Enzyme Responsive Rigid-Rod Aromatics Target “Undruggable” Phosphatases to Kill Cancer Cells in Mimetic Bone Microenvironment

Abstract

Bone metastasis remains a challenge in cancer treatment. Here we show that enzymatic responsive rigid-rod aromatics, acting as the substrates of “undruggable” phosphatases, to kill cancer cells in mimetic bone microenvironment. By phosphorylation and conjugating nitrobenzoxadiazole (NBD) to hydroxybiphenylcarboxylate (BP), we obtained pBP-NBD (1P) as a substrate of both acid and alkaline phosphatases. 1P effectively kill both metastatic castration-resistant prostate cancer cells (mCRPCs) and osteoblast mimic cells in their co-culture. 1P enter Saos2 almost instantly to target the endoplasmic reticulum (ER) of the cells. Co-culturing with Saos2 cells boosts the cellular uptake of 1P by mCRPCs. Cryo-EM reveals the nanotube structures of both 1P (2.4 Å resolution, pH 5.6) and 1 (2.2 Å resolution, pH 7.4). The helical packing of both nanotubes is identical, held together by strong pi-stackings interactions. Besides reporting the atomistic structure of nanotubes formed by the assembly of rigid-rod aromatics, this work expands the pool of molecules for designing EISA substrates that selectively target TME.

Graphical Abstract

This communication reports the first example of enzyme-instructed self-assembly (EISA) of rigid-rod aromatics for killing cancer cells in mimetic bone microenvironment. Despite considerable progress in cancer therapy, tumor metastasis still causes most of cancer-related death.¹ For example, approximately 90% of patients who die of prostate cancer have bone metastases.² Bone metastases are a vicious self-reinforcing cycle in tumor microenvironment (TME), where cancer cells promote the differentiation of osteoclasts and osteoblasts, and the increased bone turnover releases cytokines to benefit metastatic cancer cells. Moreover, bone is also a site for further spread of other metastasis.³ The current treatment of bone metastasis, Radium-223 (Ra-223)^4,5 is only palliative because the acceptable radiation dosage of Radium-223 is limited. Thus, there is an urgent need to develop novel approaches for killing cancer cells in bone metastasis sites or TME with minimal side effects. Inspired by the dual targeting mode-of-action of Radium-223, we are aiming to develop non-radioactive small molecules as a single agent that self-assemble in-situ or on-site to kill both metastatic castration-resistant prostate cancer (mCRPC) and osteoblast cells in TME.

A prominent feature of osteoblastic mCRPC at the TME is that mCRPC and osteoblasts overexpress prostatic acid phosphatase (PAP)⁶ and alkaline phosphatase (ALP)⁷, respectively. However, developing inhibitors of PAP and ALP to treat mCRPC has been unsuccessful for several reasons: (i) PAP acts as tumor suppressor,⁸ (ii) inhibitor of ALP is unable to inhibit cancer cells,⁹ and (iii) most phosphatases previously are considered as undruggable targets.^10–11 Thus, we decide to explore EISA, which kills cancer cells by enzymatic reaction and self-assembly,¹² for developing therapeutics for osteoblastic mCRPC. EISA is particularly attractive because EISA kills cancer cells without inhibiting the targeted enzymes. Moreover, ALP is metabolically inert in serum,¹³ a subtlety allowing EISA to occur at TME. Recently, a variety of small molecule substrates, including peptides,^14–18 carbohydrates,¹⁹ and lipids,^20–21 of ALP-based EISA are able to induce death of the cancer cells that overexpressing ALP because EISA allows “on-site” generating nanostructures from small molecules.²² Although peptide-based substrates are the most explored among these building blocks up-to-date, their efficacy for killing mCRPC still remains to be improved.²³ Thus, we choose to examine other nonpeptidic molecules as the self-assembling building blocks for EISA against mCRPC.

Based on aggregate advisors²⁴ and rigid-rod molecules for self-assembly, we choose a rigid-rod aromatic molecule, biphenyl, as the core motif for developing EISA substrates of PAP and ALP. Biphenyl has found various applications in liquid crystals,^25–28 dendrimers,²⁹ long-range self-assembly,³⁰ amphiphiles,³¹ peptides.³² We recently found that biphenyl enable rapid enzymatic self-assembly and hydrogelation of peptides.³³ But biphenyl has yet to be explored for EISA without involving peptides. Based on the above rationale, we phosphorylated the hydroxyl and conjugated a fluorophore (nitrobenzoxadiazole (NBD)) at the carboxylic end of hydroxybiphenylcarboxylate (BP) or hydroxyterphenylcarboxylate (TP) to produce pBP-NBD (1P) or pTP-NBD (2P), respectively. As a substrate of both PAP and ALP, 1P effectively kills both metastatic castration-resistant prostate cancer cells (mCRPCs) (VCaP or PC3) and osteoblast mimic cells (Saos2) in their co-culture. Fluorescent imaging shows that 1P enters Saos2 almost instantly to target the endoplasmic reticulum (ER) of the cells and that Saos2 in the co-culture boosts the cellular uptake of 1P or 2P by VCaP or PC3 cells. Cryo-EM reveals that the 1 (from dephosphorylating 1P at physiological pH) or 1P (at acidic pH) self-assembles to form cones with C7 symmetry, and the supramolecular cones stack helically with a 3.5 Å rise to form nanotubes. Besides reporting the atomistic structure of nanotubes formed by the assembly of rigid-rod aromatics, this work, using enzymatic reactions to form intracellular nanotubes of small rigid-rod molecules, expands the pool of molecules for designing EISA substrates that selectively target TME.

Scheme 1 shows the structures and the key synthetic step for making, 1P, 2P, and 3P, which have biphenyl, terphenyl, and phenyl motifs, respectively. While BP and hydroxylphenylcarboxylate are commercially available, TP is produced in almost quantitative yield from 4-bromo-4’-hydroxylbiphenyl and 4-carboxyphenylboronic acid by palladium-catalyzed cross coupling.³⁴ Mixing BP or hydroxylphenylcarboxylate with phosphorus pentachloride directly under heat or reacting TP with phosphorous oxychloride in pyridine produces the phosphorylated aromatic carboxylic acids. After getting all three phosphorylated compounds, we conjugated them with NBD-ethylene diamine to generate 1P, 2P, and 3P.

Scheme 1.

EISA for simultaneously killing mCRPC and osteoblast cells and the relevant synthesis.

As shown by transmission electron microscopy (TEM), adding ALP to solution of 1P (500 μM and pH7.4) transforms the amorphous aggregates of 1P to ~7 nm diameter nanotubes of 1 (Figure 1). At pH 5.6, 1P (50 μM) forms nanotubes like those of 1. These results indicate 1P self-assembling more readily at lower pH because protonation decreases phosphate repulsion. Adding PAP in 1P solution (50 μM and at pH 5.6) results in more nanotubes (Figure 1). Adding ALP to the solution of 2P also switches the aggregates of 2P into nanotubes with similar diameters (Figure S2). At pH 5.6, 2P (50 μM) transforms from aggregates into twisted nanoribbons (Figure S1). 3P exists as amorphous aggregates both at pH 7.4 or pH 5.6 and rarely forms nanotubes after adding ALP or PAP (Figure S1), indicating that noncovalent interactions from phenyl are insufficient for 3P or 3 to self-assemble into nanotubes.

Figure 1.

TEM of 1P (500 μM, pH 7.4) before and after adding ALP (0.1 U/mL) for 24 h and of 1P (50 μM, pH 5.6) before and after adding PAP (0.1 U/mL) for 24 h. Scale bar: 100 nm.

We tested 1P for its activity against six different cell lines, Saos2, SJSA1, VCaP, PC3, HepG2 and PNT1A (Figures 2A and S4). 1P shows the first day IC₅₀ around 20.9 μM against Saos2 cells. The IC₅₀ values of 1P against SJSA1 and Saos2 are comparable, except day 1. While 1P significantly inhibit VCaP and PC3 cells from day 2 and day 3, respectively, 1P is rather compatible with HepG2 and PNT1A even at day 3, with IC₅₀ around 108 μM and 89 μM. Even though 2P showed potent activity against Saos2 and SJSA1, it hardly inhibits VCaP, PC3, HepG2 and PNT1A (Figures S3 and S4). Compared to 1P and 2P, 1 and 2 is more toxic to the normal prostate cells, PNT1A, indicating that phosphorylation enhances the selectivity for targeting cancer cells. ALP inhibitor (DQB) reduces the cytotoxicity of 1P and 2P against Saos2 or SJSA1 cells (Figure S5), indicating that ALP-catalyzed EISA of 1P and 2P contributes to their cytotoxicity. The low cytotoxicity of 1P and 2P towards hepatocyte (HepG2) indicates low toxicity of 1P and 2P to liver. 3P hardly inhibits these cells (Figure S2B), consistent with the poor self-assembling ability of 3. M-β-CD, an inhibitor for caveolae mediated endocytosis, rescues Saso2 and SJSA1 treated by 1P or 2P, suggesting that 1P or 2P enters the cells via endocytosis (Figure S6).

Figure 2.

(A) IC₅₀ of 1P incubated with Saos2, SJSA1, VCaP, PC3, and HepG2 cells for 24 h. (B) Instant cell entry of 1P (100 μM). (C) Intracellular distribution of 1P (50 μM) in Saos2, SJSA1, VCaP and PC3 after 4 h incubation. Red arrows indicate the marker, green arrow indicates 1. Yellow indicates their colocalization.

As shown in Figure 2B, 1P enters Saos2 or SJSA1 within 1 minute and accumulates inside the cells. On the other hand, 1P accumulates in VCaP or PC3 cells slowly that visible fluorescence from 1 emerges in VCaP and PC3 cells after 30 and 60 minutes, respectively. The inhibitory activity of 1P correlates well with the rate of cellular uptake, except in the case of HepG2. Although 1P enters HepG2 faster (Figure S7) than entering PC3, 1P hardly inhibits HepG2, agreeing with detoxification function of hepatocytes.³⁵ 2P exhibits a similar trend as that of 1P to enter Saos2, SJSA1, VCaP, and PC3 cells (Figure S9). Agreeing with its cell compatibility, 3P enters Saos2 much slower than 1P or 2P does (Figure S8).

For Saos2 and SJSA1 cells, the distribution of 1P in cytosol resembles to that of ER (Figures 2C and S10). Co-staining 1P with ER-tracker, Lyso-tracker, or Mito-tracker in Saos2 or SJSA1 cells shows that the fluorescence from 1P mostly overlaps with the fluorescence of ER-tracker, indicating that 1P, after entering cells and being dephosphorylated, specifically localizes in ER of Saos2 and SJSA1. The ER accumulation agrees with that phosphobipenyl is a substrate of PTP1B³⁶, which mainly locates at ER³⁷. 2P localizes more in ER of Saos2 and SJSA1 cells than in lysosomes and mitochondria, and forms denser fluorescent puncta in Saos2 and SJSA1 than 1P does (Figures S11 and S12). The fluorescence of 1P in VCaP and PC3 cells overlap more with that of Lysotracker, indicating that 1P mainly undergoes dephosphorylation in the lysosomes. This result agrees with that PAP localized in lysosome.³⁸

We cocultured the mCRPC cells (e.g., VCaP or PC3) and the osteoblast mimic cells (Saos2) in a 1:1 ratio to create a mimetic bone microenvironment and tested the efficacy of 1P. As shown in Figure 3A, being incubated with 1P for 24 h, the viability of Saos2 or VCaP cells is 1.6% and 41.3%, respectively. The cell viability of the coculture of Saos2 and VCaP is 1.9% after adding 1P in the coculture. Similarly, 2P inhibits VCaP more effectively in the coculture. Moreover, Figure 3B shows the viability of PC3 cells, being incubated with 1P or 2P in the co-culture of PC3 and Saos2, is significantly lower than that of PC3 cells alone. These results suggest that Saos2 in the coculture significantly enhances the inhibitory activity of 1P or 2P against VCaP or PC3 cells. Fluorescence imaging reveals the cellular uptake of 1P by VCaP or PC3 cells (Figures 3C, D, S13–15). The fluorescence in the VCaP or PC3 cells increases faster in the coculture than being cultured alone. These results indicate that Saos2 cells, in the co-culture, boost the cellular uptake of 1P or 2P by VCaP and PC3 cells, thus leading to more effective inhibition of VCaP or PC3 cells.

Figure 3.

Viabilities of cells treated by 1P (100 μM) or 2P (50 μM) in (A) Saos2, VCaP, and the coculture of Saos2 and VCaP and in (B) Saos2, PC3, and the coculture of Saos2 and PC3. The mean fluorescence intensities of 1P (50 μM) in (C) VCaP and VCaP and the VCaP cocultured with Saos2-DsRed and (D) PC3-DsRed cells and the PC3-DsRed cocultured with Saos2 (each line representing one cell).

We used cryo-EM and determined the high-resolution nanotube structures of both 1 at pH 7.4 and 1P at pH 5.6 (Figures S16A and S17A). Possible helical symmetries were calculated from the averaged power spectrum of boxed filaments (Figures S16B and S17B) and the correct ones were found³⁹. The nanotubes of 1 and 1P have almost identical diameters (Figure 4A–B) and helical symmetries (Figure 4C–D, Table S1), with aromatic rings held together by extensive pi-stacking interactions (Figure 4E). The NBD motifs points to the center and the biphenyl groups constitute the periphery of the nanotubes. The ability of 1P to form nanotubes at pH 5.6 is consistent with the lysosomal accumulation of 1P (Figure 2A, C) and its increased activities against VCaP and PC3 cells after 48 h.

Figure 4.

A cyro-EM images of the filaments of (A) 1P and (B) 1. Scale bar 50 nm. (C) 3D reconstruction of tubes of 1P and 1 from cryo-EM. (D) Top view of tubes of 1P and 1, phosphate of 1P can be seen at the periphery. (E) Side view of tubes of 1.

In summary, this work reports a new class of EISA substrates to form nanotubes for effectively killing osteoblast model cells and mCRPCs in mimetic bone TME. The toxicity of 1 and 2 to PNT1A imply that nanotubes of 1 or 2 likely results in cell death. It illustrates the dual targeting mode-of-action of EISA substrates, which promising to kill both cancer and osteoblast cells at bone TME. Although the detailed mechanism of action of EISA of 1P or 2P in the co-culture remains to be elucidated, using the enzymatic feature of TME to boost cellular uptake of EISA substrates in mCRPCs promising a new way to target TME of mCRPCs. The approach reported here should be applicable to other aggregation-prone small molecules or drug candidates,^{24, 40} which may lead to more effective EISA substrates for targeting TME of other metastatic cancers.