English

• As a class of natural dietary phytochemicals, flavonoids are widespread in fruits, vegetables, beverages, and herbs [1]. It has been experimentally demonstrated that flavonoids are not only potent in pharmacological activities such as anti-oxidant [2], anti-microbial [3], anti-viral [4], anti-inflammatory [5], anti-platelet aggregation [6], and neuroprotective effects [7], but also exhibit protective effects against tumor [8] and diabetes [9]. Flavonoids have also been proven to possess an inestimable developing potential in the fields of nutraceutical, pharmaceutical, medicinal, and cosmetic industries. However, the weakness of extremely low solubility, low absorption, and hence unfavorable bioavailability has immensely limited their potential applications [10,11]. Therefore, it is imperative to conduct in-depth research exploring effective means aiming to enhance the solubility and hence oral bioavailability of flavonoids.

  Recently, cocrystal has been commonly proposed and accepted as an effective approach by pharmaceutical scientists for the ability to ameliorate the dissolution behaviors of active pharmaceutical ingredients (APIs) at the molecular level [12-17]. Cocrystals are multi-component assemblies by two or more neutral organic molecules into the same crystal lattice in supramolecular architecture with a definite stoichiometric ratio held together by hydrogen bonds, van der Waals forces, π-π stacking, and other non-covalent interactions [18-21]. Unlike single-component materials, the resulting multicomponent systems offer the advantages of tunable composition, and adjustable molecular arrangement, and intermolecular interactions within their solid states. Many materials can be used as the coformers. Due the diverse nature of functional groups presents in the APIs and coformer molecules which makes it more difficult to precise analysis the intermolecular interactions thus increased complexities to the cocrystal screening. Traditional cocrystal screening strategy is the trial and error method, and had the disadvantages of time-consuming, labor-intensive, and low success rate. Thus, there is a real need in the pharmaceutical industry to optimize screening methodologies in order to reduce the number of experiments required to find potential coformers. The past decades have witnessed vast research on the modification of physicochemical properties of APIs by the utilization of molecular cocrystals [22-26]. However, metallic, ionic cocrystals and zwitterionic cocrystals [27-29] are subclassifications of cocrystals that are less explored.

  Amphoteric or bipolar molecules can exist in the neutral or zwitterionic state controlled by the pH. To date, the application of bipolar molecules in cocrystallization has been investigated and reported in some cases [30-33]. For example, Wang et al. have investigated and proposed the first zwitterionic cocrystal of indomethacin with L-proline [34], in which in vitro and in vivo properties of indomethacin have been satisfactorily modified. Picolinic acid (PA, Fig. 1) is a bipolar molecule, where there are two strong functionalities: a carboxylic group and a pyridine N atom. PA having neuroprotective properties and can antagonize the neurotoxic effects of quinolinic acid [35]. The presence of multiple hydroxyl groups in the flavonoids enables them good hydrogen bond donors for the design of intermolecular interactions. Resting on the above consideration, it is of significant reasonability to design and synthesize cocrystals of hydroxyl–rich flavonoids with PA.

  Figure 1

  

  Figure 1.  The chemical structures of the 17 flavonoids and picolinic acid (neutral and zwitterionic state) used in this study.

  DownLoad: Full-Size Img PowerPoint

  In this paper, a model system consisting of PA and an extended list of 17 different flavonoids (Fig. 1) were built to investigate the cocrystallization behaviors by using a combined computational and experimental approach. LUT, GEN, TAX, DHM, quercetin (QUE), myricetin (MYR), hesperetin (HES), kaempferol (KAE), naringenin (NAR), apigenin (API), isoliquiritigenin (ILG), resveratrol (RES), 7-hydroxyisoflavone (7HIF), chrysin (CHR), baicalin (BAI), daidzein (DAI), and formononetin (FOR) were all purchased from Shanxi Huike Plant Development Co., Ltd., China. PA was obtained from Wuhan Yuancheng Co-creation Technology Co., Ltd., China. We use two different computational screening tools: the hydrogen-bond propensity (HBP) approach [36-38] and the modified molecular electrostatic potential (MEP) by us based site-pair interaction energy computations [39] to evaluate the cocrystal formation between flavonoids and PA. The HBP approach determines the probability of cocrystal formation based on the frequency of existing motifs found in similar structures deposited in the Cambridge Structural Database (CSD), giving a positive or negative possibility of cocrystal formation by subtracting the highest homodimeric propensity from the highest heterodimeric propensity [40,41]. The MEP-based method predicts the chance of cocrystal formation from energetic aspects by estimating the energy gain upon all possible intermolecular interactions between the two considered components. The predictable possibility of cocrystal formation between flavonoids and PA identified from computational data was then validated through experimental screening results. Experimentally, 5 flavonoids cannot cocrystallize with PA despite many efforts, and 12 flavonoids were found to cocrystallize with PA, among which, 4 single crystals of PA in the presence of LUT, GEN, TAX, and DHM were obtained for the first time in this paper. The prediction accuracy of HBP and MEP is 58.82% and 94.11%, respectively. The spatial structure, packing mode, and hydrogen bonding interactions were comprehensively elucidated by single-crystal X-ray diffraction (SCXRD). MEP maps of starting components ranking the maximum and minimum sites were also analyzed, which help to provide a feasible route to rationally understand the hydrogen bonding principle and Etter's rule, thus performing cocrystal design more efficiently [42-44]. It was also noted that the more general criterion for MEP would be ΔE smaller than −11 kJ/mol, the probability of the formation of a cocrystal is 50%. Based on our study, we evaluate as -ΔE ≥ 10 kJ/mol. This criterion will be increasing the efficiency helpful to predict, screen and prepare the cocrystals for further examinations. When comprehensively considered the existence status of the APIs and CCFs, MEP maps can be used as an efficient virtual screening tool in pre-screening of coformers. The thermodynamic behaviors, solubility, stability, antioxidant properties based on ABTS⁺ and DPPH^• radical-scavenging capacity, and antihypoxic properties of cardiomyocytes were all evaluated in this work. The experimental section was described in Supporting information.

  Table 1 summarizes the highest homo- and heterodimeric hydrogen-bond propensities and multi-component scores for each system. Table S1 (Supporting information) lists the summary of the outcomes of the performed grinding reactions the presence of three different solvents. For the 5 systems of TAX, DHM, MYR, QUE, KAE with PA, both HBP results and experimental results are positive; for the 5 systems of 7HIF, CHR, BAI, DAI, FOR with PA, either the HBP or experimental result is negative. The HBP method correctly predicted ~58.82% (10 of the total 17) of the experimental multi-component forms.

  Table 1

  

  Table 1.  The Computational and experimental screening results of flavonoids and PA.

  DownLoad: CSV

  No.	System	Highest propensities		Multicomonent score	HBP	-ΔE (kJ/mol)	MEP	Cocrystal observed experimentally
  Empty Cell	Empty Cell	Heterodimer	Homodimer	Empty Cell	Empty Cell	Empty Cell	Empty Cell	Empty Cell
  1	LUT-N, LUT-Z	0.48, 0.61	0.56, 0.78	–0.08, –0.17	No	2, 11	No, Yes	Yes
  2	GEN-N, GEN-Z	0.48, 0.64	0.54, 0.78	–0.06, –0.14	No	2, 14	No, Yes	Yes
  3	TAX-N, TAX-Z	0.62, 0.85	0.51, 0.76	0.11, 0.09	Yes	4, 13	No, Yes	Yes
  4	DHM-N, DHM-Z	0.63, 0.86	0.52, 0.77	0.11, 0.09	Yes	6, 16	No, Yes	Yes
  5	MYR-N, MYR-Z	0.65, 0.81	0.64, 0.77	0.01, 0.04	Yes	7, 10	No, Yes	Yes
  6	QUE-N, QUE-Z	0.62, 0.75	0.51, 0.71	0.11, 0.04	Yes	6, 17	No, Yes	Yes
  7	HES-N, HES-Z	0.44, 0.62	0.52, 0.75	–0.08, –0.13	No	5, 16	No, Yes	Yes
  8	KAE-N, KAE-Z	0.62, 0.80	0.62, 0.76	0.00, 0.04	Yes	5, 14	No, Yes	Yes
  9	NAR-N, NAR-Z	0.48, 0.62	0.55, 0.77	–0.07, –0.15	No	2, 13	No, Yes	Yes
  10	API-N, API-Z	0.49, 0.63	0.55, 0.76	–0.06, –0.13	No	3, 12	No, Yes	Yes
  11	ILG-N, ILG-Z	0.37, 0.50	0.46, 0.79	–0.09, –0.29	No	3, 12	No, Yes	Yes
  12	RES-N, RES-Z	0.50, 0.44	0.55, 0.75	–0.05, –0.31	No	8, 14	No, Yes	Yes
  13	7HIF-N, 7HIF-Z	0.43, 0.60	0.51, 0.74	–0.08, –0.14	No	1, 9	No	No
  14	BAI-N, BAI-Z	0.37, 0.59	0.47, 0.76	–0.10, –0.17	No	3, 8	No	No
  15	DAI-N, DAI-Z	0.47, 0.63	0.52, 0.74	–0.05, –0.11	No	2, 9	No	No
  16	FOR-N, FOR-Z	0.44, 0.62	0.49, 0.72	–0.05, –0.10	No	1, 9	No	No
  17	CHR-N, CHR-Z	0.44, 0.59	0.53, 0.76	0.09, –0.17	Yes, No	2, 12	No, Yes	No

  Contrary to HBP calculations, which account only for the best hydrogen bond donor-acceptor pair, the MEP method considers all possible pairs of electrostatic interactions. Herein, ΔE values of the isolated monomers and that of their assembly were calculated and presented in Table 1. There is a total of 13 copies of -ΔE values greater than 10 for at least one of the states of PA experimentally forming cocrystals with flavonoids, among which, 12 hits were verified by experimental cocrystal screening results. Experiments show fail results in the cocrystallization of 4 flavonoids (7HIF, BAI, DAI, FOR) with PA, -ΔE values of which were also smaller than 10. The MEP method successfully predicted the experimental multi-component form in 94.11% (16 out of 17) of cases. Our test set showed that MEP has a higher successful prediction rate than HBP model.

  HBP results expressed as the highest propensities toward the formation of either heterodimeric or homodimeric hydrogen bond, multi-component score between flavonoids and neutral PA (N) and zwitterionic PA (Z), MEP results expressed as -ΔE values between flavonoids and neutral PA (N) and zwitterionic PA (Z), and experimental cocrystal screening results. “Flavonoid-N” represents the system between flavonoid and neutral PA, while “flavonoid-Z” represents the system between flavonoid and zwitterionic PA.

  A reliable intramolecular proton transfer took place from the carboxyl group to the pyridine N atom within the PA molecule in all the four cocrystals, which gives the zwitterionic state of PA. The bond lengths of the C—O and C=O in the carboxylate group of PA molecule differ by less than 0.02 Å (± 0.002 Å) in these four cocrystal structures. There exists a common intramolecular O—H···O hydrogen bonding interaction between 5–hydroxyl of ring A and adjacent carbonyl O atom on position 4 of ring B within flavonoids in these four cocrystal structures. The overall bonding network in all four cocrystal structures is driven by charge-assisted hydrogen bonds between the neutral hydroxyl groups in flavonoids, and the carboxylate and protonated pyridine moieties of PA. The crystal data and structure refinement parameters of the cocrystals of LUT-PA, GEN-PA, TAX-PA, and DHM-PA were summarized in Table 2. Hydrogen bond geometrical parameters of crystal structures have been provided in Table S2 (Supporting information).

  Table 2

  

  Table 2.  Crystallographic data and structure refinement details.

  DownLoad: CSV

  Empty Cell	LUT-PA	GEN-PA	TAX-PA	DHM-PA
  Formula	C15H10O6·C6H5NO2	C15H10O5·C6H5NO2	C15H12O7·C6H5NO2	C15H12O8·H2O·2(C6H5NO2)
  Formula wt	409.34	393.34	427.36	584.48
  Crystal size (mm)	0.12×0.23×0.39	0.05×0.19×0.34	0.04×0.22×0.39	0.11×0.12×0.43
  Crystal system	Orthorhombic	Monoclinic	Monoclinic	Triclinic
  Space group	Pca21	C2/c	P21/c	P −1
  a (Å)	36.163(1)	12.268(1)	7.370(1)	7.405(1)
  b (Å)	7.161(1)	13.787(1)	18.003(1)	12.287(1)
  c (Å)	6.972(1)	21.449(1)	14.444(1)	15.056(1)
  α (deg)	90	90	90	67.41(1)
  β (deg)	90	106.49(1)	103.42(1)	84.63(1)
  γ (deg)	90	90	90	78.08(1)
  Z	4	8	4	2
  V (Å3)	1805.48(1)	3478.20(12)	1863.93(4)	1237.41(5)
  R1	0.0289	0.0441	0.0551,	0.0417
  wR2	0.0795	0.1199	0.1422	0.1090
  S	1.081	1.026	1.093	1.059
  CCDC No.	2211146	2211147	2211148	2211149

  LUT-PA crystallizes in the orthorhombic Pca2₁ space group, with one molecule each of LUT and PA in the asymmetric unit (Z' = 1, Fig. 2a). There are four hydroxyl groups in positions 5 and 7 of ring A and 3′, 4′ of ring C in the LUT molecule. LUT forms a robust primary (4′–hydroxyl) O—H···O hydrogen bond with the carbonyl O atom of PA in the asymmetric unit (Fig. 2a). There exists a strong heterodimer between charge-assisted PA and LUT: (3′–hydroxyl) O—H···O(PA), (PA) N—H···O(4′–hydroxyl) (Fig. 3a). The LUT molecule propagates through the formation of the (7–hydroxyl) O—H···O(4-carbonyl) hydrogen bond as illustrated in Fig. 3a and extended to form a layered structure viewed down the crystallographic b axis.

  Figure 2

  

  Figure 2.  The packing of the crystal structures (left) and the asymmetric unit (right) of (a) LUT-PA, (b) GEN-PA, (c) TAX-PA, and (d) DHM-PA cocrystals.

  DownLoad: Full-Size Img PowerPoint

  Figure 3

  

  Figure 3.  The hydrogen bond schemes of (a) LUT-PA, (b) GEN-PA, (c) TAX-PA, and (d) DHM-PA cocrystals.

  DownLoad: Full-Size Img PowerPoint

  GEN-PA crystallizes the monoclinic C2/c space group (Z = 8), with each asymmetric unit containing one molecule each of LUT and PA (Z' = 1, Fig. 2b). GEN contains three hydrogen bond donor sites on positions 5, 7 of ring A and 4′ of ring C. GEN forms an intermolecular (4′–hydroxyl) O—H···O=C hydrogen bond with the carbonyl group of PA in the asymmetric unit (Fig. 2b). The asymmetric units interacted via a self-assembled dimer generated by PA molecules through two symmetrically equivalent charged-assisted N—H···O hydrogen bonds between the carboxylate O anion and protonated pyridine N—H cation, thus creating a centrosymmetric tetrameric motif, which further extended into a sheet through (7–hydroxyl) O—H···O(4′–hydroxyl) hydrogen bonds between two adjacent GEN molecules, with PA molecules sandwiched between the double layers of GEN (Fig. 3b).

  TAX-PA crystallizes the monoclinic P2₁/c space group (Z = 4) with one TAX and one PA in the asymmetric unit (Z' = 1, Fig. 2c). The O₅ oxygen atom is disordered over two positions showing 50% occupancy each in the structure of TAX-PA. (7–hydroxyl) O—H···O (carbonyl) hydrogen-bonding interaction involving TAX and PA appears as a main supramolecular heterosynthon (Fig. 2c). The crystal structure features a cyclic four-membered unit containing every two molecules of TAX and PA sustained by an O—H···O hydrogen bond involving the 4′–hydroxyl of TAX and carboxylate O anion of PA (Fig. 3c). One part of the disordered component has been omitted in Fig. 3c for clarity.

  DHM-PA crystallizes the triclinic P −1 space group (Z = 2) with the asymmetric unit consisting of one molecule of DHM, two molecules of PA, and one water molecule (Z' = 1, Fig. 2d). The asymmetric unit cell showed an unexpected presence of one water molecule, which plays the bridging role of connecting two PA molecules, strengthening the hydrogen bond network. In the asymmetric unit, DHM forms the intermolecular (7–hydroxyl) O—H···O hydrogen bonding interaction with the carboxylate group of PA. Water molecules, located between two PA molecules, connect with one PA molecule via N—H···O hydrogen bond and O—H···O hydrogen bond with the carboxylate group of another PA (Fig. 2d). Neighboring two asymmetric units are linked via O—H···O hydrogen bonding interaction between the 7–hydroxyl and 5′–hydroxyl of adjacent two DHM molecules, together with O—H···O hydrogen bond involving the 4′–hydroxyl of DHM and carbonyl O atom of PA molecule (Fig. 3d).

  In this work, the maxima, and minima values from studies of MEP for LUT, GEN, TAX, DHM, and PA are implemented and presented in Fig. 4. The calculation method details were added in the supplement information. The positive or negative extreme values represent the hydrogen bonding donating and accepting abilities, respectively. The positive electrostatic potential region is illustrated in red, while the negative electrostatic potential area is marked in dark blue. As evidenced by the MEP map of neutral PA, there is a good proton donor (MEP of carboxylic acid H: +53.03 kcal/mol) and a good proton acceptor (MEP of pyridine N: −42.27 kcal/mol). The zwitterionic state of PA could act as a good acceptor owing to the global minima site of the carboxylate group (−68.93 kcal/mol), and the positive extreme site located near the N—H⁺cation (+54.37 kcal/mol). In the cases of these four flavonoids, the global maxima site is located in the region of the 7–hydroxyl group of ring A in LUT, GEN, and TAX, while in DHM it is located in the 5′–hydroxyl group of ring C and the value in the region of 7–hydroxyl group is just slightly smaller than that of 5′–hydroxyl. The MEP in the regions of the 3′–hydroxyl and 4′–hydroxyl of ring B was weaker and the MEP near the 5–hydroxyl group of ring A was the weakest. The dominating interactions that drive the formation of the four flavonoid cocrystals with zwitterionic PA basically but do not strictly obey Etter's rule that the main interaction sites in the cocrystal should first occur pairwise in the minima and maxima of the MEP, followed by the secondary ones. Interactions sometimes occur between the second best and best H-bond donor and acceptors, which is probably ascribed to steric hindrance.

  Figure 4

  

  Figure 4.  The positive and negative values (kcal/mol) from MEP of PA (neutral and zwitterionic state) and LUT, GEN, TAX, and DHM.

  DownLoad: Full-Size Img PowerPoint

  Herein, the results reveal the novelty of the synthesized cocrystals because their powder X-ray diffraction (PXRD) patterns are observed to be different from either of the starting materials. Particularly, the peaks observed from the experimental PXRD modes are consistent with those simulated results calculated from the SCXRD data, suggesting the formation of highly pure products (Fig. 5). Experimental cocrystal PXRD patterns of QUE, MYR, HES, KAE, NAR, ILG, RES and API with PA have been presented in Fig. S3 (Supporting information).

  Figure 5

  

  Figure 5.  Experimental and calculated PXRD patterns of (a) LUT-PA, (b) GEN-PA, (c) TAX-PA, (d) DHM-PA cocrystals, and raw materials of the four flavonoids and PA.

  DownLoad: Full-Size Img PowerPoint

  Thermo-analytical methods are frequently used to probe the thermodynamic stability and other thermal properties of solid forms and have served as an effective tool for detecting the formation of novel phases. Particularly, the differential scanning calorimetry (DSC) technique provides a fast and easy indication by comparing the DSC curve of the parent materials and the synthesized assemblies. DSC and Thermogravimetric (TG) analyses were carried out to investigate the thermal behaviors of the four flavonoid cocrystals with PA (Fig. 6). The TG thermograms of LUT-PA, GEN-PA, and TAX-PA cocrystals demonstrate that there is no absorption of water or solvent molecules in these compounds, as there was no loss of mass before the decomposition event. For LUT-PA, two endothermic melting peaks at 215.74 ℃ and 335.53 ℃ are ascribed to the decomposition process of PA and LUT, respectively, supported by the mass loss in the range of 179–240 ℃ (30.20%) and 320–440 ℃ (17.60%) in the TG curve. In the DSC thermogram of GEN–PA, two endothermic melting peaks at 211.98 ℃ and 300.03 ℃ are ascribed to the decomposition of PA and GEN, respectively, distinguished from the single broad endothermic peak at 301.57 ℃, which suggests the formation of a novel phase. There is a broad endothermic peak at 128.21 ℃ in TAX, which can be ascribed to the dehydration process. TAX-PA exhibits quite different thermal properties in comparison with TAX and PA pure material, which displays a single sharp endothermic melting peak at 217.6 ℃, accompanied by the decomposition process in the temperature range of 194−340 ℃ (39.32%). The DSC analysis of DHM-PA (1:2) monohydrate revealed that the dehydration process occurred with endothermic peaks of 146.66 ℃. From the TGA curve, it was found that DHM-PA (1:2) monohydrate dehydrates between 107 ℃ and 148 ℃ and a total weight loss of 2.96% (w/w, calculated value: 3.08%) was equivalent to one molecule of water, which is in good agreement with the SCXRD data. The exothermic peaks at 158.73 ℃ and 229.49 ℃ are corresponding to the decomposition temperature of the dehydrated form of DHM-PA (1:2), which is confirmed by the weight loss of 50.98% in the temperature range of 148–330 ℃ in the TG curve.

  Figure 6

  

  Figure 6.  (a) DSC thermograms for LUT-PA, GEN-PA, TAX-PA, DHM-PA, and the raw materials of corresponding API and PA. (b) TG thermograms for LUT-PA, GEN-PA, TAX-PA, DHM-PA.

  DownLoad: Full-Size Img PowerPoint

  The powder dissolution rate is an important factor to be considered in pharmaceutical development. The application of flavonoids was limited accounting for low solubility and poor bioavailability in pharmaceutical fields. Particularly, the dissolution rate of GEN-PA shows a significant enhancement in comparison with GEN (Fig. 7), and the equilibrium solubility of GEN-PA (7.73 µg/mL) exhibits approximately 89% higher than that of GEN (4.09 µg/mL) in ultrapure water. Nevertheless, it can be found that LUT-PA, TAX-PA, and DHM-PA-H₂O cocrystals exhibit comparably even reduced dissolution behaviors compared to the corresponding starting flavonoids.

  Figure 7

  

  Figure 7.  Powder dissolution profiles of four cocrystals of flavonoids in pure water. (a) GEN and GEN-PA, (b) LUT-PA, TAX-PA, DHM-PA, and the raw materials of corresponding APIs.

  DownLoad: Full-Size Img PowerPoint

  To evaluate whether the antioxidant property of flavonoids has been improved through cocrystallization, the DPPH^• and ABTS⁺ free radical-scavenging assays of four cocrystals and pure flavonoids were carried out (Fig. 8). GEN-PA (16.73% of DPPH^• inhibition), TAX-PA (75.70%), and DHM-PA (82.81%) showed a higher DPPH^• free radical-scavenging capacity than pure GEN (−1.24%), TAX (42.43%), and DHM (47.68%). LUT-PA (33.25% of ABTS⁺ inhibition) and DHM-PA (80.36%) showed higher ABTS⁺ radical-scavenging capacities than LUT (14.78%) and DHM (55.24%), respectively.

  Figure 8

  

  Figure 8.  Antioxidant studies for LUT-PA, GEN-PA, TAX-PA, DHM-PA, and the raw materials of corresponding APIs. (a) DPPH^• scavenging capacity (%), (b) ABTS⁺ scavenging capacity (%).

  DownLoad: Full-Size Img PowerPoint

  To assess whether cocrystallization of flavonoids has enhanced the antihypoxic property of cardiomyocytes, CoCl₂ hypoxia was performed to suppress H9c2 growth in vitro, and the treatment drugs were selected as GEN and DHM, which have protective effects against hypoxic injury in cardiomyocytes. We first explored the optimal concentrations of the drugs. As shown in Figs. 9a and b, CCK-8 results showed that GEN-PA (0.1, 0.3, 1, 3, 10 µmol/L) and DHM-PA-H₂O (0.03, 0.1, 0.3, 1, 3 µmol/L) were non-toxic to H9c2 cells for 24 h. The cell survival rates in GEN-PA (0.1, 0.3, 1, 3, 10 µmol/L) and DHM-PA-H₂O (0.03, 0.1, 0.3, 1, 3 µmol/L) groups were higher than the raw materials of corresponding APIs (Fig. 9, Fig. 9). Among them, concentrations of 0.1–10 µmol/L GEN and 0.1–1 µmol/L GEN-PA had no effect on cell viability, whereas concentrations of 3 and 10 µmol/L GEN-PA had better protective effects on H9c2 cells from hypoxic injury caused by CoCl₂, and the hypoxia rescue capacity of 1, 3 and 10 µmol/L GEN-PA were 8.09%, 14.70% and 22.77% respectively, higher than the corresponding APIs groups significantly. In the DHM groups, most of the concentrations of DHM and DHM-PA could alleviate hypoxic damage in H9c2 cells, and the hypoxia rescue capacity of 0.1 and 1 µmol/L GEN-PA were 32.36% and 37.74% respectively, higher than the corresponding APIs groups significantly.

  Figure 9

  

  Figure 9.  Antihypoxic studies for GEN-PA, DHM-PA, and the raw materials of corresponding APIs. (a, b) Toxicity studies. (c, d) Pharmacodynamic studies. ^#P < 0.05, ^##P < 0.01, ^###P < 0.001, compared with the model group (without sample and with CoCl₂). P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, compared with the raw materials of corresponding APIs at the same concentration (n = 3–5).

  DownLoad: Full-Size Img PowerPoint

  In conclusion, in order to extend our understanding of the synergistic association of bipolar molecules with flavonoids, PA is selected as the CCF to cocrystallize with 17 flavonoids computationally and experimentally. The MEP prediction data nicely match with experimental results. 12 cocrystals were formed experimentally, and 4 single crystals were successfully obtained for the first time. To our knowledge, this is the first zwitterionic cocrystal of GEN, TAX, and DHM. It is worth noting that the charged-assisted hydroxyl···carboxylate hydrogen bond formed between the hydroxyl groups of flavonoids and the carboxylate part of zwitterionic PA plays the dominating role in the crystalline network of the four zwitterionic flavonoid cocrystals.

  GEN-PA shows a significant enhancement of both dissolution rate and equilibrium solubility in comparison with GEN in ultrapure water. GEN-PA, TAX-PA, and DHM-PA showed a higher DPPH^• free radical-scavenging capacity; LUT-PA and DHM-PA showed higher ABTS⁺ radical-scavenging capacities; GEN-PA and DHM-PA showed better protective effects on H9c2 cells from hypoxic injury caused by CoCl₂ than corresponding pure flavonoids. This work not only provides four new crystalline forms of flavonoids but also proves the realizability of reasonable cocrystal prediction by employing MEP-based virtual cocrystal screening approach.

  Declaration of competing interest

  The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

  CRediT authorship contribution statement

  Yao Zou: Writing – original draft, Methodology, Investigation. Difei Gong: Writing – original draft, Validation, Methodology, Investigation. Haiguang Yang: Methodology, Investigation. Hongmei Yu: Validation, Methodology, Investigation. Guorong He: Data curation. Ningbo Gong: Writing – review & editing, Supervision, Project administration, Funding acquisition, Conceptualization. Lianhua Fang: Writing – review & editing, Supervision, Project administration, Formal analysis, Conceptualization. Guanhua Du: Supervision, Project administration, Conceptualization. Yang Lu: Writing – review & editing, Supervision, Project administration, Funding acquisition, Conceptualization.

  Acknowledgments

  This work was supported by the Beijing Natural Science Foundation (No. 7222261), CAMS Innovation Fund for Medical Sciences (No. 2022-I2M-1-015).

  Supplementary materials

  Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.cclet.2024.110768.

  
  
• 1. [1]

     M. Fardoun, D. Maaliki, N. Halabi, et al., Clin. Sci. 134 (2020) 1403–1432. doi: 10.1042/cs20200356

  2. [2]

     J. Melrose, Antioxidants 12 (2023) 663. doi: 10.3390/antiox12030663

  3. [3]

     K.R. Surana, E.D. Ahire, S.K. Mahajan, et al., Antimicrobial and antiinflammatory action of flavonoids, in: J.B. Harborne, T.J. Mabry, H. Mabry (Eds.), The Flavonoids, Apple Academic Press, Palm Bay Florida, 2024, pp. 263–276.

  4. [4]

     Z. Wang, L. Yang, Nutrients 15 (2023) 3443. doi: 10.3390/nu15153443

  5. [5]

     M. Park, D. Ryu, J. Cho, et al., Molecules 29 (2024) 283. doi: 10.3390/molecules29020283

  6. [6]

     J. Sharifi-Rad, C. Quispe, S. Shaheen, et al., Crit. Rev. Food Sci. Nutr. 62 (2022) 8045–8058. doi: 10.1080/10408398.2021.1924612

  7. [7]

     V. Kesharwani, S. Kabra, B.C. Semwal, et al., Neuroprotective effects of flavonoids, in: N. Sharma, D. Saini, R.K. Kesharani, et al. (Eds.), Advances in Flavonoids for Human Health and Prevention of Diseases, Apple Academic Press, Palm Bay Florida, 2024, pp. 95–123.

  8. [8]

     Y. Wang, Y. Mou, S. Lu, et al., PeerJ 12 (2024) e16711. doi: 10.7717/peerj.16711

  9. [9]

     T.P. Lam, N.V.N. Tran, L.H.D. Pham, et al., Nat. Prod. Bioprospect. 14 (2024) 4. doi: 10.1007/s13659-023-00424-w

  10. [10]

      D. Yuan, Y. Guo, F. Pu, et al., Food Chem. 430 (2023) 137115.

  11. [11]

      J. Zhao, J. Yang, Y. Xie, Int. J. Pharm. 570 (2019) 118642. doi: 10.1016/j.ijpharm.2019.118642

  12. [12]

      E. Ban, S.H. An, B. Park, et al., J. Pharm. Sci. 109 (2020) 3660–3667. doi: 10.1016/j.xphs.2020.09.030

  13. [13]

      G. Bolla, A. Nangia, Chem. Commun. 52 (2016) 8342–8360. doi: 10.1039/C6CC02943D

  14. [14]

      B. Lu, S. Liu, D. Yan, Chin. Chem. Lett. 30 (2019) 1908–1922. doi: 10.1016/j.cclet.2019.09.012

  15. [15]

      S. Li, Y. Lin, D. Yan, J. Mater. Chem. C 4 (2016) 2527–2534. doi: 10.1039/C6TC00067C

  16. [16]

      C. Xing, B. Zhou, D. Yan, et al., CCS Chem. 5 (2023) 2866–2876. doi: 10.31635/ccschem.023.202202605

  17. [17]

      C. Xing, Z. Qi, B. Zhou, et al., Angew. Chem. Int. Ed. 63 (2024) e202402634. doi: 10.1002/anie.202402634

  18. [18]

      D. Yan, A. Delori, G.O. Lloyd, et al., Angew. Chem. Int. Ed. 50 (2011) 12483–12486. doi: 10.1002/anie.201106391

  19. [19]

      D. Yan, Chem. Eur. J. 21 (2015) 4880–4896. doi: 10.1002/chem.201405456

  20. [20]

      S. Li, B. Lu, X. Fang, et al., Angew. Chem. Int. Ed. 59 (2020) 22623–22630. doi: 10.1002/anie.202009714

  21. [21]

      X.G. Yang, Z.M. Zhai, X.M. Lu, et al., ACS Cent. Sci. 6 (2020) 1169–1178. doi: 10.1021/acscentsci.0c00447

  22. [22]

      J. Xu, Q. Shi, Y. Wang, et al., Molecules 28 (2023) 613. doi: 10.3390/molecules28020613

  23. [23]

      X. Jia, H. Hao, Q. Zhang, et al., Phytomedicine 122 (2024) 155179. doi: 10.1016/j.phymed.2023.155179

  24. [24]

      M.K. Bommaka, M.K.C. Mannava, K. Suresh, et al., Cryst. Growth Des. 18 (2018) 6061–6069. doi: 10.1021/acs.cgd.8b00921

  25. [25]

      C. Almansa, R. Merce, N. Tesson, et al., Cryst. Growth Des. 17 (2017) 1884–1892. doi: 10.1021/acs.cgd.6b01848

  26. [26]

      Y. Deng, Y. Zhang, Y. Huang, et al., Cryst. Growth Des. 18 (2018) 7481–7488. doi: 10.1021/acs.cgd.8b01257

  27. [27]

      N.K. Duggirala, M.L. Perry, Ö. Almarsson, et al., Chem. Commun. 52 (2016) 640–655. doi: 10.1039/C5CC08216A

  28. [28]

      J. Yao, J.M. Chen, Y.B. Xu, et al., Cryst. Growth Des. 14 (2014) 5019–5025. doi: 10.1021/cg5005819

  29. [29]

      A. Tilborg, G. Springuel, B. Norberg, et al., CrystEngComm 15 (2013) 3341–3350. doi: 10.1039/c3ce40084k

  30. [30]

      L.L. Wang, L.Y. Wang, Y.M. Yu, et al., CrystEngComm 22 (2020) 5010–5021. doi: 10.1039/D0CE00713G

  31. [31]

      A. Gunnam, K. Suresh, R. Ganduri, et al., Chem. Commun. 52 (2016) 12610–12613. doi: 10.1039/C6CC06627E

  32. [32]

      D. Chen, T. Liu, J. Kang, et al., Adv. Mater. Interfaces 6 (2019) 1900706. doi: 10.1002/admi.201900706

  33. [33]

      H. He, Y. Huang, Q. Zhang, et al., Cryst. Growth Des. 16 (2016) 2348–2356. doi: 10.1021/acs.cgd.6b00142

  34. [34]

      L.Y. Wang, Y.M. Yu, F.B. Jiang, et al., New J. Chem. 44 (2020) 3930–3939. doi: 10.1039/c9nj06180k

  35. [35]

      M. Davidson, N. Rashidi, M.K. Hossain, et al., Int. J. Mol. Sci. 24 (2023) 2737. doi: 10.3390/ijms24032737

  36. [36]

      F. Racher, T.L. Petrick, D.E. Braun, Cryst. Growth Des. 23 (2023) 4638–4654. doi: 10.1021/acs.cgd.3c00387

  37. [37]

      J. Makadia, C.C. Seaton, M. Li, Cryst. Growth Des. 23 (2023) 3480–3495. doi: 10.1021/acs.cgd.3c00030

  38. [38]

      Y. Deng, S. Liu, Y. Jiang, et al., Pharmaceutics 15 (2023) 2174. doi: 10.3390/pharmaceutics15092174

  39. [39]

      D. Musumeci, C.A. Hunter, R. Prohens, et al., Chem. Sci. 2 (2011) 883–890. doi: 10.1039/c0sc00555j

  40. [40]

      H. Yu, B. Zhang, M. Liu, et al., Chin. Chem. Lett. 35 (2024) 109032. doi: 10.1016/j.cclet.2023.109032

  41. [41]

      A.O. Surov, A.G. Ramazanova, A.P. Voronin, et al., Pharmaceutics 15 (2023) 836. doi: 10.3390/pharmaceutics15030836

  42. [42]

      D. Yang, J. Cao, T. Heng, et al., Cryst. Growth Des. 21 (2021) 2292–2300. doi: 10.1021/acs.cgd.0c01706

  43. [43]

      M. Borovina, I. Kodrin, M. Đaković, Cryst. Growth Des. 19 (2019) 1985–1995. doi: 10.1021/acs.cgd.8b01930

  44. [44]

      B. Sandhu, S.A. Savage, H. Park, et al., Cryst. Growth Des. 22 (2022) 4105–4114. doi: 10.1021/acs.cgd.1c01486

• 

  Figure 1  The chemical structures of the 17 flavonoids and picolinic acid (neutral and zwitterionic state) used in this study.

  下载: 全尺寸图片 幻灯片
  

  Figure 2  The packing of the crystal structures (left) and the asymmetric unit (right) of (a) LUT-PA, (b) GEN-PA, (c) TAX-PA, and (d) DHM-PA cocrystals.

  下载: 全尺寸图片 幻灯片
  

  Figure 3  The hydrogen bond schemes of (a) LUT-PA, (b) GEN-PA, (c) TAX-PA, and (d) DHM-PA cocrystals.

  下载: 全尺寸图片 幻灯片
  

  Figure 4  The positive and negative values (kcal/mol) from MEP of PA (neutral and zwitterionic state) and LUT, GEN, TAX, and DHM.

  下载: 全尺寸图片 幻灯片
  

  Figure 5  Experimental and calculated PXRD patterns of (a) LUT-PA, (b) GEN-PA, (c) TAX-PA, (d) DHM-PA cocrystals, and raw materials of the four flavonoids and PA.

  下载: 全尺寸图片 幻灯片
  

  Figure 6  (a) DSC thermograms for LUT-PA, GEN-PA, TAX-PA, DHM-PA, and the raw materials of corresponding API and PA. (b) TG thermograms for LUT-PA, GEN-PA, TAX-PA, DHM-PA.

  下载: 全尺寸图片 幻灯片
  

  Figure 7  Powder dissolution profiles of four cocrystals of flavonoids in pure water. (a) GEN and GEN-PA, (b) LUT-PA, TAX-PA, DHM-PA, and the raw materials of corresponding APIs.

  下载: 全尺寸图片 幻灯片
  

  Figure 8  Antioxidant studies for LUT-PA, GEN-PA, TAX-PA, DHM-PA, and the raw materials of corresponding APIs. (a) DPPH^• scavenging capacity (%), (b) ABTS⁺ scavenging capacity (%).

  下载: 全尺寸图片 幻灯片
  

  Figure 9  Antihypoxic studies for GEN-PA, DHM-PA, and the raw materials of corresponding APIs. (a, b) Toxicity studies. (c, d) Pharmacodynamic studies. ^#P < 0.05, ^##P < 0.01, ^###P < 0.001, compared with the model group (without sample and with CoCl₂). P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, compared with the raw materials of corresponding APIs at the same concentration (n = 3–5).

  下载: 全尺寸图片 幻灯片

  Table 1.  The Computational and experimental screening results of flavonoids and PA.

  No.	System	Highest propensities		Multicomonent score	HBP	-ΔE (kJ/mol)	MEP	Cocrystal observed experimentally
  Empty Cell	Empty Cell	Heterodimer	Homodimer	Empty Cell	Empty Cell	Empty Cell	Empty Cell	Empty Cell
  1	LUT-N, LUT-Z	0.48, 0.61	0.56, 0.78	–0.08, –0.17	No	2, 11	No, Yes	Yes
  2	GEN-N, GEN-Z	0.48, 0.64	0.54, 0.78	–0.06, –0.14	No	2, 14	No, Yes	Yes
  3	TAX-N, TAX-Z	0.62, 0.85	0.51, 0.76	0.11, 0.09	Yes	4, 13	No, Yes	Yes
  4	DHM-N, DHM-Z	0.63, 0.86	0.52, 0.77	0.11, 0.09	Yes	6, 16	No, Yes	Yes
  5	MYR-N, MYR-Z	0.65, 0.81	0.64, 0.77	0.01, 0.04	Yes	7, 10	No, Yes	Yes
  6	QUE-N, QUE-Z	0.62, 0.75	0.51, 0.71	0.11, 0.04	Yes	6, 17	No, Yes	Yes
  7	HES-N, HES-Z	0.44, 0.62	0.52, 0.75	–0.08, –0.13	No	5, 16	No, Yes	Yes
  8	KAE-N, KAE-Z	0.62, 0.80	0.62, 0.76	0.00, 0.04	Yes	5, 14	No, Yes	Yes
  9	NAR-N, NAR-Z	0.48, 0.62	0.55, 0.77	–0.07, –0.15	No	2, 13	No, Yes	Yes
  10	API-N, API-Z	0.49, 0.63	0.55, 0.76	–0.06, –0.13	No	3, 12	No, Yes	Yes
  11	ILG-N, ILG-Z	0.37, 0.50	0.46, 0.79	–0.09, –0.29	No	3, 12	No, Yes	Yes
  12	RES-N, RES-Z	0.50, 0.44	0.55, 0.75	–0.05, –0.31	No	8, 14	No, Yes	Yes
  13	7HIF-N, 7HIF-Z	0.43, 0.60	0.51, 0.74	–0.08, –0.14	No	1, 9	No	No
  14	BAI-N, BAI-Z	0.37, 0.59	0.47, 0.76	–0.10, –0.17	No	3, 8	No	No
  15	DAI-N, DAI-Z	0.47, 0.63	0.52, 0.74	–0.05, –0.11	No	2, 9	No	No
  16	FOR-N, FOR-Z	0.44, 0.62	0.49, 0.72	–0.05, –0.10	No	1, 9	No	No
  17	CHR-N, CHR-Z	0.44, 0.59	0.53, 0.76	0.09, –0.17	Yes, No	2, 12	No, Yes	No

  下载: 导出CSV

  Table 2.  Crystallographic data and structure refinement details.

  Empty Cell	LUT-PA	GEN-PA	TAX-PA	DHM-PA
  Formula	C15H10O6·C6H5NO2	C15H10O5·C6H5NO2	C15H12O7·C6H5NO2	C15H12O8·H2O·2(C6H5NO2)
  Formula wt	409.34	393.34	427.36	584.48
  Crystal size (mm)	0.12×0.23×0.39	0.05×0.19×0.34	0.04×0.22×0.39	0.11×0.12×0.43
  Crystal system	Orthorhombic	Monoclinic	Monoclinic	Triclinic
  Space group	Pca21	C2/c	P21/c	P −1
  a (Å)	36.163(1)	12.268(1)	7.370(1)	7.405(1)
  b (Å)	7.161(1)	13.787(1)	18.003(1)	12.287(1)
  c (Å)	6.972(1)	21.449(1)	14.444(1)	15.056(1)
  α (deg)	90	90	90	67.41(1)
  β (deg)	90	106.49(1)	103.42(1)	84.63(1)
  γ (deg)	90	90	90	78.08(1)
  Z	4	8	4	2
  V (Å3)	1805.48(1)	3478.20(12)	1863.93(4)	1237.41(5)
  R1	0.0289	0.0441	0.0551,	0.0417
  wR2	0.0795	0.1199	0.1422	0.1090
  S	1.081	1.026	1.093	1.059
  CCDC No.	2211146	2211147	2211148	2211149

  下载: 导出CSV
•