Iodobenzohydrazides of o-iodobenzohydrazide (L1), m-iodobenzohydrazide (L2), p-iodobenzohydrazide (L3), 3,4-diiodobenzohydrazide (L4), and 2,3,5-triiodobenzohydrazide (L5) can be conveniently prepared by the refluxing reaction of hydrazine hydrate (85%) and corresponding ester in the solvent of ethanol, which were evidenced by FTIR (Supplementary Figures S1−S5), ¹H NMR, ¹³C NMR (Supplementary Figures S11−S15), and electrospray ionization mass spectrometry. Among them, L4 and L5 are prepared for the first time. As shown in Scheme 1, compounds 1–5 were obtained via the refluxing reaction of α-octamolybdates, corresponding iodobenzohydrazides and DCC with the molar ratio of 3:4:6 in the solvent of anhydrous acetonitrile. The resulting mixture was filtered to remove the insoluble white precipitates of DCU and afforded a black-red filtrate, which is diffused by ethyl ether. Two weeks later, lots of black or black-red crystals can be obtained, accompanied by a lot of white block crystals (they are proved to be octamolybdates by FTIR). It is pitiful that these crystals of title compounds cannot be collected by single-crystal X-ray diffraction, although we had tried to collect their crystals many times. According to our previous research (Wang et al., 2017), the molecular formula of compounds 1–5 are (TBA)₃ [Mo₆O₁₈(=N=NCOAr)] (Ar = Ph-2-I (1); Ph-3-I (2); Ph-4-I (3); Ph-3,4-I₂ (4); Ph-2,3,5-I₃ (5) (TBA = tetrabutylammonium), respectively. Their structures can be viewed that one terminal O_t of hexamolybdates (POM-0) was replaced by the ligand of iodobenzoyldiazenido, POM-0, and the corresponding iodobenzoyldiazenido is connected via the Mo-N multiple covalent bond (as shown in Scheme 1). Their structures are determined and evidenced by UV-Vis, FTIR, X-ray photoelectron spectroscopy (XPS), ¹H NMR, HRMS, and elemental analysis.

UV-Vis spectroscopy of title compounds was measured in their solution of acetonitrile (approximately 10^–5 M) at room temperature. As shown in Supplementary Figures S16, the lowest absorption band of compounds 1–5 are found as broadband with the peak of 370 nm for 1, 382 nm for 2, 384 nm for 3, 392 nm for 4, and 392 nm for 5. There is a bathochromic shift approximately 60 nm compared with that of their parent POMs of hexamolybdate, in which the lowest absorption arising from the electron transfer of O_t to Mo (LMCT) appeared at approximately 320 nm (Strong et al., 2000). Notably, the lowest absorption band of 1 is lower than that of other compounds; this is possible because the large steric hindrance of iodo atom in POM-1 compels the acryl group to deviate from the phenyl ring plane and destroy their conjugation between the phenyl ring and POMs (Gougoutas and Clardy, 1970). The large bathochromic shift of the lowest absorption band in title compounds indicates the formation of the multiple bond Mo=N, in which the hydrazide ligand is connected with POMs via the multiple covalent bond of Mo=N, the conjugation between the organic ligand and POMs lowers the energy level of title compounds and results in the large bathochromic shift of title compounds.

FTIR spectra are further used to study the structure of title compounds. As shown in Supplementary Figures S6−S10, in the high wavenumber region of 1,350–4,000 cm⁻¹, the peaks around approximately 2,960, 2,873 cm⁻¹ in 1, 2,960, 2,872 cm⁻¹ in 2, 2,960, 2,873 cm⁻¹ in 3, 2,960, 2,873 cm⁻¹ in 4, and 2,960, 2,873 cm⁻¹ in 5 are ascribed as the C-H stretching vibration of methyl and methylene, verifying the presence of TBA cation. The strong peak around approximately 1,638 cm⁻¹ in 1, 1,647 cm⁻¹ in 2, 1,639 cm⁻¹ in 3, 1,637 cm⁻¹ in 4, and 1,628 cm⁻¹ in 5 is attributed to the stretching vibration of C=O in compounds 1–5, indicating the presence of carbonyl in those compounds (Wang et al., 2017). There is a slight redshift approximately 5–10 cm⁻¹ relative to the υ(C=O) in corresponding ligands (1,638 cm⁻¹ in L1, 1,640 cm⁻¹ in L2, 1,649 cm⁻¹ in L3, 1,650 cm⁻¹ in L4, and 1,632 cm⁻¹ in L5). The redshift of υ(C=O) in compounds 1–5 is mainly originating from the conjugation between the organic segment of benzoyldiazenido and POMs. In the low wavenumber region of 400–1,350 cm⁻¹, the strong peak around 944 cm⁻¹ in compounds 1–5 is attributed to the stretching vibration of Mo=O_t, and their shoulder peaks around 968 cm⁻¹ are associated with the stretching vibration of Mo=N. Compared with that of the parent POMs, hexamolybdates, in which the stretching vibration of Mo=O_t appeared at 960 cm⁻¹ (Clemente-León et al., 2001), there is a redshift approximately 17 cm⁻¹ for υ(Mo=O) as one consequence of one terminal O_t substituted by benzoyldiazenido ligand. The new-formed shoulder peak around 968 cm⁻¹ can be diagnostic for the formation of Mo=N. Similarly, another peak around 799 cm⁻¹ associated with the stretching vibration of Mo-Ob in hexamolybdates is also found to be split into two peaks (798 and 767 cm⁻¹ for 1, 791 and 766 cm⁻¹ for 2, 799 and 768 cm⁻¹ for 3, 792 and 769 cm⁻¹ for 4, and 792 and 769 cm⁻¹ for 5) owing to the substitution of benzoyldiazenido ligand.

XPS was performed to further confirm their composition. Full XPS spectra of compounds 1–5 reveal the presence of C, N, O, Mo, and I in those compounds (Supplementary Figures S11−S15). Narrow XPS spectra of compounds 1–5 show the characteristic Mo (3d) doublet (Mo_3d5/2, Mo_3d3/2) approximately 231.97, 235.18 eV for 1, 231.97, 234.7 eV for 2, 231.97, 235.23 eV for 3, 231.97, 235.14 eV for 4, and 231.81, 235.05 eV for 5, in agreement with the literature values of molybdenum (VI) (231.3 eV, 235.8 eV; Wagner et al., 1977; Zhu et al., 2019). The doublet of I (3d) (I_3d5/2, I_3d3/2) is found approximately 620.71, 632.24 eV in 1, 620.87, 632.40 eV in 2, 620.88, 632.38 eV in 3, 620.74, 632.33 eV in 4, and 620.67, 632.17 eV in 5, in accordance with the literature values of iodine (I₂) (619.9 eV, 631.42 eV; Wagner et al., 1977). XPS results confirm the presence of molybdenum and iodine in compounds 1–5, and their oxidation state is +6 and -1, respectively.

¹H NMR spectra of POMs 1–5 (Supplementary Figures S21−S26) and corresponding ligands L1–L5 (Supplementary Figures S16−S20) in d ₆-DMSO solution (Supplementary Table S1) show clearly resolved signals, all of which can be unambiguously assigned. Three signals (8.29, 8.01, and 7.54 ppm) of aryl hydrogen and two signals (8.24, 8.15 ppm) of aryl hydrogen are found in the ¹H-NMR spectra of L4 and L5, respectively, indicating the successful preparation of L4 and L5. In the ¹H-NMR spectra of compounds 1–5, these signals of TBA are found approximately 3.17, 1.57, 1.32, and 0.93 ppm and further evidenced the presence of TBA. The signals of –NHNH ₂ of L1–L5 disappeared in the ¹H-NMR spectra of compounds 1–5, but the signals of aryl hydrogen around 7–8 ppm are found in that of compounds 1–5, which means the presence of aryl ring and the formation of Mo=N multiple bond in compounds 1–5. As shown in Supplementary Table S1, the chemical shifts of aryl hydrogen in compounds 1–5 show slight upfield shifts relative to that of corresponding hydrazides. Notably, only one signal of aryl hydrogen (2H) is found in 5, which appears at the place of 8.05 ppm; this is different from that of L5, which possesses two signals around 8.24 and 8.15 ppm. It can be deduced that both aryl hydrogens in 5 moved to the same chemical shifts coincidentally after the combination of L5 with hexamolybdates via multiple covalent bond of Mo-N. The changing trend of chemical shifts in compounds 1–5 is quite different from their sisters, arylimido derivatives, in which all aryl hydrogen atoms demonstrate an obvious downfield shift compared with that of corresponding free aryl amines (Wei et al., 2001; Wang et al., 2009). POM cluster plays a role of the electron-withdrawing functional group in the arylimido derivatives. However, the POM cluster of [Mo₆O₁₈(NN)]⁻³ in compounds 1–5, as an anion cluster, possessing stronger electron-donating ability compared with –NHNH₂, can alleviate the electron deficiency of carbonyl to some extent and lower the deshielding effect of aryl ring to aryl hydrogen.

Gas chromatography coupled to HRMS was used to study compounds 1–5. As shown in Supplementary Figures S27−S36, in the positive region, there are one or two molecular ionic peaks of compounds 1-4, one molecular ionic peak found m/z = 2,092.4213 in 1, 2,092.4216 in 2, 2,093.4179 in 3, and 2,218.3156 in 4 is corresponding to the molecular ion of [M + TBA]⁺ (calcd: 2,092.4286 for 1–3, calcd. 2,218.3252 for 4); another small molecular ionic peak found m/z = 1,852.1452 in 1, 1,853.1464 in 2, and 1,852.1452 in 3 is well matched with the molecular ion of [M + H]⁺ (calcd: 1,852.1511). In the negative mode, the strong peak found m/z = 1,607.8468 in 1, 1,607.8475 in 2, 1,607.8449 in 3, and 1,733.7453 in 4 is corresponding to the molecular ion of [M - TBA]^1- (calcd: 1,607.8434 for 1–3, calcd. 1,733.7400 for 4). The small peak found m/z = 1,366.5709 in 1, 1,366.5709 in 2, 1,366.5690 in 3, and 1,492.4680 in 4 is corresponding to the molecular ion of [M–2TBA + H]^1- (calcd: 1,366.5658 for 1–3, calcd. 1,492.4625 for 4). Different from that of compounds 1–4, in the HRMS of compound 5, none molecular ionic peak of [M + TBA]⁺ or [M + H]⁺ in the positive pattern except one ionic peak of [M–I + TBA]⁺ (calcd. 2,218.3252) found m/z = 2,218.3205; in the negative pattern, a very small molecular ionic peak of [M - TBA]^1- (calcd. 1,859.6439) found m/z = 1,859.6378. Moreover, two ionic peaks of [M - I- TBA]^1- (calcd. 1,732.7472) and [M–I–2TBA + H]^1- (calcd. 1,492.4697) found m/z = 1,732.7460 and 1,492.4689, respectively. This means that compound 5 is easy to lose one iodine atom under the ionization process.

These synthesized compounds were firstly evaluated for in vitro cytotoxic effects against Hep-2 (human laryngeal epithelial cancer) cell lines by the standard 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide assay (Mosmann, 1983; Priti et al., 2018); the 50% cell cytotoxic concentrations (CC₅₀s) of compounds are summarized in Table 1. As shown in Table 1, most of the compounds showed low toxicity to the tested cell lines. It is obvious that the CC₅₀s of most compounds except compound 3 are larger than that of corresponding ligands, indicating that the covalent modification can reduce the in vitro cytotoxic effects of hydrazides to some extent.

According to the in vitro cytotoxic results, the antiviral activities of title compounds were further evaluated for the inhibition of virus-induced cytopathic effects. The concentration-dependent antiviral activities are shown in Figure 1. The results showed that these tested compounds displayed moderate to good inhibitory activities against CVB3. The parent POMs, hexamolybdates (POM-0), exhibit moderate antiviral effects. Although compounds L3, L5, 3, and 5 with inhibition rates beyond 90% exhibited potent anti-CVB3 activities, they showed better inhibitory activity than the positive control ribavirin. Especially for L5 and 5, they demonstrate the lowest EC₅₀s and the highest selective index in these compounds. This is possible because there are three iodine atoms in L5 and 5, which afford them more chances to combine with target points of CVB3s via halogen oxygen interaction and hydrogen bonding. Notably, the molecular inhibition efficiency of 3 and 5 surpasses that of L3 and L5, considering the large molecular weight of 3 and 5, indicating their molecular inhibition efficiency against CVB3 enhanced after the covalent linkage between the hydrazides and POMs.

To further explore the potential antiviral activities, EC₅₀ values of the tested compounds were also evaluated, and ribavirin was used as a positive control. As shown in Table 1, EC₅₀s of L5 and 5 were 13 and 12 μg/ml, respectively. The selectivity index (SI) of L5 and 5 were 9.6 and 10.7, respectively. It means that both compounds showed more potent inhibitory activity against CVB3 than ribavirin (EC₅₀ = 24.7 μg/ml, SI = 8.3).

To this end, the confluent monolayers Hep-2 cells in a 96-well plate were infected with 100 TCID50 of CVB3 mixed with or without the tested compounds L5 and 5 at the concentration of 25 μg/ml. After 10 h, the culture media and cell lysates were collected after freeze–thaw cycles and then subjected to virus titration. Treatments with tested compounds resulted in efficient reductions in progeny virus titers (Figure 2), with a reduction of approximately 3.5 log for L5 and a 4.0 log reduction for compound 5.