**2. Materials and methods**

## **2.1 QIs determination**

Meta-analysis for QIs determination was carried out by scanning scientific literature in search of *in vitro* studies, reporting levofloxacin, ofloxacin, netilmicin, tobramycin and azithromycin MIC90 values against the most diffused bacterial species in ocular infections (*MSSA, MRSA, MSCoNS, MRCoNS*), and pre-clinical *in vivo* studies in rabbits, reporting corneal and conjunctival Cmax values of ophthalmic formulations containing the above antibiotics. In detail, as described in **Tables 1** and **2**, the following eye drops formulations were taken into account: (i) netilmicin 0.3% eye drops solution (Nettacin® eye drops); (ii) Netilmicin 0.3% gel formulation (Xanternet®); (iii) tobramycin 0.3% (Tobrex®); (iv) levofloxacin 1.5% (Iquix® and Cravit®);( v) ofloxacin 0.3% (Tarivid® and Ocuflox®); and (vi) azithromycin 1% (Azasite® and Azimycin®).

As to the MIC90, the selected references adhered to the criteria of encompassing large epidemiological/surveillance studies and including the above-cited groups of bacterial strains. As to the second parameter (Cmax), only those studies comprising the same animal species and the same regimen of ocular drug administration (single or multiple doses) were considered as eligible and comparable to each other. Then MIC90 and Cmax data were put into ratio and the QI values for the two ocular tissues and regimens were calculated.

Moreover, the same analysis was performed in human tears, comparing netilmicin 0.3% eye drops solution with tobramycin 0.3%, ofloxacin 0.3% and levofloxacin 0.5%, all in single-dose administration.


## **Table 1.**

*Cmax, MIC90 values and relative bibliographic references as to the single dose of antibiotics administration.*

## **2.2 Bacterial strains and killing curves determination**

Two eye swabs of bacterial isolates *S. aureus* 801 CT (*MRSA*) and *S. epidermidis* 829 CT (*MRSE*), belonging to SIFI's private collection were used throughout this study. These strains collection comprises different microbial eye isolates collected from 1998 to 2014 from patients with community-acquired ocular infections coming from three different geographical sites (Messina, Catania and Rome (Italy)). The above microorganisms, and their antibiotic resistance patterns, were previously identified by standard laboratory methodologies. Genetic profiles of all *S. aureus* and *S. epidermidis* isolates of the collection were previously determined (data not shown) with the aim of grouping all isolates in 'strain types'. The two methicillin-resistant microorganisms under study were each representative of the most diffused 'strain type' within the collection with regards to species and the geographical provenience. In addition, *S. aureus* ATCC 6538 and *S. epidermidis* ATCC 12228 strains (purchased from Thermo Fisher Scientific, Lenexa, KS, USA) were also included in the experiment. Time kill determination was performed as previously described by Bonfiglio *et al*. [3]. Briefly, microorganisms were grown overnight to reach a density of about 108 colony-forming unit (cfu)/ml



## **Table 2.**

*Cmax, MIC90 values and relative bibliographic references to multiple doses of antibiotics administration.*

on Mueller Hinton (MH) broth (purchased from Biogenerica, CT, Italy) and then diluted to about 1×106 cfu/ml in the same pre-warmed medium containing netilmicin (32 μg/ml; purchased from Zhejiang Zhenyuan Pharmaceutical Co., Ltd, Shaoxing, China), tobramycin (16 μg/ml; purchased from Sigma Aldrich, MI, Italy), ofloxacin, levofloxacin (4 μg/ml; purchased from Sigma Aldrich, MI, Italy), azithromycin (8 μg/ ml, purchased from Sigma Aldrich, MI, Italy) or chloramphenicol (32 μg/ml, Chemo, Spain). These antibiotics concentrations corresponded to the resistance breakpoints values according to CLSI's (2015) interpretative criteria. An antibiotic-free control was similarly inoculated. At 0, 1, 2, 6 and 24 hours after drug exposure, 0.1 ml of the culture was collected, diluted in phosphate buffered saline, inoculated (by inclusion) onto MH agar plates and incubated at 37° C for 24 hours to determine viable cfu/ml. All the experiments were performed in triplicate for three experimental days. Killing curves were constructed by plotting the log10 of cfu/ml versus the established time points (i.e 0, 1, 2, 6 and 24 h). Bactericidal activity was defined as a >3 log10 decrease of the initial inoculum size. A statistical analysis of the area under the curve (AUC) was conducted by the one-way ANOVA statistical test (GraphPad Prism 6; CA, USA).
