Enzymatic Assemblies of Thiophosphopeptides Instantly Target Golgi Apparatus and Selectively Kill Cancer Cells

Abstract

Golgi apparatus is emerging as a key signaling hub of cells and an important target for cancer therapy, but there are few approaches for targeting Golgi and selectively killing cancer cells. Here we show an unexpected result that changing an oxygen atom of the phosphoester bond in phosphopeptides by a sulfur atom enables instantly targeting Golgi apparatus (GA) and selectively killing cancer cells by enzymatic self-assembly. Specifically, conjugating cysteamine S-phosphate to the C-terminal of a self-assembling peptide generates a thiophosphopeptide. Being a substrate of alkaline phosphatase (ALP), the thiophosphopeptide undergoes rapid ALP-catalyzed dephosphorylation to form a thiopeptide that self-assembles. The thiophosphopeptide enters cells via caveolin-mediated endocytosis and macropinocytosis and instantly accumulates in GA because of dephosphorylation and formation of disulfide bonds in Golgi by themselves and with Golgi proteins. Moreover, the thiophosphopeptide potently and selectively inhibits cancer cells (e.g., HeLa) with the IC₅₀ (about 3 μM), which is an order of magnitude more potent than that of the parent phosphopeptide. This work, as the first report of thiophosphopeptide for targeting Golgi, illustrates a new molecular platform for designing enzyme and redox responsive molecules that target subcellular compartment for functions.

Graphical Abstract

The rapid dephosphorylation of a thiophosphopeptide via the enzymatic reaction generates a self-assembling thiopeptide that can target and accumulate at the Golgi apparatus for imaging Golgi and inhibiting cancer cells. This result illustrates a novel molecular platform for modulating functions resulted from enzymatic noncovalent synthesis.

Golgi apparatus (GA), a stack of flattened membrane-enclosed disks that are dynamically regulated during cell cycles in mammalian cells, is considered as the “heart” of intracellular transportation.^[1] Increasing numbers of studies have revealed that Golgi is a hub for different signaling pathways that drive the survival and migration of cancer cells.^[2] Although Golgi is emerging as an important target for cancer therapy, there are, however, few approaches for targeting Golgi.^[3] While Golgi mannosidase II inhibitors are able to inhibit cancer cells, the selectivity^[4] or efficacy^[3a] of the inhibitors remains to be improved. In addition, several studies reported the imaging of Golgi, including the commercial dyes for staining Golgi,^[5] a smart “off–on” fluorescence probe for imaging the Golgi in cancer cells,^[6] and carbon quantum dots localizing at Golgi.^[7] These imaging agents, however, require 30 minutes^[6] or longer incubation time^{[6, 8]} or pretreatment,^[5] and they have yet to lead to the approach for selectively inhibiting the cancer cells. Thus, there is an unmet need of targeting Golgi to inhibit cancer cells.

During our study of enzymatic noncovalent synthesis (ENS),^[9] we changed an oxygen atom of the phosphoester bond in a phosphopeptide (pO1) by a sulfur atom to make pS1 for fast enzymatic self-assembly. Our studies show that pS1 undergoes rapid dephosphorylation catalyzed by ALP to form S1 that assembles. Unexpectedly, treating HeLa cells with pS1 shows that S1 instantly accumulates at Golgi of the HeLa cells at the concentration as low as 500 nM. Such an enzymatic accumulation of Golgi (Scheme 1) is proportional to both the concentration of pS1 and the time of incubation. Similar rapid enzymatic accumulation also takes places in the Golgi of several other cells (e.g., Saos-2, SJSA-1, OVSAHO, HCC1937, and HEK293). Unlike pS1, the parent phosphopeptide, pO1, taking longer time for dephosphorylation than that of pS1, requires hours for cellular uptake and largely remains in endosomes. These results indicate that rapid dephosphorylation of the thiophosphate group and the resulting thiol group are critical for instantly targeting Golgi. Based on this insight, we designed pS2, a nonfluorescent analogue of pS1. Being able to undergo rapid dephosphorylation catalyzed by ALP to form S2 that exhibit a critical micelle concentration (CMC) of 9.5 μM, pS2 inhibits HeLa cells with an IC₅₀ value about 3 μM, an order of magnitude more potent than that of the parent phosphopeptide. Preliminary mechanistic studies indicate that (i) the thiophosphopeptides enter cells via both caveolin-mediated endocytosis and macropinocytosis, (ii) disulfide bond formation is essential for Golgi targeting, and (iii) the level of ALP of cells contributes to the rate of the accumulation of the resulting thiopeptide assemblies at the Golgi. Providing the first case of targeting Golgi based on fast enzymatic kinetics and redox reaction at the oxidative environment of Golgi, this work illustrates a new molecular platform for designing enzyme and redox responsive molecules that target subcellular compartment for functions.

Scheme 1.

Illustration of thiophosphopeptides instantly targeting the Golgi apparatus by enzymatic assembling and forming disulfide bonds.

As shown in Scheme 1, pS1 consists of three segments: (i) 4-nitro-2,1,3-benzoxadiazole (NBD), a fluorophore that emits bright green fluorescence in hydrophobic environment of supramolecular assemblies;^[10] (ii) D-diphenylalanine (ff), a hydrophobic building block, which enables self-assembly^[11] and resists proteolysis; (iii) thiophosphate group, a substrate of ALP for enzymatic self-assembly.^[12] Solid phase peptide synthesis (SPPS) of NBD-ff followed by a conjugation of cysteamine S-phosphate^[13] generates pS1 (Scheme S2) in a good yield. Such a design ensures the fast dephosphorylation of the thiophosphopeptide by ALP (Figure S1). While pS1 exhibits critical micelle concentration (CMC) of 6.0 μM, S1 has the CMC of 2.4 μM (Figure S2). Transmission electron microscopy (TEM) reveals that pS1, at 5 μM form micelles, which turn into nanofibers after ALP converts pS1 to S1 (Figure 1A and 1B). The formation of micelles of pS1 likely facilitates the cellular uptake by caveolin-mediated endocytosis, similar to the cellular uptake of peptide amphiphiles.^[14]

Figure 1.

(A) Chemical structures of pS1 and S1 and (B) the TEM images of pS1 (5 μM) before and after the addition of ALP (0.1 U/mL) for 24 h. Scale bars = 100 nm. CLSM image of HeLa cells (C) at 0 min immediately after the addition of pS1. The Golgi of HeLa is marked by arrows.

We first incubated HeLa cells with CellLight^® Golgi-RFP^[15] for 24 hours to transfect RFP at the Golgi, then incubated the HeLa cells with pS1 (10 μM) for 8 minutes. The fluorescence of S1 appears almost instantly after adding pS1 in the culture of HeLa cells (Figure 1C). This rate is, at least an order of magnitude, faster than previously reported probes.^[5-7] The fluorescence from the assemblies of S1 overlaps with the red fluorescence from all Golgi-RFP. In addition, there is no bleed-through from the Golgi-RFP staining in green (NBD) channel (Figure S3). These results confirm that pS1 targets the Golgi of the HeLa cells (Figure 2). The intensity of the fluorescence at the Golgi increases significantly with the time of incubation of pS1, about 7 times enhancement from 1 minute to 8 minutes. Except the bright fluorescence at the Golgi and the dim fluorescence at the endoplasmic reticulum (ER), the rest of intracellular and extracellular regions of the HeLa remain dark. This result indicates that enzymatic assembly of S1 occurs at the Golgi of HeLa cells, agreeing with the observation of ALP at Golgi of HeLa cells.^[16] Live cell imaging over 20 minutes (Video S1) reveals that the fluorescence of the assemblies of S1 emerges at the Golgi prior to diffusing to ER, likely resulted from Golgi-ER retrograde traffic,^[1b] which is inevitable due to interorganelle communication. Adjusting the concentration of pS1 or the treatment time according to the characteristics of different cell lines would minimize the fluorescent background due to Golgi-ER transport (vide infra (Figure 3)). The differential interference contrast (DIC) image of HeLa cells treated with 10 μM of pS1 for 20 min (Figure S4) shows well-spread HeLa cells, agreeing with the IC₅₀ value (38 μM) of pS1 (Figure S5) against the HeLa cells and excluding that pS1 enters cells due to cell death.

Figure 2.

CLSM images of HeLa cells (A) stained with CellLight^® Golgi-RFP after treating with pS1 for 8 minutes and (B) treated with pS1 for 1, 4, and 8 min. Scale bars = 20 μm and [pS1] = 10 μM.

Figure 3.

CLSM images of HeLa cells treated with pS1 (10 μM, 5 μM, 2 μM) for 4 min (top), S1 (10 μM, 5 μM, 2 μM) for 30 min (middle) and pO1 (100 μM, 50 μM, 20 μM) for 4 h. Scale bars = 20 μm.

The concentration of pS1 is another important factor determines the rate of accumulation of S1 at Golgi. We compared the Golgi fluorescence by fixing the incubation time at 4 minutes and varying the concentration pS1 at 10, 5, and 2 μM (Figure 3). At 10 μM, bright green fluorescence presents in Golgi and weak fluorescence in ER; at 5 μM, green fluorescence clearly still presents at the Golgi, but little at the ER; at 2 μM, much weaker fluorescence at the Golgi. Further decreasing the concentration of pS1 to 500 nM still results in Golgi accumulation of S1 in HeLa cells. Though the brightness of S1 at Golgi is weaker at the beginning of the addition, distinctive fluorescence appears at the Golgi after 15 minutes (Figure S6). The accumulation rate of S1 at Golgi, quantified by the increase of the fluorescence intensity, is concentration dependent (Figure S7). Using a Golgi disruptor, brefeldin A (BFA), is able to abolish the accumulation of S1 (Figure S8). These results indicate that enzymatic formation and self-assembly of S1 in situ at Golgi enables the instant targeting of Golgi. In addition, the concentration needed of pS1 for targeting Golgi is orders of magnitude lower than the previously reported probes.^[5-7]

Unlike pS1, S1, at the concentration of 10 μM and being incubated with HeLa cells for 8 minutes, hardly results in any fluorescence in the cells (Figure S9A), suggesting slower cell uptake of S1 than that of pS1 and indicating the importance of enzymatic dephosphorylation for targeting Golgi. As shown in Figure 3, after incubating S1 (10 μM) with HeLa cells for 30 minutes, some green fluorescence at the Golgi of the HeLa cells, with the fluorescent intensity similar to that of HeLa cells incubated with pS1 (2 μM) for 4 minutes. After 30 minutes incubation, when the concentrations of S1 are at 5 and 2 μM, there is very weak at the Golgi and no fluorescence in cell at all, respectively. These results suggest that S1 enters the HeLa cells less efficiently than pS1 does. Moreover, pO1 (the parent compound of pS1) produces almost no fluorescence (Figure S9B) in the HeLa cells after 8 minutes incubation and at the concentration of 10 μM. In fact, after 4 hours of incubation of pO1 and HeLa cells, there are several scattered fluorescent puncta in cells, with fluorescent intensity being proportional to the concentrations of pO1 (from 100 to 50 and to 20 μM). These results indicate that the assemblies of O1, formed by dephosphorylation, largely are retained in endosomes or lysosomes (Figure 3, Scheme S3). The above results confirm the unique ability of pS1 for instantly targeting Golgi.

To further understand the mechanism of Golgi-accumulation of S1 assemblies resulted from the rapid enzymatic dephosphorylation of pS1, we examined the rate of fluorescence increase in Golgi for 16 minutes in the HeLa cells treated pS1 and several inhibitors, using the fluorescence increase in the Golgi of the HeLa cells incubated with pS1 only as the reference (Figure 4A). Using phospholipase C (PLC),^[17] an enzyme that cleaves glycosylphosphatidylinositol (GPI) anchor, to remove ALP from the cell membrane results in slightly faster increase of fluorescence in the Golgi, confirming that ALP at Golgi dephosphorylates pS1 and indicating that pericellular dephosphorylation of pS1 by the ALP on plasma membrane slightly slows down the accumulation of S1 at the Golgi. Methyl-β-cyclodextrin (mβCD) significantly decreases the rate of the fluorescence increase at the Golgi, indicating that pS1 (at 10 μM) likely enters cells via caveolin-mediated endocytosis. As a potent inhibitor of actin polymerization,^[18] cytochalasin D (CytD) decreases the accumulation of S1 at Golgi in a concentration-dependent manner (Figure S10). This result agrees with the critical role of actin dynamics in cellular uptake, indicating that pS1 also enters the cells via marcopinocytosis.^[19] Both the phosphatase inhibitor cocktail set 3 (PIC) and the tissue nonspecific ALP inhibitor (DQB^[20]) reduce the rate of the fluorescence increase at the Golgi, with PIC more effectively inhibiting the accumulation than DQB. These results suggest that other phosphatases, in addition to ALP, also contribute to the enzymatic accumulation of S1 at Golgi and agree with sorting of ALP at the Golgi before secretion.^[21]

Figure 4.

Time-dependent mean fluorescence intensity of Golgi in (A) HeLa cells pretreated with PLC (phospholipase C, 0.2 U, 30 min), mβCD (5 mM, 30 min), PIC (phosphatase inhibitor cocktail set 3, 4000×, 30 min), DQB (20 μM, 30 min), respectively, and then treated with pS1 (10 μM), and the pS1 treated HeLa cells in a fresh medium. (B) Time-dependent mean fluorescence intensity of Golgi in different cell lines treated with pS1 (10 μM).

We also examined the exocytosis of the accumulated S1. HeLa cells were pretreated with pS1 (10 μM) for 1 h, and then the medium containing pS1 was replaced by the fresh medium without pS1. The curve is shown as “fresh medium” in Figure 4A. The intensity of fluorescence at the Golgi of the HeLa cells drops only slightly over time, confirming that the enzymatically formed assemblies of S1 are largely trapped in the Golgi. Inhibiting protein disulfide isomerases (PDIs) decrease disulfide bonds of cysteine rich proteins (CRPs)^[22] that are transported to Golgi, contributing to a slight decrease of the rate of Golgi accumulation of S1 (Figure S11). This result implies the peptide assemblies likely form disulfide bonds with CRPs. This observation agrees with that dimers of S1 (or S2) form in the cell lysate treated with pS1 (or pS2) (Figure S12), suggesting that certain extent of covalent linkage between assemblies likely contributes the retention of the assemblies in the Golgi. Moreover, the addition of N-ethylmaleimide (NEM), a molecule known to block the formation of disulfide bond of cysteine^[23], almost completely eliminates the accumulation of S1 at Golgi (Figure S13), further supporting that S1 forms disulfide bond with CRPs at Golgi.

To examine the applicability of the process illustrated in Scheme 1 for targeting Golgi of other cells, we incubated pS1 with several other cancer cell lines (Saos-2, SJSA-1, OVSAHO, HCC1937, HepG2, OVCAR-4, SKOV-3, MCF-7) and immortalized normal cell lines (HEK293 and HS-5) and examined the rates of fluorescent increase at the Golgi of the cells (Figure 4B). The fluorescence intensities increase significantly at the Golgi of Saos-2, SJSA-1, OVSAHO and HCC1937 cells, slightly in those of HepG2 and OVCAR4 cells, and much slowly in those of SKOV-3, MCF-7, HEK293 and HS-5 cells. These results largely agree with expression levels of ALP in these cell (Figure S14).^[24] One exception is HepG2, which expresses higher level of ALP than OVSAHO, but exhibits slower fluorescence increase at Golgi. High level of glutathione in hepatocytes^[25] likely antagonizes the accumulation of S1 in the Golgi. This observation supports that oxidative Golgi environment favors disulphide bond formation, thus the retention of the assemblies of S1 at the Golgi.

We synthesized a nonfluorescent analogue (pS2) of pS1 by using naphthyl group to replace NBD (Figure 5A). Being similar as pS1, pS2 undergoes rapid dephosphorylation by ALP to form S2 (Figure S15). Compared with its oxophosphate analogue pO2,^[26] pS2 shows much faster dephosphorylation. For example, being incubated with ALP (0.1 U/mL) for about 16 minutes, pS2 nearly fully converted to S2, while the maximum conversion ratio of pO2 to O2 remains at about 50% at the same duration (Figure S15). The CMC of pS2 is 9.5 μM, and the CMC of its dephosphorylated product, S2, is 4.3 μM (Figure S16), indicating that both compounds have an excellent self-assembling ability. We further tested the cytotoxicity of pS2 against HeLa, HEK293, and HS-5 cells and found that the IC₅₀ values are 2.8 μM, >100 μM and >100 μM, respectively (Figure 5B). The inhibitory activity of pS2 against HeLa cells is an order of magnitude higher than that of pO2 (Figure S17). The difference between these IC₅₀ values agrees with the difference of the rate of enzymatic assemblies in Golgi of the cells, indicating that selectively targeting the Golgi is mainly resulted from fast enzyme kinetics. This result also confirms that pS2 is more selective than S2 against cancer cells (Figure S18). In addition, several commonly used inhibitors (Z-VAD-FMK, NAc, Nec-1, DFO, Fer-1, and disulfiram)^[27] of cell death are unable to rescuing HeLa cells from pS2, but an oncosis inhibitor (PD 150606)^[27] can partially rescue the HeLa cells from pS2 (Figures S19 and S20). These results suggest a unique cell death resulted from the molecular processes defined by the enzymatic reactions and assemblies of thiophosphopeptides (e.g., pS1 and pS2) at the Golgi apparatus. Treating the HeLa cells incubated with pS2 by the tetracysteine probe, FlAsH-EDT₂,^[28] results in the fluorescent at the Golgi of the HeLa cells (shown by the arrow in Figures 5C, S21). The fluorescence of Figure 5C appears weaker than that in Figure 3 because FlAsH-EDT₂, as a secondary staining agent, only interacts some of multimers of thiols in assemblies of pS2. This observation is further confirmed by adding Golgi marker in the incubation (Figure S22). These results indicate that S2 self-assembles at the Golgi to organize the multiple C-terminal thiols into arrangement similar to that of tetracysteine.

Figure 5.

(A) Structures of pS2 and S2. (B) IC₅₀ of pS2 against HeLa cells, HEK293 cells and HS-5 cells. (C) CLSM images of HeLa cells treated with pS2 (10 μM, 4 h) and then stained by a tetracysteine probe, FlAsH-EDT₂. Scale bar = 20 μm.

In summary, this work illustrates that rapid dephosphorylation of thiophosphopeptides and the subsequent disulfide bond formation enables instantly targeting of Golgi apparatus and selectively inhibiting the cancer cells. Our observations agree with several known facts: (i) CRPs are enriched in Golgi,^[29] (ii) a significant level of oxidation occurs in the Golgi membrane,^[30] (iii) ALP, as an “almost perfect” enzyme^[31] being anchored on the cell membrane by glycosylphosphatidylinositol (GPI) and overexpressed on certain cancer cell,^{[16, 32]} is known to be sorted as oligomers at the Golgi before secretion.^[21] This work also underscored the importance to identify the targets of thiophosphopeptides and the assemblies of thiopeptides for design different molecules capable of engaging the same pathways. Since thiophosphopeptides^[33] are much less developed than phosphopeptides, replace NBD with other functional motifs (e.g. 10-hydroxycamptothecine)^[34] in the thiophosphopeptide may lead to new discoveries. Enzymatic assemblies by thiophosphopeptides, besides acting as a molecular process to reveal the state of Golgi, may provide substrates for thiol-click chemistry^[35] or for integration thiol groups in other supramolecular assemblies.^[36]