Minocycline Might Be an Adjunctive Therapy Option for the Treatment of COVID-19: In Silico Screening, Structure-anity Relationship, and Literature Review

Up to now, there is no specic therapy for the globally ongoing COVID-19. To explore potential inhibitors of Severe Acute Respiratory Syndrome-Coronavirus-2 (SARS-CoV-2) for the treatment of novel coronavirus disease (COVID-19), in silico screening of 135 clinical drugs was performed targeting on 3-chymotrypsin-like protease (3CLpro, or M pro ). Six drugs including anti-HIV drug (raltegravir), antibacterial drugs (cefonicid, cefoperazone, minocycline), and antidiabetic drugs (canaglifozin, glyburide) showed high binding anities ( ≤ -8.5 kcal/mol) and interesting binding conformations compared with the designed co-crystal ligand N3 (-7.7 kcal/mol). In which the antibiotic minocycline, an inhibitor of bacterial ribosomal rRNA, showed the highest binding anity (-9.6 kcal/mol). Valuable hydrogen bonding and hydrophobic interactions were found between minocycline and M pro active site. Beside the hydrogen bond with Cys145, minocycline formed a Pi-Cation with His41, which strongly supported minocycline as a Michael Addition acceptor to bind with the catalytic site of M pro . The structure-anity relationship was studied based on molecular docking of minocycline analogues. Literature review found that minocycline had both in vitro and in vivo broad-spectrum antiviral as well as anti-inammatory activities, and the levels of a broad-spectrum of biological markers during minocycline administration were opposed to those of COVID-19 condition (both severe and non-severe). Minocycline deserves in vitro and in vivo evaluations as SARS-CoV-2 inhibitor as well as a randomized controlled trial to investigate the ecacy for COVID-19. These studies will shed new light on an adjuvant treatment strategy for this potentially lethal viral disease.


Introduction
The novel coronavirus (Severe Acute Respiratory Syndrome-Coronavirus-2, SARS-CoV-2) disease (COVID- 19) was rst detected in Wuhan, China in December 2019. As the global ongoing epidemic of SARS-CoV-2, there have been more than 7,500,000 people infected and more than 40,000 patients died according to the WHO. COVID-19 is an acute infectious disease with common symptoms including fever, cough, shortness of breath, and dyspnea. In more severe cases, infection causes pneumonia, severe acute respiratory syndrome, kidney failure, or even death [1]. Although several anti-viral (mainly anti-HIV) drugs and hydroxychloroquine were investigated the e cacy for the COVID-19, the outcomes of these trials structure and catalytic mechanism of SARS-CoV M pro allows it as an important target for anti-coronavirus drug development.
There are three strategies for developing an anti-SARS-CoV-2 drug [4]. The rst is to test existing broadspectrum antivirals for their metabolism, used dosages, e cacy and side effects are clear. However, broad-spectrum antivirals cannot kill SARS-CoV-2 with clear targeting and the side effects should not be underestimated [1]. The second is to high-throughput-screen for SARS-CoV-2 therapeutic candidates from existing clinical drugs ("old drugs") [5,6]. The third is to develop a new drug from scratch, however, the development and registration procedure of a new antiviral would be theoretically time-consuming [7]. For the novel virus, screening of potential "old drug" molecules for COVID-19 would be the fastest way. Since old drugs have been prepared, the medication has su cient experience, and the safety and pharmacokinetic parameters are well known, with the in vivo e cacy in animal model, it could be approved by the Green Channel or the hospital ethics committee for clinical use [1].
In this study, we established a small-scale "old" drug database (clinical drugs being used in Ordos Central Hospital and existing antivirals) according to Linpinski's Rule of 5, and conducted in silico screening of potential M pro inhibitors from the database by molecular docking. Binding a nity and interaction as well as structure-a nity relationship were analyzed to better understand the potentiality. Finally, literature support for the potentiality of anti-SARS-CoV-2 and the treatment of COVID-19 was reviewed and analyzed. This study will provide contributions to the transient ongoing infectious disease.

Pharmacophore study of the co-crystal ligand N3
The crystal structure of SARS-CoV-2 M pro in complex with a designed small ligand N3 to 2.1-Å resolution (PDB code: 6LU7) had been determined by Professors Zihe Rao and Haitao Yang's research team from ShanghaiTech University [8]. The protein coordinates of the M pro used in this study were in-time donated by Zihe Rao et al. Based on the structure, key helixes/loops, amino acid residues, and hydrophobic interactions in binding site were investigated and the pharmacophore of N3 was summarized, which was used as a control in the following in silico study.

Small-scale drug database
According to Linpinski's Rule of 5 (e.g., molecular mass less than 500 Da, no more than 5 hydrogen bond donors, no more than 10 hydrogen bond acceptors, and an octanol-water partition coe cient not greater than 5) [

Molecular docking by AutoDock Vina
We applied a work ow for molecular docking which was described in our previous work [10]. The chain B (co-crystal ligand N3 in 6LU7) and chain C (water molecules) were deleted, and chain A was prepared for docking within the molecular modeling software package Chimera 1.10.2 (National Institutes of Health, Bethesda, MD, USA) [11]. Adding of polar hydrogens and kollman charges, gasteiger computing and grid box parameters de ning were done using MGL tools 1.5.6 (The Scripps Research Institute, La Jolla, CA, USA) [12,13].
All the ligands were set as exible and the receptor was set as rigid. Docking calculations were performed using AutoDock Vina 1.1.2 software (The Scripps Research Institute, La Jolla, CA, USA) [14]. A search grid box was set to cover the whole surface of M pro protein to collect all possible orientations and conformations of the ligand paired with the protein (including compounds outside the active site). For which, the center was set as: center_x=-23.982, center_y = 12.114, center_z = 57.466, and the size was set as: size_x = 58, size_y = 78, size_z = 66. Spacing angstrom was set as 1.000, and the exhaustiveness was set as 100. The default settings and the AutoDock Vina scoring function were applied.
Totally 9 binding modes were generated by AutoDock Vina for each compound, and the mode (even outside the active site) with highest binding a nity was selected as the most predictable. Visual investigation and analysis of ligand-protein interactions were performed using PyMOL V.1.5 (Schrodinger LLC, New York, NY, USA). Binding conformation, a nity, and receptor-ligand interaction were analyzed with N3 as a control.

Re-docking study by Discovery Studio
The accurate 3D protein structure of M pro was de ned as the receptor and optimized by hydrogenation, dehydration and removing redundant residues. Location of the originally contained ligand N3 in the cocrystal was de ned as the active binding site with radius as 13.890841 which could cover the best binding region. The X, Y, and Z centers were − 10.797, 12.536, and 68.905, respectively. Molecular structure of N3 was also prepared and converted to 3D structure and its energy was minimized. The molecular docking was performed using CDOCKER tool. -CDOCKER_ENERGY and -CDOCKER_INTERACTION_ENERGY was used to score the interaction between receptor and ligand.
Discovery Studio 2016 software (Biovia, San Diego, CA, USA) was used for the docking, visualization, and analysis.

Structure-a nity relationship of promising drug
Considering certain molecular structure of a ligand underlies certain ligand-receptor binding conformation, the generated information is relatively limited. Investigation of promising drug analogues might provide further information. After the in silico screening and the re-docking simulation of promising drug, the analogues were collected by referring to literatures from PubMed, Elsevier, Springer, and Google Scholar, and a small-scale analogue database was established. Then one-by-one docking of the analogues targeting on M pro was performed. Based on the analogues' binding a nities, the structurea nity relationship of the promising drug was summarized.

Literature review of promising drug
Considering anti-SARS-CoV-2 potentiality of the promising drug is simulation, we further questioned whether the promising drug have documented antiviral activities to the virus genera and whether the promising drug possess biological activities associated with the pathological changes in COVID-19 condition. We searched PubMed, Elsevier, Springer, and Google Scholar for articles describing SARS-CoV-2 virus, COVID-19 condition, and the biological changes during the promising drug use. The effect of COVID-19 and the promising drug on human biological changes were summarized and analyzed.
As shown in the diagram of Fig. 1, a covalent bond between the Sγ atom of Cys145 and the Cβ of the vinyl group is formed, which means the Michael Addition that is critical in the catalytic mechanism has occurred [3]. The lactam functional group at P1 site inserts into the subsite S1 (consists of residues Phe140, Asn142, Glu166, His163, His172, and Leu141 as well as two waters) and forms a hydrogen bond with His163, while the functional group Leu at P2 site inserts deeply into the hydrophobic subsite S2 (consists of residues His41, Met49, Tyr54, Met165, and Asp187) [8]. The functional group Val at P3 site is solvent-exposed tolerating a variety of functional group substitutions. The functional group Ala at P4 side is in a hydrophobic pocket (consists of residues Met165, Leu167, Phe185, Gln189, and Gln192). P5 site makes van der Waals interactions with Pro168, Thr190, and Ala191, while the hydrophobic aromatic ring of N3 forms van der Waals contacts with Thr24 and Thr25. Besides, N3 forms multiple hydrogen bond interactions with the active site residues, helping to lock the inhibitor inside the binding pocket, which determines the inhibition of the enzyme as well as the coronavirus replication [8].

In silico screening by AutoDock Vina
For validation of docking simulation, N3 was re-docked into M pro [10]. The docking work ow (described in 2.3) allowed top-ranked and reproduced binding conformation which was close to those of the 6LU7 cocrystal structure (checked by PyMOL, RMSD of 1.126 Å). According to AutoDock Vina, binding a nity ≤ -0.0 kcal/mol means the receptor and ligand could automatically bind together. In this study, molecule with binding a nity ≤ -8.5 kcal/mol was treated to be potential based on recent reports on in silico screening of SARS-CoV-2 M pro inhibitors [15].
All the 135 "old" drug structures, biological activities, targets, and top-ranked binding a nities were summarized (Supporting information Table. S1). In which, 6 molecules including anti-HIV drug (raltegravir), antibacterial drugs (cefonicid, cefoperazone, minocycline), and antidiabetic drugs (canaglifozin, glyburide) showed high a nities (≤ -8.5 kcal/mol) as well as interesting binding conformations (bound to the M pro active site and formed interesting interactions with key residues). In particular, the antibiotic minocycline, an inhibitor of bacterial ribosomal rRNA, showed the highest binding a nity (-9.6 kcal/mol) compared with N3 (-7.7 kcal/mol). The results indicated that these small molecular drugs might be M pro inhibitors of SARS-CoV-2. However, the docking simulation was just potential, evaluations by in vitro cell-based models and in vivo animal models would provide further information whether these drugs could be used as SARS-CoV-2 inhibitors.

Re-docking of N3 and minocycline by Discovery Studio
To gain further validation of the docking simulation, re-docking of known ligand with the target and comparison of docking results with the published co-crystal as well as the comparison of docking results generated by different software are academically consensus.
From the CDOCKER results generated by Discovery Studio, N3 ( Fig. 2A) formed conventional hydrogen bonds with residues Phe140, His163, His164, Glu166, Gln189, and Thr190. The isoxazole group formed Pi-Alkyl interaction with Ala191 and Pro168, and hydrophobic aromatic ring formed van der Waals' forces with residues Thr24, Thr25, Leu27 and Cys145. In addition, N3 molecule forms covalent bonds with multiple residues of M pro . The docking results were closely consistent with the co-crystal structure, indicating that the CDOCKER docking model was validated and suitable for in silico screening of M pro inhibitors.
For minocycline (Fig. 2B), it contains multiple hydrophilic groups which formed conventional hydrogen bonding networks with key residues Phe140, Gly143, Cys145, His164, and Glu166 in the active site. The hydrophobic aromatic rings formed van der Waals' forces with multiple amino acid residues of M pro . It is commonly accepted that covalent bond formed between the Cys145-His41 catalytic dyad and the designed compound would increase the M pro inhibition potency, resembling the intermediate during substrate cleavage [3]. Beside the hydrogen bond between Cys145 and 2-carboxamide, a critical Pi-Cation formed between His41 and 4-dimethylamino group, which strongly supported minocycline as a Michael Addition acceptor binding with the exact catalytic site to inhibit M pro . These results indicated that the multiple including critical interactions stabilized minocycline bound with M pro in a low-energy state, which was required for M pro selection and antiviral activity.

Structure-a nity relationship of minocycline
Minocycline analogues were collected by referring to the literatures from PubMed, Elsevier, Springer, and Google Scholar. Finally a 44-compound (in which 21 compounds were clinical drugs) small-scale database was established. After molecular docking targeting on M pro , the chemical structures and topranked binding a nities of the minocycline analogues were summarized (Supporting information Table. S2).
Indeed, minocycline showed a promising highest binding a nity among all the 44 analogues. Structures simply derivable from minocycline (containing the main octahydrotetracene-2-carboxamide skeleton) were analyzed and the structure-a nity relationship was summarized (Fig. 3). From the structure-a nity relationship, carbonyl functional groups should be kept and the middle hydroxyl group might be better if changed to be carbonyl. Furthermore, the terminal 2-carboxamide could be modi ed with moderate (not too long) moiety. On the 4,7-bis(dimethylamino) side, 4-dimethylamino group is critical for the high a nity, and the S-stereochemistry of C4 is better than the R-stereochemistry, which was also veri ed from the previous binding mode investigation that it could form the key covalent bond with His41.

Literature review of minocycline
Due to the limited resource, we could not test the antiviral activities of minocycline in SARS-CoV-2 cell and animal models. Instead, we did the literature review to nd the support information. From PubMed, Elsevier, Springer, and Google Scholar databases, articles describing COVID-19 and minocycline use until 2020, Mar 6th were searched. Totally 165 papers including published and preprints were found, in which 25 papers were associated with the biochemical indexes of COVID-19 patients, and 22 papers were associated with the anti-in ammatory effect of minocycline. The effect of COVID-19 condition and minocycline on selected biomarkers including erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), lactose dehydrogenase (LD), D-dimer, CD4 + T cell, CD8 + T cell, tumor necrosis factor (TNF)-α, interferon (IFN)-γ, interleukin (IL)-6, and IL-10 were summarized and analyzed (Table 1).
A variety of clinical data con rmed that the in ammatory factor "storm" (IFS) existed and played an important role in severe or fast-progressive COVID-19 condition. Among the selected biomarkers, the increase of ESR, CRP, IL-6, and D-dimer helps to judge the COVID-19 progression [15]. Besides, the counts of lymphocytes, total T cells, CD4 + T cells, and CD8 + T cells were dramatically reduced, while patients in decline period showing decreased levels of TNF-α, IFN-γ, IL-6, IL-10 and restored counts of T cell [16,17]. IL-6 and IL-10 are found to be the core cytokines that are consistently found to be elevated in patients with CRS [18,19]. The increases of IL-6 and IL-10 levels in COVID-19 patients were found to run parallel with the severity of the disease [20]. Consistent decrease of CD4 + T and CD8 + T cells, spleen damage, and lymphocyte depletion exist in COVID-19 patients [21]. In severe COVID-19 patients, the elevation of CRP level and white blood cell count might be accompanying with bacterial infection, and antibiotics were sometimes prescribed [22]. ESR, erythrocyte sedimentation rate; CRP, C-reactive protein; LD, lactose dehydrogenase; TNF-α, tumor necrosis factor α; IFN-γ, Interferon γ; IL-, interleukin;↑, signi cantly increased; ↓, signi cantly decreased.
It is critical to control the IFS in COVID-19, as shown in Table 1, the levels of broad-spectrum of selected biological markers associated with minocycline administration were opposed to those of COVID-19 condition (both severe and non-severe). In terms of TNF-α, IFN-γ, IL-6 and IL-10, the affected levels by minocycline on physical or infectious challenge obviously opposite to those in COVID-19 condition, which was strongly supported by in vivo and in vitro data. These ndings suggested that minocycline, a safe, inexpensive, and readily available antibiotic, could be considered as an adjunctive therapeutic option for severe and fast-progressive COVID-19 patients.

In silico screening of SARS-CoV-2 M pro inhibitors
There is still no approved speci c therapy for the ongoing COVID-19. For the treatment, screening of potential "old drug" molecules for COVID-19 would be the fastest way. Since old drugs have been prepared, the medication has su cient experience, and the safety and pharmacokinetic parameters are well known, with the in vivo e cacy in animal model, it could be approved by the Green Channel or the hospital ethics committee for clinical use [1].
M pro , which is highly conserved among all coronavirus, is a good target for the development of a single antiviral agent or in combination with other potential therapies to provide an effective rst line of defense against all coronavirus-associated diseases [8]. The co-crystal structure of SARS-CoV-2 M pro complexed with N3 is a good model for identifying inhibitor lead through in silico screening. Based on the "New uses of old drugs" research concept, 135 old drugs were screened targeting on M pro in this study. A group of "old" drugs were screened out, in which the antibiotic drug minocycline showed highest a nity. Docking simulation and the structure-a nity relationship study found critical covalent bond formed between the active Cys145-His41 catalytic dyad and minocycline, which helped us to better understand why the functional groups as well as the tetracycline skeleton could be suitable for the M pro active-site binding and interaction. The in silico screening is just simulation, in vitro and in vivo anti-SARS-CoV-2 evaluations of minocycline will provide useful information.

Biological activities of minocycline
Minocycline is an FDA-approved, second-generation tetracycline class antibiotic with an established safety pro le that has been used in clinic for more than 30 years. Minocycline binds to the bacterial 30S ribosomal subunit, inhibiting the binding of RNA to ribosomes, and interferes with protein synthesis [51]. The main associated pharmacological conditions of minocycline were both gram-positive/negative bacterial infections and the more recent multidrug resistant Acinetobacter baumannii [52]. In spite of this, minocycline appears to have broad-spectrum antiviral activities: reducing West Nile Virus titers in brainderived cell types, reducing Japanese encephalitis-induced damage in neuronal cells inhibiting H7N9 replication in human lung epithelial cells, and attenuating pathogenic immune responses during infection with human and simian immunode ciency virus (HIV/SIV) [48,[53][54][55][56]. Moreover, based on molecular docking and dynamic studies, minocycline was proposed as potential antiviral therapy against Congo Crimean hemorrhagic fever virus to inhibit the binding of virus to host nucleoprotein [57]. In a randomized controlled trial of dengue hemorrhagic fever patients, compared with standard-of-care, combination therapy with doxycycline (analogue of minocycline) signi cantly decreased the TNF and IL-6 levels, and mortality [58].
Literatures showed that the broad anti-IFS activity of minocycline opposed those of many biological markers of COVID-19 condition. In addition, ACE2 (the functional receptor for SARS-CoV-2) is present in multiple human organs including nervous system and skeletal muscle [59]. Severe COVID-19 patients were more likely to develop neurological symptoms [28]. Minocycline attenuates T cell and microglia activity to impair cytokine production in T cell-microglia interaction [45]. Due to the small size and lipophilic nature, minocycline might cross into tissue compartments with potentially therapeutic concentrations.

IFS-suppressing treatment of COVID-19
IFS is an inappropriate immune response that is caused by rapidly proliferating and highly activated T cells, more than 100 in ammatory factors are released, and subsequently lead to tissue damage and organ failure [60]. In ammatory responses triggered by viral infection play a crucial role in pulmonary pathology severity [61]. Suppressing the IFS to reduce lung in ammation might be a valuable treatment method. High doses of glucocorticoid were widely applied during the outbreaks of SARS and Middle East Respiratory Syndrome coronavirus infections to suppress lung in ammation and immune responses [62][63][64]. However, it appeared to be associated with side effects, such as secondary bacterial infection, osteoporosis, etc. Therefore, glucocorticoid was not generally recommended for severe COVID-19 due to its inhibition of immune responses and pathogen clearance [17]. However, the immune imbalance and bacterial infection often appear in the later stages of COVID-19 progression, the e cacy of antiviral drugs might remain unsatisfactory or insu cient [2]. The antibiotics and glucocorticoid were sometimes administered according to the clinical characteristics and physicians' discretion [31].
Adjunctive therapy for anti-IFS but not globally downregulate the host immune response may hold promise for better outcomes of COVID-19. Thalidomide, an immunomodulatory and anti-in ammatory agent, was case-reported the protective effect on lung injury and immunological stress caused by COVID-19 pneumonia in combination with antiviral drugs and low-dose glucocorticoid [17]. The acute pulmonary effusion symptoms and the elevated in ammatory cytokine pro les including IFN-γ, IL-6 and IL-10 were reduced, and the number of lymphocytes recovered [65]. In a two-year randomized controlled trial on early seropositive rheumatoid arthritis patients, minocycline achieved better anti-in ammatory outcomes than hydroxychloroquine [66]. Chloroquine was included in the 6th version of Diagnostic and Treatment Protocol for COVID-19 in China due to the in vitro anti-SARS-CoV-2 results and in vivo anti-in ammatory activity. In case of hydroxychloroquine, it is chemically and biologically similar but safer than chloroquine, and has been suggested by scholars and included in the Shanghai local diagnostic and treatment guidelines for COVID-19 [66,67]. Whether minocycline is better than chloroquine, hydroxychloroquine and thalidomide in COVID-19 deserves a randomized controlled trial.

Minocycline as an adjunctive COVID-19 therapy option
Based on the above ndings, we concluded that minocycline could be considered as an adjunctive therapy option for the treatment as well as the chronic sequela of COVID-19. For one thing, it has anti-IFS activity against biomarkers that appear to be pathologically changed in COVID-19 condition. For another, it could prevent or treat the bacterial/mycoplasma infection along with COVID-19 progression. The most important point is that it has broad-spectrum antiviral activities and potential anti-SARS-CoV-2 activity. Chloroquine (hydroxychloroquine), steroids, and high-dose of urinastatin have been proposed in several COVID-19 treatment guidelines in China for the anti-in ammatory consideration. Minocycline might be a good alternative/adjunctive option of the proposed drugs, and it is likely that the anti-IFS bene t would be more signi cant in severe and fast-progressive infected populations. Anyway, in vitro and in vivo evaluations of minocycline as SARS-CoV-2 inhibitor as well as a randomized controlled trial to investigate the e cacy for COVID-19 treatment will shed new light on an adjuvant treatment strategy for this potentially lethal viral disease.

Conclusion
In conclusion, we did an in silico screening of 135 clinical drugs targeting on M pro of the novel SARS-CoV-2, and the antibiotic minocycline, an inhibitor of bacterial ribosomal rRNA, showed the highest binding a nity (-9.6 kcal/mol). Valuable hydrogen bonding including critical bonds with the Cys145-His41 catalytic dyad and hydrophobic interactions were found between minocycline and M pro active site. The structure-a nity relationship study helped to explain the conformation suitability of minocycline. Literature review found that minocycline had both in vitro and in vivo broad-spectrum antiviral as well as anti-in ammatory activities, and the levels of a broad-spectrum of biological markers during minocycline administration were opposed to those of COVID-19 condition. Minocycline deserves in vitro and in vivo evaluations as SARS-CoV-2 inhibitor as well as a randomized controlled trial to investigate the e cacy for COVID-19. These studies will shed new light on an adjuvant treatment strategy for this potentially lethal viral disease.
Declarations Figure 1 N3 bound in Mpro active site and the diagram of the interactions in the co-crystal.P1, P1', P2, P3, P4 and P5 sites of N3 are indicated, 2Fo-Fc density map is shown aroundN3 molecule in blue mesh, C145-A in yellow mesh, and water in blue mesh [8].
The key residue is shown in stick, hydrogen bond is shown in black dashed line, and water is shown as red sphere.

Figure 1
N3 bound in Mpro active site and the diagram of the interactions in the co-crystal.P1, P1', P2, P3, P4 and P5 sites of N3 are indicated, 2Fo-Fc density map is shown aroundN3 molecule in blue mesh, C145-A in yellow mesh, and water in blue mesh [8].
The key residue is shown in stick, hydrogen bond is shown in black dashed line, and water is shown as red sphere.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.