*2.1. Chemistry*

The detailed synthesis of compound **1** derivatives (substituted at the C-4 position with allyl (**6**), hydroxymethyl (**10**), acetoxymethyl (**11**), methoxymethyl (**12**)), novel 2,6-dimethyl curcumin (DMC) derivatives (**13** and **15**), and the ester hydrolysis products of derivatives showing potential anticancer activity (**2m**, **3m**, **4m**, **5m**, **6m**, **16**, and **17**) are depicted in Schemes 3 and 4. As shown in Scheme 3, compound **6** was synthesized according to a procedure established for the synthesis of **2**–**5**. Compound **7**, prepared by the esterification of curcumin with 2,2,5-trimethyl-1,3-dioxane-5-carboxylic acid, was treated with 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU) and formaldehyde in THF to produce intermediary compound **8**. The latter was used without column chromatographic purification in acid-promoted hydrolysis at room temperature to produce **9** with an overall yield of 6% for two synthetic steps.

Compound **9**, however, proved unstable, slowly decomposing even when kept at 4 ◦C and presenting an NMR spectrum comprising a messy jungle of peaks after being dissolved in DMSO-d6 for 1 week at room temperature. The low stability of **9** may have been due to self- or cross-hydrolysis, such that the two hydroxyl groups at the C-4 alkyl chains cleaved the ester linkage of the 2,2-bis(hydroxymethyl)propionate group. Previously, we had found that upon exposure to alcohol for several hours, the ester linkages of **1** were broken under neutral or basic conditions, leading to the conversion of **1** to curcumin and several unknown products. Therefore, we modified the hydroxymethyl group of **9** and were able to produce compounds **10** and **11**. To prepare **10**, compound **9** was acetylated with acetyl chloride in the presence of Et3N, and the resulting di-acetylation intermediate was hydrolyzed by hydrochloric acid. Compound **11** was obtained with an overall yield of 5% following a similar three-step reaction modified for methylation. The stability of these two products at room temperature may have been due to the conversion of the labile hydroxyl groups of the C-4 alkyl chains of **10** and **11** into non-nucleophilic methyl and acetyl groups, respectively, which prevented degradation.

**Scheme 3.** *Reagents and Conditions:* (**a**) Formaldehyde in THF, DBU, 0 ◦C to rt, 2 h; (**b**) HCl, MeOH, rt, 1 h, 6% for two steps; (**c**) Et3N, acetyl chloride, DCM, rt, 12 h; then HCl, MeOH, rt, 1 h, 3% for three steps; (**d**) K2CO3, CH3I, DMF, rt, 6 h; then HCl, MeOH, rt, 1 h, 4% for three steps; (**e**) Formaldehyde in THF, DBU, 0 ◦C to rt, 2 h, 42%; (**f**) Compound 13, Et3N, DCM, rt, 12 h; and (**g**) HCl, MeOH, 1 h, 7% for two steps.

Because DMC is a dimethylated derivative of curcumin [26] that exhibits better anticancer activity and stability than curcumin, we synthesized a novel DMC derivative (compound **12**) in which C-4 was substituted with two bulky 2,2-bis(hydroxymethyl)propionate groups. To produce this derivative, commercially available DMC was reacted with DBU and formaldehyde in THF. The reaction of **12** with 2,2,5-trimethyl-1,3-dioxane-5-carbonyl chloride **13** [27] in the presence of Et3N provided a diol intermediate, which was then hydrolyzed by hydrochloric acid in MeOH to yield compound **14**. In addition, as illustrated in Scheme 4, compounds **2m**–**6m** were prepared through the NaOMe-mediated hydrolysis of their corresponding parent compounds **2**–**6**, with high yield.

**Scheme 4.** *Reagents and Conditions*: (**a**) NaOMe, MeOH, rt, 2 h, 82–87%; (**b**) K2CO3, alkyl iodide, DMF, rt, 20 h; and (**c**) NaOMe, MeOH, rt, 3 h, 28–33% for two steps.

For SAR analysis, compounds **15** and **16**, which were substituted with two straight alkyl chains with three or four carbons at the C4 position, were prepared from curcumin bisacetate **17** [28] via a two-step procedure comprising K2CO3-induced dialkylation and NaOMe-mediated hydrolysis. For the synthesis of four possible phase I metabolites of compound **2**, compound **2m** was subjected to hydrogenation reactions. As shown in Scheme 5, compound **18** is a partially hydrogenated product that can be prepared in 19% yield by treating compound **2m** with H2(g) (1.0 atm, balloon) and a catalytic amount of 10% *w*/*w* Pd/C in ethyl acetate. When the solvent was changed from ethyl acetate to methanol, the hydrogenation reaction yielded the fully hydrogenated diketone **19** in 52% yield and β-hydroxy ketone **20** in 23% yield. Finally, the reduction of **20** with NaBH4 in MeOH at room temperature (rt) for 2 h provided the desired diol compound **21** in 31% yield.

**Scheme 5.** *Reagents and Conditions*: (**a**) 10% *w*/*w* Pd/C, ethyl acetate, rt, 1 h, 19%; (**b**) 10% *w*/*w* Pd/C, MeOH, rt, 2 h; and (**c**) NaBH4, MeOH, 0 ◦C to rt, 2 h, 31%.

#### *2.2. Structure–Activity Relationship*

As mentioned above, we synthesized **2**–**5** and evaluated their anti-proliferative activity against the MDA-MB-231 and HCT-116 cell lines (72-h treatment). These compounds, along with all newly synthesized compounds, were screened for anti-proliferative activity against the same TNBC and colon cancer cell lines. The results of the in vitro assay are compiled in Table 1.


**Table 1.** The anti-proliferative effects of curcumin derivatives against MDA-MB-231 and HCT-116 cell lines.

<sup>a</sup> All data presented are means from at least three experiments with standard deviations of the value quoted.

The anticancer effects of **2**–**5** against both cell lines were still better or at least similar to that of **1**, although the drug treatment duration was reduced to 48 h. The newly synthesized compound **6** had a higher anti-proliferative activity against HCT-116. Compound **9** was less potent against the cell lines than curcumin and **1**, which was used as a reference compound in the current study. We hypothesize that the two polar and hydrophilic side chains at the C-4 position in **9** are the cause for the weak anticancer effects observed for this compound. Another possibility is that **9** decomposed during the screening due to its low stability. Compound **10**, having two acetoxymethyl groups rather than hydroxymethyl groups at the C-4 side chains, showed significant improvement in in vitro anticancer activities compared to **9**; however, it was still less potent than **1**. Compound **11**, substituted with two methoxymethyl groups at the C-4 position, displayed poor anticancer activity with high IC50 values for both cell lines used.

After listing the compounds in descending order based on their potency against HCT-116 (**6** > **3** > **2** > **5** > **1** > **10** > **4** >> **11** >> **9**), we concluded that higher polarity and hydrophilicity of the C-4 side chain are detrimental to the anticancer activity of compound **1** derivatives. Alternatively, the DMC derivative 4,4-dihydroxymethyl dimethoxycurcumin (compound **12**) displayed low anti-proliferative activity. When the two hydroxyl moieties of **12** both esterified with 2,2-bis(hydroxymethyl)propionic acid, the resulting compound **14** exhibited activities comparable to those of **12**, which correlates with our previous findings regarding the unfavorable effect of having a hydrophilic moiety at the C-4 position.

Subsequently, the activities of the ester hydrolysis products **2m**–**6m** were explored. Unexpectedly, **5m** displayed substantial improvement in anticancer activity by two- to four-fold relative to that of parent **5**. Compound **2m** exhibited activity similar to that of **2**, whereas **3m**, **4m**, and **6m** showed inferior activity than their corresponding parent compounds **3**, **4**, and **6**. Further screening of **15** and **16** revealed an obvious SAR, such that the anticancer activity of these hydrolysis compounds decreased with the increasing length of the C-4 alkyl side chains. In addition, lower structural volume was seemingly preferred to sterically congested side chains, as demonstrated by the analysis of **2m**, **3m**, **15**, **4m**, and **16**.

Among all examined compounds, **2** and its ester hydrolysis product **2m** possessed significant anticancer activity against MDA-MB-231 and HCT-116; therefore, **2** was chosen for further studies. Notably, the solubility of **2** was more significant than that of curcumin as was evident from the vehicle of **2** for the following in vivo study. The maximum solubility of **2** in the aqueous vehicle comprising 5% ethanol, 10% Tween 80, and 85% saline was approximately 200 mg/mL, while curcumin was almost insoluble in the vehicle. Based on the logic for the metabolism of curcumin [29] and an ester-type prodrug [30], we speculated that the in vivo metabolism of **2** would readily produce **2m**. Subsequently, the latter would be converted into glucuronide **22** and sulfate **23** through the phase II metabolism conjugation of hydrogenated products **18**, **19**, **20**, and **21** via the phase I metabolic pathway (Scheme 6). To verify our assumptions, the preliminary pharmacokinetic evaluation of **2** was executed.

**Scheme 6.** In vivo metabolism of **2**.

### *2.3. The Preliminary Pharmacokinetic Evaluation of 2*

The preliminary in vivo pharmacokinetic evaluation of **2** was designed to identify the anticipated metabolites and evaluate the plasma stability of **2** and **2m**. The four possible phase I metabolites **18**, **19**, **20**, and **21** were synthesized and subjected to anti-proliferative screening against MDA-MB-231 and HCT-116. Compound **18** exhibited moderate to weak anti-proliferative activity (MDA-MB-231, IC50 = 36.48 ± 0.34 μM; HCT-116, IC50 = 9.64 ± 0.21 μM) and the others are inactive against both cell lines (IC50 ≥ 100 μM). The assessment of **2** and relevant metabolites was performed in male Sprague-Dawley *rats* after the oral administration of **2** at a dose of 100 mg/kg. The LC signals of **2**, **2m**, and **18**–**21** was recognized unequivocally by comparison to the signals of standard compounds. The direct analysis of **22** and **23** was not easily attainable due to the low extraction efficiency and high probability of sample loss in LC column. It has been well-documented that curcumin glucuronide and curcumin sulfate can be transformed back into curcumin by the enzyme-mediated hydrolysis reaction [31]. Accordingly, the treatment of **22** and **23** with enzyme that contains sulfatase and glucuronidase is supposed to produce **2m**. In the current study, the amounts of **22** and **23** were not calculated individually but counted as a summary value based on the disparities between the LC signal of **2m** in enzyme-treated and -untreated serum samples.

In practice, LC-MS analysis of one serum sample which was collected at 20 min post-dosing demonstrated the existence of **2**, **2m**, **18**, **19**, **20**, **21**, **22** and **23** (Figure S1 in the Supporting Information). Shown in Figure 1 are the LC signal count-time profiles for **2**, **2m**, **18**, **19**, and the comparison of **2m**, **22**, and **23**, following oral administration. The analysis of **20** and **21** is not illustrated due to their low LC signal response and non-significant anticancer activity. The maximum blood content of **2** was achieved at 20 min, while the second and third highest appeared at 120 min and 360 min, respectively, following administration. The multiple peaks of **2** observed in the profile could be attributed to an enterohepatic circulation cycle.

**Figure 1.** The amount of **2**, **2m**, **18**, **19**, **22**, and **23** in serum at different time points following a single oral administration of **2** at a dose of 100 mg/kg. The blood samples were collected at 5, 10, 20, 30, 60, 120, 240, 360, and 480 min after the administration of **2**. The samples were deproteinated and analyzed by HPLC.

Such behavior, extensively studied in the phenolic drugs, was beneficial in maintaining the circulating concentration of free-form curcumin [32] and may extend the pharmacological effect of drug [33]. The hydrolysis metabolite **2m** emerged from 10 min and reached its maximum blood content at 120 min, indicating that a medium-to-high esterase-mediated biotransformation proceeded in vivo to yield the efficient absorption of **2m**. Subsequently, the phase II metabolic process of **2m** progressed quickly to produce **22** and **23**. The ratio of **2m** to its phase II metabolites (**22** and **23**) is about 1:4 based on the area under the curves in part (e) of Figure 1.

The LC signal count-time profiles of **18** and **19** are analogous to that of **2m**, implying that phase I reduction reaction also occurred rapidly. Compared to non-formulated curcumin with a short elimination period of 28 min in the rat model [34], **2** exhibited superior plasma stability. Otherwise, the active metabolite **2m** is equally potent as **2** with regard to anticancer activity, and the maximum blood content of **2m** appeared after that of **2** could be considered as an extension of drug efficacy of **2**. Thus, the pharmacokinetic outcome was to guide the following in vivo efficacy study of **2**.

#### *2.4. In Vivo Antitumor E*ffi*cacy of 2*

The therapeutic effect of **2** was evaluated on HCT-116 colon tumor xenograft bearing nude mice. The oral administration of **2** at doses of 50, 100, or 150 mg/kg bw per day were performed after tumors reached approximately 100 mm<sup>3</sup> and continued for 30 consecutive days. 5-Fluorouracil treatment (30 mg/kg bw, i.p., QOD) was used as a positive control. The group treated with compound **2** (100 mg/kg) had a statistically significant reduction in tumor volumes, with a 45% tumor inhibition ratio observed when compared with the cancerous control mice. Positive control group exhibited a capability to cause 40% tumor-growth inhibition. There was a clear dose- and time-dependent inhibitory effect on tumor size in the other two groups, in which at 50 and 150 mg/kg doses the volume of the HCT-116 colon tumor was reduced to 73% and 40% of the control, respectively (Figure 2). The body weight of the mice was recorded prior to dosing every 3 d, and the average body weight of each group is depicted in Figure 3. Otherwise, they were monitored for visible signs of toxicity and behavioral changes 1 h after each administration. No obvious adverse effects were observed between the **2**-administrated and control groups.

**Figure 2.** The efficacy study of **2**. In vivo antitumor activity of compound **2** was evaluated in HCT-116 tumor xenograft bearing nude mice. Compound **2** (50, 100, or 150 mg/kg, p.o.) was administered once daily and 5-Fluorouracil was administrated (30 mg/kg, i.p.) every other day. The results are the mean ± SEM of six mice in each group. \*\*\* *P* < 0.001 compared to untreated control.

**Figure 3.** Mean body weight-time profile of **2**- administrated and control groups.

#### **3. Materials and Methods**

#### *3.1. General Information*

The reactions were performed under an air atmosphere unless otherwise stated. All solvents and reagents were employed as received. Analytical thin layer chromatography (TLC) was performed on SiO2 60 F254 plates and flash column chromatography was carried out using SiO2 60 (particle size 0.040–0.055 mm, 230–400 mesh), both of which are available from E. Merck (Darmstadt, Germany). Visualization was performed under UV irradiation at 254 nm followed by staining with aqueous potassium permanganate [KMnO4 (3 g) and K2CO3 (20 g) in 300 mL of H2O containing 5 mL of an aqueous solution of NaOH (5%, *w*/*v*)] and charring by heat gun. 1H- and 13C-NMR spectra were recorded on a 500 FT NMR instrument (Bruker, Billerica, MA, USA). Chloroform-*d* and methanol-*d* were used as solvents and TMS (δ = 0.00 ppm) as an internal standard. Chemical shifts are reported as δ values in ppm as referenced to TMS. Multiplicities are recorded as s (singlet), d (doublet), t (triplet), q (quartet), quint (quintet), sext (sextet), sept (septet), dd (doublet of doublets), dt (doublet of triplets), br (broad), m (multiplet). Coupling constants (*J*) are expressed in Hz. LRMS and HRMS were measured by a JMS-HX110 spectrometer (JEOL, Tokyo, Japan) and spectroscopic data were recorded as *m*/*z* values. Melting points were measured using an Electrothermal instrument (Anatec Yanaco Inc., Kyoto, Japan).
