Exploring Verrucosidin Derivatives with Glucose-Uptake-Stimulatory Activity from Penicillium cellarum Using MS/MS-Based Molecular Networking

Under the guidance of LC-MS/MS-based molecular networking, seven new verrucosidin derivatives, penicicellarusins A-G (3–9), were isolated together with three known analogues from the fungus Penicillium cellarum. The structures of the new compounds were determined by a combination of NMR, mass and electronic circular dichroism spectral data analysis. The absolute configuration of penicyrone A (10) was corrected based on X-ray diffraction analyses. Bioactivity screening indicated that compounds 1, 2, and 4 showed much stronger promising hypoglycemic activity than the positive drug (rosiglitazone) in the range of 25–100 μM, which represents a potential new class of hypoglycemic agents. Preliminary structure-activity relationship analysis indicates that the formation of epoxy ring on C6-C7 in the structures is important for the glucose uptake-stimulating activity. The gene cluster for the biosynthesis of 1–12 is identified by sequencing the genome of P. cellarum and similarity analysis with the gene cluster of verrucosidins in P. polonicum.


Introduction
Fungi have attracted much attention of chemists and biologists due to their potential in producing bioactive secondary metabolites with diverse chemical skeletons [1,2]. Verrucosidins produced by Penicillium strains belong to a family of highly reducing fungal polyketides that are characterized with 2H-pyran-2-one and dicyclic fused 3,6dioxabicyclo[3.1.0]hexane moieties interlinked by a polyene chain [3][4][5][6]. They have been reported to display important bioactivities, such as antitumor [7,8], antivirus [9], antibacterial [3,10], and neurological activities [11]. In order to explore in depth this kind of compounds with unique chemical structure and diverse biological activities, we explored Penicillium strains collected in our lab searching for verrucosidin analogues.
Molecular networking analyses include acquisition and similarity comparison of mass spectral fragment data, cluster grouping and visualization [12,13]. More recently, the MS/MS-based molecular networking has been demonstrated to be powerful in dereplicating known natural products from a targeted extract and searching for new analogues with with the specific skeleton.Examples included thermoactinoamide A with moderate antiproliferative activity from Thermoactinomyces vulgaris DSM 43016 [14], suffranidines A-C with significant neuritogenic activity from Flueggea suffruticosa [15], and trilliumoside D with strong cytotoxicity against MOLT-4 cell lines from Trillium tschonoskii maxim [16]. To explore new reducing fungal polyketides from fungi, we applied the LC-MS/MS-based molecular networking for new verrucosidins from Penicillium strains using deoxyverrucosidin that was deposited in our compound library as the probing agent.
The EtOAc extracts of Penicillium strains fermented on solid culture were first analyzed by high performance liquid chromatography (HPLC) with UV diode array detection (DAD) to find fungi potentially producing verrucosidin derivatives ( Figure S1). In this work, the target isolation was further conducted on the selected fungus P. cellarum YM1 under the guidance of LC-MS/MS-based molecular networking ( Figure S2). As a result, seven new verrucosidins, penicicellarusins A-G (3)(4)(5)(6)(7)(8)(9), as well as five known verrucosidins (compounds 1, 2 and 10-12) were identified from the culture of P. cellarum YM1 ( Figure  1). The isolated compounds were evaluated for anti-bacterial effect, cytotoxicity, and glucose uptake-stimulating activities. This work described the details of the isolation, structure elucidation, and biological activities of the isolated secondary metabolites from P. cellarum YM1.

Fungal Material
The strain Penicillium sp. YM1 used in this work was isolated from mildewed corn, collected in China, in September 2017. The sequences of RPB2 (MT898427), Ben A (MT898428), and CaM (MT898429) of our fungus were deposited in GenBank and employed for phylogenetic analysis. The fungus is similar to P. cellarum in forming hyaline, roughened stipes with bearing terminal terverticillate penicillii; and producing typically two rami per stipe, which are usually hyaline, roughened, appressed or only narrowly divergent; and having four to five metulae typically per ramus, which are usually hyaline, roughened, appressed or only narrowly divergent as well; and producing typically six to eight per metula phialides, which are usually hyaline, smooth, ampulliform, slender; and with pale green conidia that were typically smooth, globose to sometimes subglobose [17,18]. The phylogenetic analyses based on a combined dataset of RPB2, Ben A, and CaM was conducted by using PhyML v.3.0, with 1000 bootstrap replicates presented that our taxon grouped with the other taxa of P. cellarum with strongly maximum likelihood bootstrap proportions value ( Figure S3). In consideration of the morphological features and phylogeny, this fungus was identified as P. cellarum YM1.

Fermentation and Extraction
P. cellarum was cultured on slant of PDA at 28 • C for 10 days. To prepare inoculum, the spores of the strain on the plate were collected with 0.01% sterile solution of Tween 80 (BTL, Warsaw, Poland) and adjusted to 1 × 10 6 CFU/mL. A large-scale fermentation was done in 40 × 500 mL Fernbach culture flasks containing 80 g of rice in 110 mL of distilled water (each with 0.5 mL of spore suspension) and incubated at 28 • C for 3 weeks. The fermented rice substrates were extracted with EtOAc (3 × 4 L) with the aid of ultrasonication, and the organic solvent was filtered and evaporated to dryness under vacuum to afford the crude extract (33.7 g).

Alkaline Hydrolysis of Compound 8 and 9
Alkaline hydrolysis reaction was carried out following a previously described method [20]. Each compound (2.0 mg) was dissolved and hydrolyzed with 2 M NaOH/MeOH at 25 • C for 3 h. Then neutralized with 1 N HCl/MeOH and extracted with chloroform for two times (10 mL × 2). Methyl esters of the fatty acids were identified by GC-MS. The GC-MS was operated in EI mode (70 eV) scanning from 40 to 500 amu.

Bioinformatic Analyses
To identify Biosynthetic Gene Clusters (BGCs) in the genomes of P. cellarum YM1, antiSMASH 6.2 was used and only clusters containing a putative PKS similar to both VerA and CtvA protein were further considered [21,22]. The proteins in these clusters were additionally blasted against P. polonicum and Aspergillus terreus var. aureus to verify their presence. To find functional domains and predict a putative function, we resort to NCBI BLAST using Non-Redundant database and Interproscan.

Computation Section
Systematic conformational analyses for 5a, 5b, 5c and 5d were performed using the CONFLEX softwre (version 7 Rev. A; CONFLEX Corporation, Tokyo, Japan) via the MMFF94 molecular mechanics force field. Using TDDFT at B3LYP/6-31+G(d,p) basis set level, the MMFF94 conformers were further optimized in methanol with PCM model. The stationary points have been checked as the true minima of the potential energy surface by verifying that they do not exhibit vibrational imaginary frequencies. ECD spectra were calculated by TD-DFT using a Gaussian function at the PBE1PBE/6-311G* level. Using Boltzmann statistics, equilibrium populations of conformers at 298.15 K were calculated from their relative free energies (∆G). According to Boltzmann weighting of main conformers, the overall ECD spectra were then generated [23]. , and Aspergilus fumigatus (CGMCC 3.5835) were carried out as previously described method [24]. The inhibition rate was calculated and plotted versus test concentrations to afford the MIC. MIC values were defined as the minimum concentration of compounds that inhibited visible microbial growth. All the experiments were performed in triplicate.

Cytotoxicity Assay
Cytotoxicity test against A549, HepG2, and K562 cell lines was carried out as previously described method [25]. Taxol, 5-flourouracil, and cisplatin were used as the positive controls.

2-[N-(7-Nitrobenz-2-oxa-1,3-diazol-4-yl)amino]-2-deoxy-D-glucose (2-NBDG) Glucose Uptake Assay
This experiment was consistent with those reported in our previous work [26]. The HepG2 hepatoma cells were cultured in DMEM supplemented with 10% fetal bovine serum (FBS; Gibco, NY, USA), 100 U/mL penicillin/streptomycin. The cells reaching confluence were treated with 10 −6 M insulin for 24 h to generate insulin resistance. Compounds or positive drug (rosiglitazone) were mixed and incubated for 24 h, with the final concentration of 100, 50, 25, and 12.5 µM; then, 100 nM of insulin was added and incubated for 30 min at 37 • C followed by addition of 50 µM (2-NBDG). After that, cells were washed with ice-cold PBS and 100 µL FBS-free DMEM was added to each well. The level of 2-NBDG uptake was determined on microplate reader (Bio-Tek Instruments, VT, USA) at 485 nm excitation and 528 nm emission. All data were handled with GraphPad Prism 5 and reported as mean ± SD of three independent experiments.

Results
In this study, the MS/MS-based molecular networking strategy was applied for target isolation of new verrucosidins. First, Penicillium strains were cultured on rice substrates and the resulting EtOAc extracts was screened by HPLC-UV-DAD analysis ( Figure S1). Then, the ethyl acetate extract of P. cellarum YM1 that produced secondary metabolites with similar retention time and UV characteristics to those of deoxyverrucosidin was further investigated by UPLC-HRMS/MS. The LC-MS/MS data were used to generate a visualized molecular networking that was further annotated by Cytoscape 3.8.2 ( Figure S2).
In details, the HPLC-HRMS/MS analysis in the positive ion mode was conducted on the ethyl acetate extract from P. cellarum YM1 with deoxyverrucosidin as the phishing probe. The obtained fragmentation data were organized by molecular networking, yielding a metabolite-level view of the data. Individual MS/MS spectrum was organized into 106 clusters consisting of 899 connected nodes ( Figure S2). Using the MS/MS data of deoxyverrucosidin as "seed" spectra, an initial focal point (a blue hexagon with m/z 401.232) was generated in the global molecular networking. A close examination of the molecular network indicated some nodes connected to deoxyverrucosidin (Figure 2), which predicted the presence of potential natural analogs. Under the guidance of MS/MS-based molecular networkings, seven new verrucosidins, namely penicicellarusins A-I (3-9), in addition to five known polyketides verrucosidin (1) [4,27], normethylverrucosidin (2) [27], penicyrone A-B (10-11) [28], and deoxyverrucosidin (12) [29] were obtained by the isolation workflow. The structures of known compounds were determined by comparing their NMR and MS data with literature data.
Compound 1 was obtained as white needle-like crystals and identified as verrucosidin by comparison of the NMR data reported in the literature [4,27] and the single-crystal X-ray crystallographic analysis ( Figure 3).   Figure S2.
Compound 1 was obtained as white needle-like crystals and identified as verrucosidin by comparison of the NMR data reported in the literature [4,27] and the single-crystal X-ray crystallographic analysis (Figure 3).   Figure 3. The X-ray crystallographic structure of 1 and 10.
Penicicellarusin B (4) was obtained as yellow oil with the molecular formula C 23 H 30 O 7 and nine degrees of unsaturation, as deduced from HRESIMS data. The 1D NMR spectroscopic data of 4 (Table 1) were similar with those of 3, except for the lack of a singlet methyl group and the presence of one additional olefinic proton at δ 5.66 in 4. HMBC correlations of H-2 (δ 5.66) to C-1, C-3 and C-4, H 3 -17 to C-3 and C-5, H 3 -24 to C-3 confirmed the structural changes on the 2H-pyran-2-one moiety in 4. A further comprehensive analysis of its 1 H-1 H COSY, HMQC, and HMBC spectra assigned the planar structure of 4 ( Figure 4).
The In the experimental ECD spectrum, compound 3 showed similar Cotton effects as verrucosidin (1) (Figure 6), supporting the same configuration at C-6 and C-7 between 1 and 3. Thus, compound 3 was assigned a 6R, 7S, 12S, 13S, 14R, 15R configuration and named penicicellarusin A. Penicicellarusin B (4) was obtained as yellow oil with the molecular formula C23H30O7 and nine degrees of unsaturation, as deduced from HRESIMS data. The 1D NMR spectroscopic data of 4 (Table 1)  Compound 5 was assigned the molecular formula of C24H34O7 (eight degree of unsaturation) on the basis of its HRESIMS at m/z 457.2190 [M+Na] + and NMR data ( Table 2). The 1 H-, 13 C-NMR, and UV spectra of 5 were similar with those of verrucisidinol [6], with the notable difference in the 1 H-NMR data of C-3, C-4, C-5, and H-7 (Table 2). A comprehensive analysis of its 2D NMR spectra including 1 H-1 H COSY, HMQC, and HMBC experiments confirmed the planar structure of 5 ( Figure 4).
The partial relative configuration of 5 was confirmed by a NOESY experiment ( Figure  5). The geometry of C8=C9 and C10=C11 were confirmed to be E by analysis of the NOESY observations. The key NOESY correlations of H-11 with H-13 and H3-22, H3-23 with H-13 and H3-22, and H-15 with H3-21 supported the same relative configurations on furan ring as verrucisidinol [6]. Considering the same biosynthesis origin, compound 5 is deduced to share the same absolute configuration with those of 1-4 in the furan ring. In addtion, the optical rotation data of 5 (  configurations (5a, 5b, 5c and 5d, Figure 7) were calculated using time-dependent density functional theory (TDDFT) at PBE1PBE/6-311 G* level with PCM model in methanol, and 60 exciting states were calculated. By comparison of the experimental and simulated ECD curves (Figure 7), the experimental ECD was match  Table 2). The 1 H-, 13 C-NMR, and UV spectra of 5 were similar with those of verrucisidinol [6], with the notable difference in the 1 H-NMR data of C-3, C-4, C-5, and H-7 (Table 2). A comprehensive analysis of its 2D NMR spectra including 1 H-1 H COSY, HMQC, and HMBC experiments confirmed the planar structure of 5 ( Figure 4).
The partial relative configuration of 5 was confirmed by a NOESY experiment ( Figure 5). The geometry of C 8 =C 9 and C 10 =C 11 were confirmed to be E by analysis of the NOESY observations. The key NOESY correlations of H-11 with H-13 and H 3 -22, H 3 -23 with H-13 and H 3 -22, and H-15 with H 3 -21 supported the same relative configurations on furan ring as verrucisidinol [6]. Considering the same biosynthesis origin, compound 5 is deduced to share the same absolute configuration with those of 1-4 in the furan ring. In addtion, the optical rotation data of 5 ([α]  To determine the absolute configurations at C-6 and C-7, ECD calculation method was applied. The four configurations (5a, 5b, 5c and 5d, Figure 7) were calculated using time-dependent density functional theory (TDDFT) at PBE1PBE/6-311 G* level with PCM model in methanol, and 60 exciting states were calculated. By comparison of the experimental and simulated ECD curves (Figure 7), the experimental ECD was match better with 5a (6R, 7S, 12S, 13S, 14R, and 15R). Thus, the compound 5 was assigned as 6R, 7S, 12S, 13S, 14R, and 15R, and named as penicicellarusin C.
The molecular formula of penicicellarusin D (6) was determined to be C 23 H 32 O 7 with the unsaturation degrees of eight on the basis of the HRESIMS data at m/z 443.2047 [M+Na] + (calcd. for C 23 H 32 O 7 Na m/z 443.2040) and NMR data ( Table 2). The NMR data of 6 were similar to those of 5 except for the absence of one singlet methyl group. The key HMBC correlations from H-2 (δ 5.60) to C-1, C-3 and C-4, as well as the upfield shift of C-2 (δ 88.5) confirmed the disappearance of the methyl group on the C-2 position in 6. Furthermore, Compound 6 showed similar Cotton effects in the experimental CD spectrum with those of 5 ( Figure S5), which assigned the absolute configurations of 6 as 6R, 7S, 12S, 13S, 14R, and 15R. It was designated as penicicellarusin D.
Penicicellarusins E-G (compounds 7-9) were determined to be fatty acid esters of 5 by interpretation of the HRESIMS, 1D and 2D NMR data (Table 3, Figures S6 and S7), and ECD spectra ( Figure S8 To assign the structure of fatty acid moieties, compounds 7-9 was hydrolyzed with alkaline solution followed by methyl esterification. The fatty acid chain in 7-9 was determined to be the palmitic acid, the oleic acid, and the linoleic acid, respectively, by comparison of the retention time and MS spectrum with those of standards by GC-MS analysis ( Figure S9). Compounds 7-9 showed similar Cotton effects in the experimental CD spectrum with those of 5 ( Figure S8), which assigned their absolute configurations as 6R, 7S, 12S, 13S, 14R, and 15R.
The molecular formula of penicicellarusin D (6) was determined to be C23H32O7 with the unsaturation degrees of eight on the basis of the HRESIMS data at m/z 443.2047 [M+Na] + (calcd. for C23H32O7Na m/z 443.2040) and NMR data ( Table 2). The NMR data of 6 were similar to those of 5 except for the absence of one singlet methyl group. The key HMBC correlations from H-2 (δ 5.60) to C-1, C-3 and C-4, as well as the upfield shift of C-2 (δ 88.5) confirmed the disappearance of the methyl group on the C-2 position in 6. Furthermore, Compound 6 showed similar Cotton effects in the experimental CD spectrum with those of 5 ( Figure S5), which assigned the absolute configurations of 6 as 6R, 7S, 12S, 13S, 14R, and 15R. It was designated as penicicellarusin D.
Penicicellarusins E-G (compounds 7-9) were determined to be fatty acid esters of 5 by interpretation of the HRESIMS, 1D and 2D NMR data (Table 3, Figure S6 and S7), and ECD spectra ( Figure S8 To assign the structure of fatty acid moieties, compounds 7-9 was hydrolyzed with alkaline solution followed by methyl esterification. The fatty acid chain in 7-9 was determined to be the palmitic acid, the oleic acid, and the linoleic acid, respectively, by comparison of the retention time and MS spectrum with those of standards by GC-MS analysis ( Figure S9). Compounds 7-9 showed similar Cotton effects in the experimental CD spectrum with those of 5 ( Figure S8), which assigned their absolute configurations as 6R, 7S, 12S, 13S, 14R, and 15R.
With the help of single-crystal X-ray crystallographic analysis, the 6R, 9R, 12S, 13S, 14R, and 15R absolute configuration of penicyrone A (10) was determined. The value of the Flack absolute structure parameter 0.03 (8) was obtained, and a perspective ORTEP plot was shown in Figure 3 (CCDC 2039558). According to X-ray diffraction analysis, the configuration at the C-6 positions in 10 was 6R, instead of 6S reported in the literature [28]. The structures of other known compounds were determined by comparing spectroscopic data with those in the literature. With the help of single-crystal X-ray crystallographic analysis, the 6R, 9R, 12S, 13S, 14R, and 15R absolute configuration of penicyrone A (10) was determined. The value of the Flack absolute structure parameter 0.03 (8) was obtained, and a perspective ORTEP plot was shown in Figure 3 (CCDC 2039558). According to X-ray diffraction analysis, the configuration at the C-6 positions in 10 was 6R, instead of 6S reported in the literature [28]. The structures of other known compounds were determined by comparing spectroscopic data with those in the literature.
To explore the bioactivities of verrucosidins, compounds 1-12 were evaluated for the antimicrobial effect, cytotoxic activity, and hypoglycemic activity. As a result, 1-12 showed no significant bioactivity in the antimicrobial assays and cytotoxicity assays at the dose of 100 µM. However, Compounds 1-4 were found to enhance the insulin-stimulated uptake of 2-NBDG in insulin-resistant HepG2 cells with the EC 50 values at 47.2 ± 1.2, 9.9 ± 2.5, 93.2 ± 1.2 and 40.2 ± 1.3 µM, respectively, while the other compounds showed no significant activity ( Figure 8). In particular, compounds 1, 2, and 4 showed much stronger activity than the positive drug (rosiglitazone) in the range of 25-100 µM. To explore the bioactivities of verrucosidins, compounds 1-12 were evaluated for the antimicrobial effect, cytotoxic activity, and hypoglycemic activity. As a result, 1-12 showed no significant bioactivity in the antimicrobial assays and cytotoxicity assays at the dose of 100 µM. However, Compounds 1-4 were found to enhance the insulin-stimulated uptake of 2-NBDG in insulin-resistant HepG2 cells with the EC50 values at 47.2 ± 1.2, 9.9 ± 2.5, 93.2 ± 1.2 and 40.2 ± 1.3 µM, respectively, while the other compounds showed no significant activity (Figure 8). In particular, compounds 1, 2, and 4 showed much stronger activity than the positive drug (rosiglitazone) in the range of 25-100 µM.

Discussion
Up to now, less than 20 verrucosidins and structurally-related compounds have been found in fungi. In this study, seven new verrucosidin derivatives (compounds 3-9), were

Discussion
Up to now, less than 20 verrucosidins and structurally-related compounds have been found in fungi. In this study, seven new verrucosidin derivatives (compounds 3-9), were isolated together with five previously identified compounds from the fermentation products of fungus P. cellarum, suggesting that this fungus is an important producer of verrucosidins.
Verrucosidins share similar structural features with the citreoviridins, including a methylated α-pyrone, a polyene linker, and a tetrahydrofuran ring. The citreoviridin biosynthetic gene cluster containing a polyketide synthase (CtvA), a SAM-dependent methyltransferase (CtvB), a flavin-dependent monooxygenase (CtvC), and a hydrolase (CtvD) has been identified in Aspergillus terreus var. aureus [21]. CtvC is the only monooxygenase in the cluster, which can iteratively oxidize the terminal triene portion of the precursor into a bisepoxide moiety. As a regioselective hydrolase, CtvD can transform the bisepoxide moiety into a tetrahydrofuran ring moiety [21,30]. In addition, the verrucosidin bosynthetic gene cluster was confirmed by constructing deletion mutants for verA gene coding for HR-PKS known to be the key enzyme of the biosynthesis. Different from citreoviridin, the bosynthetic gene cluster for verrucosidin in the genome of P. polonicum contains two flavin-dependent monooxygenases, VerC1 and VerC2, which means that the cluster can synthesize compounds with higher oxidation degree [22]. However, the enzymes involved in the biosynthesis of verrucosidin are largely uncharacterized.
In our work, with the genome of P. cellarum YM1 sequenced in our group, we searched for the gene cluster for penicicellarusins. By similarity analysis with the polyketide synthase gene ctvA and verA, the putative gene cluster celA was found in the genome of P. cellarum (Table 4 and Figure S11). Further bioinformatic analysis revealed seven genes, the polyketide synthase gene (celA), the SAM-dependent methyltransferase gene (celB), the flavin-dependent monooxygenase (celC1 and celC2), the cytochrome P450 gene (celD), the acyl-acyltransferase gene (celF), and the lyase gene (celH) potentially involved in the biosynthesis of penicicellarusins in P. cellarum. Based on above evidence, we propose the biosynthetic pathway of 1-12 (Figures 9 and S10). Verrucosidin (1) could be formed from 12 by oxidation of the olefinic bond, and further oxidation produces 3-6. Compounds 7-9 were biosynthesized by esterification of 5 with different fatty acids. Compounds 10 and 11 can be transformed from 5 or 6 by dehydration reaction.

Conclusions
In summary, a MS/MS-based molecular networking for the target discovery of verrucosidin-like polyketides was established in this study. The stereochemistry of the new compounds was determined by electronic circular dichroism (ECD) methods or comparison of experimental ECD spectra. The absolute configuration of penicyrone A (10) was corrected based on X-ray diffraction analyses. Bioactivity screening indicated that compounds 1, 2, and 4 showed much stronger promising hypoglycemic activity than the positive drug (rosiglitazone) in the range of 25-100 µM. The promising hypoglycemic activity is an interesting new bioactivity for this class of compounds. This work further proved the efficacy of the molecular networking in discovering natural products with unique structural features.
Supplementary Materials: The following are available online at www.mdpi.com/xxx/s1, Figure S1: The HPLC profiles of metabolites extracted from the culture medium of Penicillium strains, Figure  S2: The molecular network obtained by combining the LC-MS/MS analyses of extracts from P. cellarum YM1, Figure S3: Phylogenetic analysis and morphological characters of P. cellarum YM1, Figure  S4: Most stable conformers of 5 in solvated model calculations at the B3LYP/6-31+G(d,p) level (d), Figure S5: Experimental CD spectra of 5 and 6 in MeOH, Figure S6: Selected key HMBC and 1 H-1 H COSY correlations of 7-9, Figure S7: Selected key NOE correlations of 7-9, Figure S8: Experimental CD spectra of 5 and 7-9 in MeOH, Figure S9: GC-MS analysis of methyl linoleate, methyl oleate and products of alkaline hydrolysis-methyl esterification of compounds 8 and 9, Figure S10: Gene cluster schematic illustrating comparative organization of the penicicellarusin, verrucosidin, and citreoviridin, Figures S11-S40: NMR spectra for compounds 3-9. Early studies have demonstrated that verrucosidins and structurally-related compounds are endowed with several interesting bioactivities, such as antibacterial activities [3,10], antitumor [7,8], antiviral [9], and neurological activities [11]. In this work, it was found that compounds 1-4 show promising hypoglycemic activity, especially compounds 2 and 4. Preliminary structure-activity relationship showed that the formation of epoxy three-membered ring on C 6 -C 7 in the structures contributes greatly for the glucose uptakeenhancing activity in insulin-resistant HepG2 cells. The promising hypoglycemic activity is an interesting new bioactivity for this class of compounds.

Conclusions
In summary, a MS/MS-based molecular networking for the target discovery of verrucosidin-like polyketides was established in this study. The stereochemistry of the new compounds was determined by electronic circular dichroism (ECD) methods or comparison of experimental ECD spectra. The absolute configuration of penicyrone A (10) was corrected based on X-ray diffraction analyses. Bioactivity screening indicated that compounds 1, 2, and 4 showed much stronger promising hypoglycemic activity than the positive drug (rosiglitazone) in the range of 25-100 µM. The promising hypoglycemic activity is an interesting new bioactivity for this class of compounds. This work further proved the efficacy of the molecular networking in discovering natural products with unique structural features.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/ 10.3390/jof8020143/s1, Figure S1: The HPLC profiles of metabolites extracted from the culture medium of Penicillium strains, Figure S2: The molecular network obtained by combining the LC-MS/MS analyses of extracts from P. cellarum YM1, Figure S3: Phylogenetic analysis and morphological characters of P. cellarum YM1, Figure S4: Most stable conformers of 5 in solvated model calculations at the B3LYP/6-31+G(d,p) level (d), Figure S5: Experimental CD spectra of 5 and 6 in MeOH, Figure S6: Selected key HMBC and 1 H-1 H COSY correlations of 7-9, Figure S7: Selected key NOE correlations of 7-9, Figure S8: Experimental CD spectra of 5 and 7-9 in MeOH, Figure S9: GC-MS analysis of methyl linoleate, methyl oleate and products of alkaline hydrolysis-methyl esterification of compounds 8 and 9, Figure S10: Gene cluster schematic illustrating comparative organization of the penicicellarusin, verrucosidin, and citreoviridin, Figures S11-S40: NMR spectra for compounds 3-9.