₄P [M - H]^- 374.93; found, 374.92.

General procedure for the synthesis of compound 11–16

Compound 10 (50mg, 0.13 mmol, 1.0 equiv.), boronic acid (0.20 mmol, 1.5 equiv.), K₂CO₃ (71.9 mg, 0.52 mmol, 4.0 equiv.) and Pd(PPh₃)₄ (11.6 mg, 0.01 mmol, 0.1 equiv.) were suspended in 2 mL of solvent (DMF: 1.8 mL, H₂O: 0.2 mL) under argon atmosphere. The reaction mixture was heated at 90°C and the reaction progress was monitored with LC-MS. Upon completion, 1 ml of 1M HCl aqueous solution was added into the reaction mixture. The resulted mixture was purified by prep HPLC to afford the products.

Compound 11. Yield 23.8 mg (49%). ¹H NMR (500 MHz, DMSO-d₆) δ 8.34 (d, J = 8.0 Hz, 1H), 7.91 – 7.86 (m, 2H), 7.76 (d, J = 8.0 Hz, 1H), 7.72 (d, J = 8.2 Hz, 1H), 7.63 (d, J = 8.1 Hz, 1H), 7.61 – 7.52 (m, 3H), 7.49 (m, 1H), 7.39 (ddd, J = 8.2, 7.2, 1.1 Hz, 1H). ¹³C NMR (126 MHz, DMSO-d₆) δ 162.79, 156.28, 153.21, 135.67, 129.33 (3C), 128.87, 128.61, 128.03, 127.56, 126.88, 126.29, 123.38, 123.14, 122.66, 122.36, 120.42, 111.89. MS (ESI): calcd for C₁₉H₁₃F₂O₄P [M - H]^- 373.05; found, 373.03.

Compound 12. Yield 20.8 mg (41%). ¹H NMR (500 MHz, DMSO-d₆) δ 9.78 (brd, 1H), 8.32 (d, J = 8.0 Hz, 1H), 7.78 – 7.65 (m, 4H), 7.63 – 7.49 (m, 2H), 7.42 – 7.33 (m, 1H), 7.01 – 6.92 (m, 2H). ¹³C NMR (126 MHz, DMSO-d₆) δ 162.79, 158.30, 156.22, 153.05, 130.55 (2C), 128.45, 127.68, 127.04, 126.25, 126.18, 126.13, 123.26, 123.11, 122.54, 122.46, 119,65, 116.16, 111.83. MS (ESI): calcd for C₁₉H₁₃F₂O₅P [M - H]^- 389.05; found, 389.04.

Compound 13. Yield 23.3 mg (39%). ¹H NMR (500 MHz, DMSO-d₆) δ 8.33 (d, J = 7.8 Hz, 1H), 8.00 – 7.90 (m, 2H), 7.79 (d, J = 8.0 Hz, 1H), 7.73 (d, J = 7.7 Hz, 1H), 7.70 – 7.62 (m, 2H), 7.54 (t, J = 7.7 Hz, 1H), 7.40 (t, J = 7.6 Hz, 1H). ¹³C NMR (126 MHz, DMSO-d₆) δ 160.62, 158.63, 156.28, 153.37, 133.63, 131.55, 129.84, 128.87, 128.22, 127.84, 127.72, 126.36, 124.88, 123.56, 122.82, 122.45, 122.19, 120.59, 114.15, 111.88. MS (ESI): calcd for C₂₀H₁₁F₆O₄P [M - H]^- 459.03; found, 459.06.

Compound 14. Yield 24.2 mg (45%). ¹H NMR (500 MHz, DMSO-d₆) ¹H NMR (500 MHz, DMSO) δ 8.37 (d, J = 8.0 Hz, 1H), 7.86 (d, J = 8.1 Hz, 1H), 7.82 (d, J = 4.0 Hz, 1H), 7.78 (d, J = 8.2 Hz, 1H), 7.58 – 7.52 (m, 2H), 7.37 (t, J = 7.6 Hz, 1H), 7.29 (d, J = 4.0 Hz, 1H). ¹³C NMR (126 MHz, DMSO-d₆) δ 162.78, 156.15, 151.44, 136.01, 129.97, 128.69, 128.32, 127.21, 126.71, 123.90, 123.62, 123.29, 122.89, 122.33, 119.24, 111.82. MS (ESI): calcd for C₁₇H₁₀ClF₂O₄PS [M - H]^- 412.97; found, 412.98.

Compound 15. Yield 21.3 mg (41%). ¹H NMR (500 MHz, DMSO-d₆) δ 8.33 (d, J = 8.0 Hz, 1H), 7.58 – 7.51 (m, 2H), 7.47 (t, J = 7.7 Hz, 1H), 7.42 (d, J = 7.7 Hz, 1H), 7.38 – 7.27 (m, 6H), 6.01 (s, 1H), 5.70 (s, 1H). ¹³C NMR (126 MHz, DMSO-d₆) δ 156.11, 153.70, 143.80, 139.85, 129.01 (3C), 128.59, 128.56, 128.34, 127.93, 127.23 (3C), 126.20, 123.33, 122.70, 122.28, 120.34, 118.74, 111.79. MS (ESI): calcd for C₂₁H₁₅F₂O₄P [M - H]^- 399.07; found, 399.07.

Compound 16 Yield 25.5 mg (49%). ¹H NMR (500 MHz, DMSO-d₆) δ 8.34 (d, J = 7.1 Hz, 1H), 7.87 (d, J = 7.4 Hz, 1H), 7.80 (d, J = 8.2 Hz, 1H), 7.76 – 7.70 (m, 3H), 7.63 (d, J = 16.6 Hz, 1H), 7.56 (s, 2H), 7.45 – 7.41 (m, 2H), 7.39 – 7.31 (m, 2H). ¹³C NMR (126 MHz, DMSO-d₆) δ 156.14, 153.44, 137.21, 133.20, 129.35 (3C), 128.84, 128.50, 127.35 (3C), 126.39, 124.62, 123.94, 123.41, 122.99, 122.45, 121.84, 111.86. MS (ESI): calcd for C₂₁H₁₅F₂O₄P [M - H]^- 399.07; found, 399.06.

Synthesis of compound 17.

Intermediate 10 (50 mg, 0.13 mmol, 1.0 equiv.), ethynylbenzene (16.3 mg, 0.16 mmol, 1.2 equiv.), CuI (1.9 mg, 0.01 mmol, 0.1 equiv.), Pd(PPh₃)₄ (11.6 mg, 0.01 mmol, 0.1 equiv.) and triethylamine (109 μL, 0.78 mmol, 6.0 equiv.) were suspended in 2 mL of DMF under argon atmosphere. The reaction mixture was heated at 90°C and the reaction progress was monitored with LC-MS. Upon completion, 1 ml of 1M HCl aqueous solution was added into the reaction mixture. The resulted mixture was purified by prep HPLC to afford the products. Yield 29.0 mg (56%). ¹H NMR (500 MHz, DMSO-d₆) δ 8.35 (d, J = 8.0 Hz, 1H), 7.85 – 7.77 (m, 2H), 7.70 – 7.64 (m, 2H), 7.59 (d, J = 7.1 Hz, 1H), 7.55 (d, J = 7.9 Hz, 1H), 7.51 – 7.44 (m, 3H), 7.40 (t, J = 7.6 Hz, 1H).¹³C NMR (126 MHz, DMSO-d₆). δ 162.79, 156.09, 155.75, 133.06, 132.05 (3C), 129.79, 129.38 (3C), 128.55, 127.25, 123.36, 122.78, 122.32, 111.61, 108.02, 95.21, 84.02. MS (ESI): calcd for C₂₁H₁₃F₂O₄P [M - H]^- 397.05; found, 397.04.

Synthesis of compound 18.

4-bromodibenzo[b,d]furan (50 mg, 0.20 mmol, 1.0 equiv.), (1-phenylvinyl)boronic acid (44.4 mg, 0.30 mmol, 1.5 equiv.), K₂CO₃ (41.5 mg, 0.30 mmol, 1.5 equiv.) and Pd(PPh₃)₄ (23.2 mg, 0.02 mmol, 0.1 equiv.) were suspended in 2 mL of solvent (DMF: 1.8 mL, H₂O: 0.2 mL) under argon atmosphere. The reaction mixture was heated at 90°C and the reaction progress was monitored with LC-MS. Upon completion, 1 ml of 1M HCl aqueous solution was added into the reaction mixture. The resulted mixture was purified by prep HPLC to afford the products. Yield 38.4 mg (71%). ¹H NMR (500 MHz, DMSO-d₆) δ 7.57 (d, J = 8.3 Hz, 1H), 7.49 – 7.44 (m, 1H), 7.42 – 7.34 (m, 4H), 7.32 (d, J = 11.3 Hz, 5H), 7.22 (dt, J = 6.3, 2.2 Hz, 1H), 5.92 (s, 1H), 5.73 (s, 1H). ¹³C NMR (126 MHz, DMSO-d₆) δ 155.79, 153.49, 144.35, 140.33, 133.80, 129.46, 129.27, 128.94, 128.63, 128.46, 128.14, 127.44, 125.86, 124.50, 123.93, 123.68, 121.67, 121.20, 118.17, 112.17. MS (ESI): calcd for C₂₀H₁₄O [M + H]⁺ 271.10; found, 271.11.

Recombinant Protein Production

The coding sequences of human CDC14A isoform 2 (codons 1–413) and CDC14B isoform 3 (codons 54–448) were codon-optimized for E. coli, synthesized, and cloned into an entry vector for Gateway cloning (Invitrogen) by Twist Biosciences. The coding sequences were sub-cloned into pDEST15 for expression as N-terminal glutathione-S-transferase (GST) fusion proteins using Gateway cloning. Mutations were made in pDEST15 vectors by inverse PCR using the In-Fusion cloning system (Takara Bio) as recommended by the supplier and confirmed by whole plasmid sequencing. GST-Cdc14 enzymes were expressed in 1 L 2xYT cultures (10 g/L yeast extract, 16 g/L tryptone, 5 g/L NaCl) of BL21-AI cells by induction with 0.1% L-arabinose for 18 hours at 25°C. Cells were lysed with 1 mg/mL lysozyme for 30 min on ice in 30 mL 25 mM HEPES pH 7.5, 500 mM NaCl, 2 mM EDTA, 0.1% Triton X-100, 10% glycerol, 1 mM PMSF, 10 μM leupeptin, 1 μM pepstatin, 5 μM bestatin, 1 mM benzamidine and 1,000 units Universal Nuclease (Thermo Fisher Scientific). Extracts were clarified by centrifugation at 35,000xg for 30 min at 4°C and the soluble fraction was loaded onto 0.5 mL bed volume of Glutathione Superflow Agarose (Thermo Fisher Scientific) equilibrated with 10 mL 25 mM HEPES pH 7.5, 500 mM NaCl, and 10% glycerol. The column was washed with 10 mL 25 mM HEPES pH 7.5, 500 mM NaCl, and 10% glycerol. Bound protein was subsequently eluted with 25 mM HEPES pH 7.5, 500 mM NaCl, 10% glycerol, and 20 mM reduced glutathione. Peak fractions determined by Bradford assay, were pooled, and dialyzed overnight into 25 mM HEPES pH 7.5, 300 mM NaCl, 2 mM EDTA, 0.1% 2-mercaptoethanol, and 40% glycerol prior to storage at −80°C in small aliquots.

The proteins for the selectivity panel (TC-PTP, Aa 1–387; PTP1B, Aa 1–321; SHP-1, Aa 245–543; SHP-2, Aa 224–528; LYP, Aa 1–294; STEP, Aa 258–539; HePTP, Aa 22–360; PTP-PEST, Aa 5–304; FAP-1, Aa 2124–2485; PTPα, Aa 173–793; PTPε, Aa 107–697; CD45, Aa 620–1236; mouse MKP5, Aa 320–647; VHZ, Aa 1–150; Larforin, Aa 1–331; LMPTP, Aa 1–158.) were cloned into pET-21a(+) vector. Bacterial BL21(DE3) (Novagen) was used as an expression host, and the induction of protein expression was carried out in LB media with 1 mM IPTG at 18 °C overnight. Cell pellets were stored at −80 °C for subsequent protein purification. Protein purification was conducted at 4 °C. Frozen cell pellets were lysed by sonication in 40 ml cold lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 5 mM imidazole, and 1 mM PMSF) per liter cell pellet. Cell lysates were clarified by centrifugation using a Bechman JA-18 rotor for 15 min at 6,000 rpm. The supernatant was incubated with HisPur Ni-NTA resin (Thermo Scientific) for 2 h, and then packed onto a column and washed with 50 resin volume of buffer A (50 mM Tris-HCl, pH 8.0, 500 mM NaCl, 5 mM imidazole). The HIS-tagged proteins were eluted with Buffer B (50 mM Tris-HCl, pH=8.0, 500 mM NaCl, 300 mM imidazole,). Pooled HIS-protein-containing fractions were concentrated, loaded onto a HiLoad 26/600 Superdex 75 column (GE Healthcare Biosciences), and eluted with storage buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1 mM DTT, 10% glycerol). Proteins used for inhibition assays were purified using Ni-NTA resin (Qiagen) followed by size exclusion column chromatography (ÄKTA pure, cytiva) and the purity was determined to be >95% by SDS-PAGE and Coomassie staining. The protein was aliquoted and stored at −80 °C.

Dynamic Light Scattering (DLS) Assay

In a 384-well, low-volume, non-treated, black with clear bottom assay plate (Corning 3540), compounds were dissolved in DMG buffer (50 mM DMG, pH 7.0, 1 mM EDTA, 18 mM NaCl, 0.01% Triton X-100) to reach a final volume of 30 μL and a concentration of 1 mM (15 μM for ketoconazole). The DMSO concentration was maintained at 1% (v/v) for all solutions. The bottom of the plate was wiped off using Kimwipe to eliminate interferences from dust and other particles. The plate was centrifuged using a ThermoFisher Scientific Heraeus Multifuge X3 centrifuge and incubated in the instrument for 30 min to ensure temperature equilibrium. The data was collected using DyanaPro Plate Reader II with the following parameter:

Number of times to measure each well: 1, Number of times to scan the selected regions: 1, Enable Auto-attenuation: Yes, Acquire an image of each well: Yes, Number of DLS acquisitions: 10, Temperature: 25°C, DLS Acquisition Time: 5 s.

The data cutoffs were set between 10 μs and 1 s with a filter to mask acquisitions with an autocorrelation intensity outside of the baseline limit of 1 ± 0.02. The data was analyzed and plotted using GraphPad Prism 9.2.0

IC₅₀ and K_i Determination

PTP activity was assayed using 6,8-Difluoro-4-Methylumbelliferyl Phosphate (DiFMUP) as a substrate in 3,3-Dimethylglutaric acid (DMG) buffer (50 mM DMG, pH 7.0, 1 mM EDTA, 18 mM NaCl, 0.01% Triton X-100) at 25 °C. To determine the IC₅₀ values, the assays were performed in 96-well plates (Corning Costar 3915). The reaction was initiated by the addition of enzyme (the final concentration was 2.5 nM) to a reaction mixture containing DiFMUP (final concentration close to Km values of each enzyme) and inhibitors (serially diluted by 2-fold/step for a total of 11 concentrations). The final reaction volume is 200 μL. The reaction was allowed to proceed for 10 minute and then quenched by the addition of 40 μL of a 160 μM solution of bpV(Phen). The fluorescence signal was measured using a CLARIOstar Plus Microplate Spectrophotometer (BMG Labtech) using excitation and emission wavelength of 340 nm and 450 nm, respectively. Data were fitted using Prism GraphPad 9.2.0 with the build in 4-parameter IC₅₀ equation.

To determine the mode of inhibition, the reactions were initiated by the addition of enzyme (2.5 nM final concenrtation) to the reaction mixtures (0.2 mL) containing various concentrations of DiFMUP (serially diluted by 2-fold/step for a total of 8 concentrations) with inhibitor at the specified concentration.. The reaction was allowed to proceed for 10 minute and then quenched by the addition of 40 μL of a 160 μM solution of bpV(Phen). The fluorescence signal was measured using a CLARIOstar Plus Microplate Spectrophotometer (BMG Labtech) using excitation and emission wavelength of 340 nm and 450 nm, respectively. Data were fitted using Prism GraphPad 9.2.0 with the build in competitive inhibition equation.

Molecular Docking Studies

Molecular docking studies were performed on a previously reported hCDC14B crystal structure (PDB 1OHC²³) using the Maestro suite version 2021–3 (Schrödinger, LLC, New York, NY). The protein was prepared using the Protein Preparation Workflow at pH 7.4 ± 1.0. Missing side chains and loops were filled using Prime. Hydrogen bond assignments were optimized using PROPKA, and the catalytic C314 was manually ionized via the 3D Builder module. Lastly, a restrained minimization was performed to 0.30Å RMSD using the OPLS4 force field, and all waters further than 5 Å from heteroatoms were removed. The receptor grid was generated using the Receptor Grid Generation module. The prepared 1OHC structure was aligned to PDB 1OHE²³, and the coordinates of the 1OHE peptide ligand was used to define the grid site in 1OHC (using default settings). Ligand preparation was conducted using the LigPrep module with the OPSL4 force field. Ionization states were generated using Epik at pH 7.0 ± 1.0 Docking was performed using Glide in the standard precision (SP) mode using default settings. The top 10 poses were retained for visual evaluation.

Docking studies for compound 9 with other PTPs was conducted in a similar manner. Docking studies were performed on PTP1B (PDB 1PXH⁸⁸), MKP-5 (1ZZW⁸⁹), VHZ (2IMG⁹⁰), CD45 (1YGU⁹¹), and Laforin (4RKK⁹²). Prior to preparation, organic solvents and ions present from the crystallography experiments were removed. For 4RKK, the maltohexose chains and phosphates were removed from the crystal structure, and S266 was mutated to cysteine to mimic the enzymatically active protein. All proteins were subsequently prepared using the above protocol, and the catalytic cysteines for each PTP were manually ionized using the 3D Builder similar as above. With the exception of MKP-5 (1ZZW) and Laforin (4RKK), the receptor grid was generated under default settings using the co-crystallized ligand to define the grid. For MKP-5 (1ZZW), the grid was centered on residues A410 and N379. For Laforin (4RKK), the grid was centered on the XYZ coordinates 52.47, 71.22, and 122.78, and a hydrogen bond constraint was placed on the R272 side chain, which is known to interact with substrate phosphate groups. compound 9 was docked into each PTP using the standard precision (SP) setting of Glide, and up to 10 top poses were retained for visual evaluation.

Molecular Dynamics

Molecular dynamics were carried out using Desmond (Desmond Molecular Dynamics System, D. E. Shaw Research, New York, NY, 2021. Maestro-Desmond Interoperability Tools, Schrödinger, New York, NY, 2021.). The docking model of compound 15 bound to hCDC14B was set up for MD studies with the System Builder. The TIP3P solvent model was utilized, and an orthorhombic box shape was used when constructing the boundary conditions. The box size was calculated using buffer, the box volume was minimized, and the OPSL4 force field was utilized. Counterions were added to neutralize the system and were excluded from the region within 20Å of the ligand. The salt concentration used was 0.15M. A 50ns simulation time was used, with a trajectory of 100ps and approximately 500 frames. The simulation was run at 300K, and the interactions analysis was selected prior to the run.

Cell Permeability Studies

The cell permeability of tested compounds was determined with Corning^® BioCoat^® Pre-coated PAMPA Plate System as per the manufacturer’s guidance. In specific, the compound solutions were prepared by diluting 10 mM DMSO stock solutions in PBS (in most cases we used a final concentration of 200 μM). The compound solutions were added to the wells (300 μL/well) of the receiver plate and PBS was added to the wells (200 μL/well) of the pre-coated filter plate. The filter plate was then coupled with the receiver plate, and the plate assembly was incubated at room temperature without agitation for five hours. At the end of the incubation, the plates were separated, and 150 μL solution from each well of both the filter plate and the receiver plate was sampled. The final concentrations of compounds in both donor wells and acceptor wells were analyzed by LC-MS analysis. Permeability of the compounds was calculated using the following formula:

Permeability (cm/s):

$${\text{P}}_{\mathrm{e}}=\frac{-\text{ln}\left(1-\frac{{\text{C}}_{\mathrm{A}}\left(\text{t}\right)}{{\text{C}}_{\text{eq}}}\right)}{\text{A}\times \left(\frac{1}{{\text{V}}_{\mathrm{D}}}+\frac{1}{{\text{V}}_{\mathrm{A}}}\right)\times \text{t}}$$

$${\text{C}}_{\text{e}\mathrm{q}}=\frac{{\text{C}}_{\mathrm{D}}\left(\text{t}\right)\times {\text{V}}_{\mathrm{D}}+{\text{C}}_{\mathrm{A}}\left(\text{t}\right)\times {\text{V}}_{\mathrm{A}}}{{\text{V}}_{\mathrm{D}}+{\text{V}}_{\mathrm{A}}}$$

whereas: A = filter area (0.3 cm²), ${\text{V}}_{\mathrm{D}}$ = donor well volume (0.3 mL), ${\text{V}}_{\mathrm{A}}$ = acceptor well volume (0.2 mL), t = incubation time (seconds), ${\text{C}}_{\mathrm{A}}\left(\text{t}\right)$ = compound concentration in acceptor well at time t, ${\text{C}}_{\mathbf{D}}\left(\text{t}\right)$ = compound concentration in donor well at time t.

Differential Scanning Fluorimetry (DSF) Assay

DSF experiments were performed in a Roche LC480 Light Cycler II qPCR machine. Solutions of protein (assay buffer: 50 mM DMG, pH 7.0, 1 mM EDTA, 18 mM NaCl), Sypro Orange (Thermo Fisher Scientific, S6650) and tested compounds or DMSO were added to the wells of a 96-well PCR plate (USA Scientific, 1402–9590). The final concentration for protein, compound, and Sypro Orange are 10 μM, 100 μM, and 15x, respectively. The plates were sealed with transparent films (EXCEL Scientific, TS-RT2–100) and heated in the qPCR machine from 25 to 95 °C in increments of 1.0 °C. The 96 well-plate assay data was analyzed using Roche LC480 Light Cycler II software.

Cell Culture

The U2OS and HEK293 cells were grown in DMEM cell culture media supplemented with 10% fetal bovine serum (no fetal bovine serum supplement for cytotoxicity assay), penicillin (50 units/mL), and streptomycin (50 μg/mL) in a 37°C incubator containing 5% CO₂. The Jurkat cells were grown in RPMI 1640 cell culture media supplemented with 10% fetal bovine serum (no fetal bovine serum supplement for cytotoxicity assay), penicillin (50 units/mL), and streptomycin (50 μg/mL) in a 37°C incubator containing 5% CO₂. 5 μM Mitomycin C was used as needed to induce DNA damage.

Immunoblotting

Cultured cells were lysed with ice cold lysis buffer (50 mM Tris (pH 8.0), 150 mM NaCl, 10% Glycerol, 1% Triton-X-100) supplied with phosphatase inhibitor and protease inhibitor mixture. Equal amounts of protein were resolved by SDS-PAGE, transferred to nitrocellulose membrane and subjected to immunoblotting.

Cytotoxicity Evaluation

For cytotoxicity test, 5,000 cells were plated onto 96-well plates in 100 μL medium and treated with compound 15 concentration varying from 0 to 50 μM. After 24 hours of incubation, the compound was washed off followed by addition of 100 μL fresh media and 10 μL cell counting kit 8 (CCK-8) reagent. OD470 was measured after 3 hours of incubation and used to calculate cytotoxicity.

PK Study

All the in vivo studies were performed under an animal protocol (1511001324) approved by the Institutional Animal Care & Use Committee of the Purdue University, in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. For PK study, C57BL6 female mice (25–30g body weight) were injected intraperitoneally with 25 or 50 mg/kg DU-14 dissolved in 0.4 ml saline. Blood samples were collected through tail vein at indicated time points after injection. Isoflurane was used as an anesthetic. All blood samples were centrifuged at 1,500 g for 5 minutes, and plasma was separated and stored at –80°C until analysis by a validated method based on reversed-phase liquid chromatography coupled to mass-spectrometric detection (LC/MS) using a previously published procedure⁹³.

Abbreviations

DiFMUP

6,8-Difluoro-4-Methylumbelliferyl Phosphate

DLS

dynamic light scattering

F₂Pmp

phosphonodifluoromethyl phenylalanine

hCDC14A

human Cell Division Cycle 14A

hCDC14B

human Cell Division Cycle 14B

IP

intraperitoneal

MD

molecular dynamic

PAMPA

parallel artificial membrane permeability assay

PhFP

p-(difluorophenylmethyl) phosphonic acid

PK

pharmacokinetics

PO

oral administration

PTKs

protein tyrosine kinases

PTPs

protein tyrosine phosphatases

pSer

phosphoserine

pTyr

phosphotyrosine

RMSD

root mean square deviation

RMSF

root mean square fluctuation

SD

standard deviation

tPSA

topological polar surface area