Anthraquinones as Potential Antibiofilm Agents Against Methicillin-Resistant Staphylococcus aureus

Biofilms formed by methicillin-resistant Staphylococcus aureus (MRSA) are one of the contributing factors to recurrent nosocomial infection in humans. There is currently no specific treatment targeting on biofilms in clinical trials approved by FDA, and antibiotics remain the primary therapeutic strategy. In this study, two anthraquinone compounds isolated from a rare actinobacterial strain Kitasatospora albolonga R62, 3,8-dihydroxy-l-methylanthraquinon-2-carboxylic acid (1) and 3,6,8-trihydroxy-1-methylanthraquinone-2-carboxylic acid (2), together with their 10 commercial analogs 3–12 were evaluated for antibacterial and antibiofilm activities against MRSA, which led to the discovery of two potential antibiofilm anthraquinone compounds anthraquinone-2-carboxlic acid (6) and rhein (12). The structure-activity relationship analysis of these anthraquinones indicated that the hydroxyl group at the C-2 position of the anthraquinone skeleton played an important role in inhibiting biofilm formation at high concentrations, while the carboxyl group at the same C-2 position had a great influence on the antibacterial activity and biofilm eradication activity. The results of crystal violet and methyl thiazolyl tetrazolium staining assays, as well as scanning electron microscope and confocal scanning laser microscopy imaging of compounds 6 and 12 treatment groups showed that both compounds could disrupt preformed MRSA biofilms possibly by killing or dispersing biofilm cells. RNA-Seq was subsequently used for the preliminary elucidation of the mechanism of biofilm eradication, and the results showed upregulation of phosphate transport-related genes in the overlapping differentially expressed genes of both compound treatment groups. Herein, we propose that anthraquinone compounds 6 and 12 could be considered promising candidates for the development of antibiofilm agents.


INTRODUCTION
Bacterial biofilms are surface or interphase-attached microbial communities that are encapsulated in self-secreted extracellular matrix comprising largely of proteins, polysaccharides, nucleic acids, and lipids (Costerton et al., 1999;Flemming and Wingender, 2010). Bacteria growing inside biofilms are more extremely resistant to hostile environment, antimicrobial agents, and mechanical stresses than their planktonic counterparts (Gilbert et al., 2002;Fux et al., 2005;Hoiby et al., 2010Hoiby et al., , 2011. It is reported that more than 80% of chronic infections are related with the formation of biofilms by pathogens and very difficult to tackle (Potera, 1999;Lebeaux et al., 2013).

Sample Collection and Bacterial Identification
The soil sample was collected from the rhizosphere sediment of the mangrove plant Kandelia candel (L.) Druce in Mai Po Nature Reserve, Hong Kong (114.05 • E, 22.49 • N). Strain R62 was isolated from the sample through stepwise purification. The 16S rRNA gene of strain R62 was amplified by two universal primers of 27F (5 -AGAGTTTGATCMTGGCTCAG-3 ) and 1492R (5 -TACGGYTACCTTGTTACGACTT-3 ). The 16S rRNA gene sequence displayed 100% similarity with K. albolonga NBRC 13465.

Fermentation, Extraction, and Compound Isolation
The actinobacterial strain K. albolonga R62 was distributed to 250 ml Erlenmeyer flask containing 80 ml SGTYP medium (5 g of starch, 5 g of glucose, 1 g of peptone, 1 g of tryptone, 1 g of yeast extract, and 17 g artificial sea salt is dissolved in 1 L double-distilled water, pH 7.4-7.6) and fermented at 28 • C, 200 rpm for 5 days. In total, nearly 15 L of culture broth was centrifuged at 10,000 × g for 15 min to remove cells. The supernatant was extracted using ethyl acetate (ν/ν 1:3) three times to yield an EtOAc extract (3.0 g) after concentration. The extract was separated by ODS open column elution with a gradient of MeOH-H 2 O (20-100%, at intervals of 10%) to afford nine fractions (Fr. 1-Fr. 9). HPLC-MS analysis of these fractions showed that natural products are mainly concentrated in Fr. 3 and Fr. 4. Both fractions were isolated and purified by semipreparative HPLC (Waters, Parsippany, NJ, United States) FIGURE 1 | Chemical structures of anthraquinones mentioned in the article. Compounds 1 and 2 were isolated from Kitasatospora albolonga R62 and compounds 3-12 were commercially obtained.

The Determination of Minimum Inhibitory Concentration
The minimum inhibitory concentrations (MICs) of AQs were determined according to the Clinical and Laboratory Standards Institute (CLSI) guideline. The experiment was conducted in 96-well microplates (Corning, Corning, NY, United States). In brief, a stock solution of 20 µg/µl of each compound was prepared in DMSO, diluted 100 times with bacterial culture (∼5 × 10 5 CFU/ml) to 200 µg/ml, and further diluted by a serial twofold dilution to obtain the tested solutions with five concentrations ranging from 100 to 6.25 µg/ml. One microliter of DMSO was used as the negative control. Cell growth was monitored by the absorbance at OD 595 using the spectrophotometer (Varioskan Flash, Thermo Scientific, Waltham, MA, United States).

The Evaluation of Inhibitory Activity Against MRSA Biofilm Formation
The inhibition of static biofilm formation was tested as previously described (Stepanovic et al., 2007;Nair et al., 2016;Yin et al., 2019). After overnight cultivation, 100 µl/well of the bacterial culture, diluted in LB broth with 0.5% glucose (about 5 × 10 6 CFU/ml), was aliquoted into 96-well microplates (Corning, United States) with 1 µl of different concentrations (20 or 5 µg/µl) of each compound or DMSO and incubated at 37 • C for 24 h. Using the same method, 1 ml/well of the diluted culture was added into 24-well microplates (Corning, United States) with 2 µl of compounds (100 or 25 µg/µl) or DMSO. After incubation, each well was rinsed with 1 × PBS to remove non-adherent cells and then dried at 37 • C. CV staining was used to determine the remaining total biofilm biomass. Biofilms were stained with 0.5% CV for 10 min before washing. The washed biofilm mass was then dissolved in 30% acetic acid, and then the absorbance was measured at 550 nm.

Biofilm Eradication Assay
The 24-h mature biofilms containing no compounds or DMSO preformed in 24-well microplates according to the described method (Nair et al., 2016;Yin et al., 2019). Each well was rinsed twice by 1 × PBS, followed by adding fresh LB broth with 0.5% glucose and 2 µl of different concentrations (25 and 100 µg/ml) of each compound or DMSO, respectively, then incubated at 37 • C for another 24 h. The rest of the washing and CV staining steps were the same as the static biofilm formation assay. Additionally, 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazoliumbromide (MTT; HiMedia, West Chester, PA, United States) staining was conducted to show the number of viable cells in the remaining biofilms at 550 nm. Both CV staining and MTT staining assay were used to determine the eradication assay of compounds 6 and 12 in the gradient-diluted concentrations (6.25, 12.5, 25, 50, and 100 µ g/ml).

Biofilm Eradication Imaging
Bacterial biofilms were grown on the glass coverslips which were at the bottom of 24-well microplates at the beginning. The biofilm eradication assay was then carried out according to the method described in Yin et al. (2019) and Nair et al. (2016). For SEM imaging (Nithya et al., 2010;Sultana et al., 2016;Yin et al., 2019), before drying, the biofilms were fixed by 2.5% glutaraldehyde in 0.1 M phosphate for 24 h, followed by rinsing with 1 × PBS and dehydrated by a gradient of ethanol (50,70,80,90,95, and 100% ethanol for 10 min each). The biofilms were then subjected to analysis using a SEM (SU-70, HITACHI, Japan) after air drying and sputter coating with gold (50 s). The CSLM (TCS SP8, Leica, Germany) analysis was implemented as describe by the reported articles (Nithya et al., 2010;Dusane et al., 2011;Yin et al., 2019). In brief, the biofilms were washed with 1 × PBS after 24 h of treatment and dried at 37 • C for 30 min, then stained with the LIVE/DEAD BacLight Bacterial Viability Kit (L7007, Invitrogen, Carlsbad, CA, United States) in the dark. The biofilms were observed using a ×20 objective, and images were acquired with 1,024 × 1,024 resolutions. The CLSM was conducted at 488 and 561 nm.

RNA Extraction, Library Construction, Sequencing, and Data Analysis
Samples were collected according to the method mentioned in the biofilm eradication assay. Mature MRSA biofilms were treated with compound 6 or 12 at the concentration of 100 µg/ml and washed by 1 × PBS three times after 8 h.
The same volume of DMSO-treated mature biofilms was used as control. Total RNA of all the samples was extracted using Trizol reagent (Invitrogen) according to the manufacturer's instructions. RNA qualification was examined by 1% agarose gel electrophoresis and the NanoDrop spectrophotometer (Thermo Fisher Scientific). Strand-specific libraries were constructed using the TruSeq RNA sample preparation kit (Illumina, San Diego, CA, United States) and sequencing with Illumina nova-seq6000. The quality of raw data was checked by FastQC version 0.11.2. 1 High-quality clean reads were mapped to reference gene 1 http://www.bioinformatics.babraham.ac.uk/projects/fastqc/ S. aureus A9754 sequence (genome assembly: ASM17799v1) using STAR (2.5.3a) (Dobin et al., 2013). All mapped reads were then annotated by BLASTX against NCBI non-redundant protein database (nr) and Swiss-Prot. Gene IDs of the whole detected sequence were named according to the information on the ensemble bacterial database for analysis. 2 The fragments per kilobase of transcript per million mapped reads (FPKM) for the mapped sequences of each sample were calculated and normalized by Cuffnorm (Trapnell et al., 2012), and the differential expression of genes between control and compound treatment groups were analyzed with DESeq2 (version 1.16.1) (Love et al., 2014). Some differentially expressed genes (DEGs; p < 0.05 and absolute fold change ≥ 2) were chosen for subsequent analysis of gene function and signaling pathway enrichment based on searching against Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) database (Sun et al., 2017;Wang et al., 2018). The GO functional annotation of the DEGs was determined with topGO package (Alexa and Rahnenfuhrer, 2010) for each gene clusters, and the KEGG pathway enrichment analysis was performed using phyper algorithm in R 3.6.1 3 against the KEGG database. Significantly enriched GO terms and KEGG pathways were selected using a threshold of p-value of ≤0.05.

qRT-PCR
To confirm the RNA-Seq data, the relative expression of some related genes was assessed by qRT-PCR. Total RNA was extracted by the E.Z.N.A. R Bacterial RNA Kit (Omega Bio-Tec, Norcross, GA, United States). cDNA was synthesized using a PrimeScript RT reagent kit with a gDNA Eraser (TaKaRa, Tokyo, Japan). qRT-PCR reactions (20 µl in total volume with 4 µl of cDNA product) were carried out using TaKaRa SYBR Premix Ex Taq (Tli RNaseH Plus) in a CFX96 Real-Time System (BioRad, Hercules, CA, United States). All the primers used in the present study had been listed in the Supplementary Table 1.
Housekeeping gene pyk was used as an internal reference to normalize the qRT-PCR date (Tan et al., 2015). Three independent replicates with four replications were tested in each qRT-PCR run.

Statistical Analysis
All the experiments were performed in at least three independent experiments and three biological replicates each time. Singlefactor analysis of variances followed with Ryan-Einot-Gabriel-Welsch F-test as the post hoc method were performed to test the differences of the antibiofilm activity between 12 different compounds using SPSS software 16.0 in

Biofilm Inhibition Activity of Natural Products 1 and 2 at Concentrations Lower Than MIC Values
The preliminary screening results of antibiofilm activities of compounds 1 and 2 revealed that compound 1 had a significant effect on decreasing MRSA biofilm formation by >50% at the concentration of 200 µg/ml while its MIC value was higher than 200 µg/ml (Figure 2), but compound 2 showed no obvious bactericidal activity or antibiofilm activity at the same concentration.

Antibacterial and Antibiofilm Activities of Anthraquinones
The antibacterial and antibiofilm activities of the other 10 commercial AQs 3-12 (Figure 1) against MRSA ATCC43300 were investigated to analyze their SARs. Among them, compounds 6 and 12 displayed antibacterial activity with MIC values of 100 and 12.5 µg/ml, respectively. The other compounds did not show antimicrobial activity against planktonic MRSA cells at the concentration of 200 µg/ml (Table 1). Meanwhile, antibiofilm activities (including the inhibitory effects on the formation of MRSA biofilms and the eradication abilities against preformed biofilms) of compounds 3-12 were evaluated at concentrations of 200 and 50 µg/ml by the CV staining method (Figure 3). Compounds 5, 6, 8, 9, 11, and 12 could all strongly inhibit the formation of MRSA biofilms by >50% at the concentration of 50 µg/ml. It is worth noting that the inhibition rates were 75.2% (±3.3%) and 96.4% (±0.3%) for compounds 6 and 12, respectively. On the contrary, compound 7 promoted biofilm forming which was not parallel to the trend of this class of compounds ( Figure 3A). In addition to their good biofilm formation inhibitory activities, compounds 6 and 12 also showed good eradication activity against mature MRSA biofilms. Compound 6 could eradicate 54.2% (±5.9%) and 57.7% (±3.8%) of the mature MRSA biofilms at 200 and 50 µg/ml, and compound 12 could eradicate 44.0% (±3.8%) and 26.5% (±3.2%) at 200 and 50 µg/ml, respectively ( Figure 3B). Based on their outstanding performance in both inhibiting and eradicating MRSA biofilms, we therefore chose compounds 6 and 12 as representatives for further antibiofilm study.

Biofilm Eradication Activity of Compounds 6 and 12
Two kinds of staining methods were conducted to further qualify the dose-dependent effects of compounds 6 and 12 against MRSA biofilms. The remaining total biomass in biofilms was determined by the CV staining method, while the metabolic activity of the viable cells was investigated by MTT staining method. Both compounds displayed only weak to moderate effect on reducing the total biomass of 24 h mature MRSA biofilms at concentrations ranging from 12.5 to 200 µg/ml (Figures 4A,B, 5A,B). However, these two compounds could significantly disperse the living cells inside biofilms and showed strong eradicating efficiency with concentrations ranging from 12.5 to 200 µg/ml (Figures 4C,D, 5C,D). The remaining rates of viable biofilm cells were 47.9% (±3.8%) and 34.5% (±6.1%) at the lowest tested concentration of 12.5 µg/ml, respectively.

SEM and CLSM Analysis of Biofilm Structures After Treatment With Compounds 6 or 12
SEM and CLSM were used to analyze the 24 h preformed mature MRSA biofilm structures after treatment with compounds 6 and 12 for another 24 h. The images of SEM showed that the total MRSA biofilm biomass was significantly reduced when treated with compounds 6 and 12 at concentrations of 200 and 100 µg/ml (Figure 6). There were not many differences between compounds 6and 12-treated groups. It also should be noticed that the cell walls and membranes of the biofilm cells after treatment with both compounds remained as intact as those in the negative control group. Meanwhile, the viable bacteria within biofilms could be represented by the thickness and fluorescence intensity of the three-dimensional structures of MRSA biofilms in CLSM images (Figure 7), which showed that compound 12 at 50 and 100 µg/ml was highly effective in removing viable cells, and compound 6 at 100 and 200 µg/ml could also remove most of living bacteria.

Gene Expression Analysis
All the sequencing reads of gene expression of MRSA cells under treatment with 100 µg/ml of compounds 6 and 12 have been submitted to the NCBI Gene Expression Omnibus (GEO) with accession number GSE147157. Overall, among 3,050 unigenes of the reference strain S. aureus A9754, 2,989 unigenes (98%) matched those in the NR database and 2,088 (68.46%) matched those in Swiss-Prot. After treatment with compounds 6 and 12, 237 and 302 DEGs were detected compared with the control group, respectively ( Figure 8A). A total of 84 and 153 genes were down-and up-regulated respectively in compound 6-treated group, while there were 145 downregulated and 157 upregulated genes in compound 12-treated group (Figures 8B,C). The Venn diagrams showed that 166 DEGs (including 73 downregulated and 93 upregulated) had the same expression trend in both treatment groups (Figures 8A-C). The heatmaps revealed the induced and suppressed transcripts in the treatment groups compared with the control group (Figures 8D,E).

Functional Analyses of Differentially Expressed Genes
To further understand the function of DEGs, GO enrichment and KEGG pathway analyses of two treatment groups were performed respectively (Figure 9). According to GO classification analysis, DEGs were classified into three categories: biological process, cellular component, and molecular function. The GO enrichment analyses showed that the treatment groups had great different effects on the transcriptome of biofilm cells (Figures 9A,B). As in the biological process, all the GO term-related genes in the compound 6 treatment group were upregulated while only 2 out of 10 term-related genes were upregulated in the compound 12 treatment group. On the other hand, there were also similar changes in both treatment groups, such as the upregulation of nitrate metabolic process (GO: 0042126) and nitrate assimilation (GO: 0042128) in biological processes and the four terms enriched in molecular function were the same in both treatment groups. KEGG pathway analysis was used to clarify the biological function of genes (Figures 9C,D). A total of seven and nine pathways were significantly enriched in the compounds 6 and 12 treatment group, respectively. Among the pathway enrichment in the compound 6 treatment group, there were 17, 14, and 8 DEGs belonging to microbial metabolism in diverse environments, biosynthesis of amino acids, and two-component system, respectively. In the compound 12 treatment group, there were 10 DEGs that belonged to pyrimidine metabolism, 9 DEGs belonged to two-component system, and 8 DEGs belonged to glycolysis/gluconeogenesis pathway.

Overlapped Differentially Expressed Genes of Two Treatments
A total of 166 DEGs, including 73 downregulated and 93 upregulated, overlapped in compounds 6 and 12 treatment groups. As shown in Figure 10A, the majority of downregulated DEGs were enriched to cellular component ontology, of which 26 genes were enriched in cell periphery (GO:0071944), 23 genes in plasma membrane (GO:0005886), and 23 genes in membrane (GO:0016020). The GO terms related to compound biosynthesis and metabolic processes were frequently shown in biological process category, and the kinase-and transferase-related terms were mainly enriched in molecular function category. Meanwhile, the upregulated categories ( Figure 10B) showed that metabolic process-and carbolic process-related GO terms were prevalent in the biological process category, oxidoreductase activity, and binding and transporter activity in the molecular function category. Among these three categories, most of DEGs were enriched in biological process with 30 genes enriched in single-organism metabolic process (GO:0044710), followed by 22 genes in small molecule metabolic process (GO:0044281) and 16 genes in oxidation-reduction process (GO:0055114); 16 genes enriched in oxidoreductase activity (GO:0016491) were the largest group in molecular function category; oxidoreductase complex (GO:1990204) with two significant genes was annotated in cellular component category.

qRT-PCR Validation
To validate the RNA-Seq results, some genes involved in phosphate-specific transport system (phoU and pstS), ironacquisition/transport (isdC, SKAG_01383 and SKAG_001385), and other genes (lrgA, mhqA, hrtA, and SKAG_00613) were selected to conduct the qRT-PCR analysis (Figure 11 and Supplementary Tables 1, 2). The results showed that isdC, lrgA, SKAG_00613, SKAG_01383, and SKAG_001385 genes were significantly downregulated in two groups. On the contrary, phoU, pstS, and hrtA were significantly upregulated in both groups. What is more, mhqA was significantly upregulated in compound 6 treatment group, while it had no significant change in compound 12 treatment group. In general, most of the qRT-PCR results of these selected genes were consistent with the RNA-Seq results, indicating that the RNA-Seq data were reliable.

DISCUSSION
In this study, a total of 10 structurally related AQs including two microbe-derived natural products DMCA (1) and TMCA (2) and their eight analogs were tested for antibiofilm activity. Among these compounds, only anthraquinone-2-carboxlic acid (6) and rhein (12) exhibited bactericidal activity with MIC values lower than 200 µg/ml. Both compounds had a carboxyl group attached to the C-2 position of the anthraquinone skeleton, which is different from other tested anthraquinones. For antibiofilm activity, some of the hydroxy-anthraquinones have been reported for their biofilm-inhibiting activity, and previous SAR analysis suggested that the hydroxyl groups at the C-1 and C-2 positions of the anthraquinone skeleton should play an important role in inhibiting biofilm formation at low concentrations (lower than 10 µg/ml) (Manoharan et al., 2017). Herein, it is interesting to note that the single hydroxyl group at the C-1 and C-2 positions exerted different biofilm inhibition activities at high concentrations (50 and 200 µg/ml). When comparing compounds 3-5, compound 5 showed a higher rate of biofilm inhibition than compounds 3 and 4, suggesting that the substitution of a hydroxyl group at the C-2 position was more important than the C-1 position in enhancing the prevention capability of MRSA biofilm formation. This was also supported by the good biofilm inhibitory activity of compounds 8 and 9, both of which also have a C-2 hydroxyl group. Surprisingly, compound 7 with two symmetric hydroxyl groups at C-2 promoted biofilm formation instead, which was different from all the other compounds and requires further investigation. Consistent with the literature (Manoharan et al., 2017), at high concentrations, compounds 8, 9, and 11 still had excellent inhibitory performance for MRSA biofilm, but compounds 3, 4, 7, and 10 could not suppress biofilm formation.
The biofilm eradication activity of these hydroxy-and carboxy-substituted AQs was then analyzed, and the SAR study revealed that: (i) for the hydroxyl AQs 5, 7, and 9, the biofilm eradication rates of these three compounds were all around 20-30% at the concentration of 200 µg/ml, suggesting that the number of hydroxyl substitutions might not have much correlation with their activity; (ii) for the carboxylic AQs 3, 6, and 12, compounds 6 and 12 showed higher biofilm eradication efficiency (the respective eradication rate of 54.2 ± 5.9 and 44.0 ± 5.3% at 200 µg/ml, and 57.7 ± 3.8 and 26.5 ± 3.2% at 50 µg/ml for compounds 6 and 12, respectively) than that of compound 3, indicating that the carboxyl group at C-2 position of the anthraquinone skeleton might be very important for improving eradication activity against MRSA biofilm; and (iii) combining the results of all the SAR analysis of hydroxy-and carboxy-AQs, we could conclude that the AQs including the carboxyl group at C-2 position of anthraquinone mother nucleus exhibited better MRSA biofilm eradication activity than the hydroxyl ones. Therefore, we suggest that the carboxyl substitution attached at C-2 position of anthraquinone should be considered a key group for further chemical design and development of antibiofilm AQs. In addition, it is particularly noteworthy that the biofilm eradication activity of carboxyl-substituted anthraquinone compounds 6 and 12 was reported for the first time in this study.
Since compounds 6 and 12 displayed promising eradication activity against MRSA biofilm, different staining methods, microscopy images, and RNA-Seq analysis were employed for more in-depth study. It was found that the MTT staining test always displayed a higher biofilm eradication rate than the CV staining test under the same concentrations of both compounds, which lead us to wonder whether their biofilm eradication effects against the preformed MRSA biofilms mostly involved decreasing the viable cells within biofilms rather than reducing the total biomass. Thus, the SEM imaging of the biofilm structures treated with these two compounds was carried out, and results showed visibly structural disruption of the treated biofilms. And the CLSM imaging analysis revealed that a significant decrease of the living biofilm cells in both treatment groups. All these results indicated that compounds 6 and 12 could gently reduce the total biofilm biomass but effectively remove the viable biofilm cells when exerting their antibiofilm activity.
For RNA-Seq analysis, we focused on the comparison of our transcriptomic results with those reported in literature (Monds et al., 2001;O'May et al., 2009;Neznansky and Opatowsky, 2014;Wang et al., 2017;Abouelhassan et al., 2018) due to the currently ambiguous and incomplete mechanism of the MRSA biofilm eradication. Recently, a halogenated phenazine has been reported to eradicate MRSA biofilms depending on an ironstarvation mechanism (Abouelhassan et al., 2018). The RNA-Seq results of compounds 6 and 12 treatment groups showed that most of the iron-acquisition-and iron-transport-related genes (isdABCEFGHI, sbnAEI, and mntAB) were downregulated or insignificantly differentially expressed instead of upregulated (Supplementary Table 2). Even though the iron ion might not be involved with the working mechanism of our compounds, it provided us a clue to analyze other ions. Meanwhile, among the overlapped DEGs of the two experimental groups, except for those involved in metabolic-and catabolic-related progress or function, many of the other upregulated genes belonged to transport GO terms: anion transport (GO:0006820), inorganic phosphate transmembrane transporter activity (GO:0005315), and phosphate transmembrane transporter activity (GO:1901677). It has been reported that the phosphatespecific transport system (Pst system; Supplementary Table 2), including five genes pstSCAB-phoU, has a high affinity for uptake of inorganic phosphate (P i ) and is very important for cell growth. What is more, the pst mutants revealed that these genes are necessary for the biofilm formation of different bacteria (Monds et al., 2001;O'May et al., 2009;Neznansky and Opatowsky, 2014;Wang et al., 2017). In our study, the upregulation of these genes suggested that P i uptake was increased after treatment of compound 6 or 12. The redundant P i may be used to synthesize high-molecular-weight inorganic polyphosphate to store energy for cell metabolism and other purpose (Monds et al., 2001;Wang et al., 2017). We speculated that compounds 6 and 12 might influence phosphate homeostasis in MRSA which in turn led to the disassembly of biofilms, and more experiments are required to confirm this hypothesis.
As for the safety of compounds 6 and 12, numerous studies have been carried out to study their biological activities and related molecular mechanisms (Sheng et al., 2011;Zhao et al., 2011;KoraMagazi et al., 2016;Panigrahi et al., 2016;Park et al., 2016a,b;Yuan et al., 2016). Take the cytotoxicity for instance, compound 6 had been demonstrated to have antiinflammatory activity without cellular toxicity at concentrations below 26.8 µg/ml (Park et al., 2016a,b). On the contrary, compound 12, which has been well studied in liver diseases, could trigger liver cytotoxicity in the long-term use, and the cytotoxicity could take place quickly when used over 2.8 µg/ml (Sheng et al., 2011;Zhao et al., 2011;KoraMagazi et al., 2016;Panigrahi et al., 2016;Yuan et al., 2016). Herein, although both compounds displayed biofilm eradication activity at the lowest effective concentration of 12.5 µg/ml, it seems that compound 6 might be safer than compound 12 when used as an antibiofilm agent.
In our study, the antibacterial and antibiofilm activities against MRSA of a series of structurally related AQs have been well evaluated. The SAR analysis demonstrated that different positions of carboxyl and hydroxyl substitutions of the anthraquinone skeleton can significantly affect antibiofilm activity of these compounds. A hydroxyl group at the C-2 position of the anthraquinone skeleton was much more important than the C-1 position for improving their inhibitory activity against MRSA biofilm formation. Carboxylic AQs were more effective in the eradication of MRSA biofilms than the hydroxyl ones. The results of MTT staining and CV staining test, SEM and CLSM images, and RNA sequencing revealed that the MRSA biofilm eradication mechanism of compounds 6 and 12 might involve dispersion of biofilm cells and disruption of phosphate homeostasis.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary Material.

AUTHOR CONTRIBUTIONS
Z-MS and YX designed the project. Z-MS and K-LW carried out experiments, interpreted the results, and wrote the manuscript. K-LW, YX, and C-YW funded the research and revised the manuscript. KZ, L-MY, and YZ helped in the RNA-Seq analysis. J-LZ provided a lot of help during the experiment. All authors contributed to the article and approved the submitted version.