This in vitro study used commercially available American Type Culture Collection (ATCC) reference strains and did not involve live animals, clinical samples, or human subjects. Therefore, ethical approval was not needed.

The study was conducted from January to March 2025 at the Dairy Production Laboratory, Universitas Jenderal Soedirman, Purwokerto, Central Java, Indonesia.

This study followed a three-phase sequential workflow. First, nine CLO:surfactant:co-surfactant ratios were tested for formulation screening to find candidates that meet the clarity criterion (transmittance >90%), followed by optimization using an I-optimal mixture design. Second, the optimized formulation underwent physicochemical characterization and thermodynamic stability testing. Third, the in vitro antimicrobial activity of the stable CLON against bovine mastitis-associated pathogens was assessed.

The main materials included CLO (from a local commercial distiller, Cilacap, Indonesia), Tween 80 (≥95%, Iberchem, Jakarta, Indonesia), and Polyethylene glycol (PEG) 400 (≥95%, Green Pharma, Selangor, Malaysia). Mueller–Hinton Agar and broth (MHA/MHB), Sabouraud Dextrose Agar (SDA), and RPMI-1640 were sourced from HiMedia (Mumbai, India). Cultures of Staphylococcus aureus ATCC 6538P, Escherichia coli ATCC 25922, and Candida albicans ATCC 14053 (PT. Agritama Sinergi Inovasi, Bandung, Indonesia) were stored at 4°C with monthly subculturing and revived on MHA (37°C, 24 h) for bacteria and SDA (30°C, 48 h) for yeast. Other materials included deionized water (pH 6.8–7.0) and physiological saline (0.9% NaCl).

The instruments used included an ultrasonic bath (35 W, 0.8 L capacity), Digital Hotplate Magnetic Stirrer-Homogenizer H3770-HS (Benchmark Scientific, Sayreville, NJ, USA), Multiskan Sky UV–Vis microplate spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA), Shimadzu GC–MS-QP2020NX (Kyoto, Japan), digital pH meter (Hanna HI991001, Woonsocket, RI, USA), particle size analyzer (Microtrac Nanotrac Wave II, Montgomeryville, PA, USA), JEOL JEM-1400 Transmission Electron Microscope (Akishima, Tokyo, Japan), and Hettich ROTOFIX 32 A centrifuge (Tuttlingen, Germany).

CLO was extracted from fallen leaves of Syzygium aromaticum (L.) Merr. & L. M. Perry, botanically authenticated in January 2026 at the Herbarium of the Biology Faculty, Universitas Jenderal Soedirman, Purwokerto, Indonesia (voucher No. PUNS-Syamsi-001). The collected leaves were washed and sun-dried for 2–3 days, then steam-distilled in a mild steel vessel (water capacity ~500 L, batch size ~200 kg) for 6–8 h at approximately 100°C, producing 1.5%–3.0% oil (dry weight basis). The crude oil was treated with 5% citric acid to chelate iron contaminants, followed by overnight treatment with 10% anhydrous Na₂SO₄ to remove residual water, and then decanted to obtain a pale yellow oil [10].

The essential oil components were analyzed by GC–MS using an HP-5MS capillary column with helium as the carrier gas (1.00 mL/min). Split injection (10:1) was performed at 250°C, and the oven temperature was programmed from 50°C to 250°C at 5°C/min. Compounds were identified in scan mode (m/z 35–600) using the NIST database (Table 1).

CLON was developed using the SNEDDS method [6, 7]. Nine combinations of CLO:Tween 80:PEG 400 (ratios of 1:1:1 to 1:9:1, total volume 10 mL, v/v) were tested in triplicate. CLO was mixed with Tween 80 using a magnetic stirrer (500 rpm, 10 min, 27°C ± 2°C). This moderate stirring speed ensured thorough mixing while reducing air incorporation and thermal degradation of volatile compounds [11].

After adding PEG 400 and mixing for 10 min, the coarse emulsion was sonicated at 40 kHz (power density approximately 0.44 W/mL, based on 35 W output in an 80 mL working volume, at 40°C for 600 seconds). The 40 kHz frequency was chosen because it lies within the optimal range for creating nanoemulsions without causing excessive cavitation damage to essential oil components [12].

Formulations were visually assessed under ambient light [13], and transmittance was measured at 650 nm after dilution in deionized water (1:50 ratio; 200–300 μL per well) [6]. Formulations with clear appearance and transmittance >90% at the lowest surfactant ratio were selected [14].

The chosen formulation was optimized using an I-optimal mixture design with three components: X₁ (CLO), X₂ (Tween 80), and X₃ (PEG 400), constrained to a total of 100%. The responses measured were transmittance (Y₁, aiming for maximum), pH (Y₂, aiming for 6.8–7.2), and emulsification time (Y₃ aiming for less than 1 min). Data were analyzed with Design-Expert software version 13 [15, 16]. Model selection employed a linear sequential approach:

Transmittance was measured as described above. The pH was measured at 27°C ± 2°C using a calibrated digital pH meter (buffers pH 4.0 and 7.0) with a 5 mL sample. Emulsification time was determined by mixing 100 μL of CLON with 50 mL of physiological saline (0.9% NaCl, pH 7–7.4) under stirring at 100 rpm (37°C), and the time needed for the solution to become clear was recorded [17].

The optimized formulation was prepared as described in the pre-formulation step and stored at 27°C ± 2°C for characterization within 24 h.

Droplet size (DS), polydispersity index (PDI), and zeta potential (ZP) were measured with a particle size analyzer (PSA). Samples were diluted in deionized water (1:100 v/v) and gently homogenized (350 rpm, 30–60 s). Measurements were conducted in triplicate at 25°C using a scattering angle of 173° (backscatter), with deionized water as the dispersant (viscosity: 0.8872 cP; refractive index: 1.330).

Morphological characteristics were examined using transmission electron microscopy (TEM). Samples were diluted (1:1000) in deionized water, deposited onto Formvar-carbon-coated 300-mesh copper grids (Ted Pella, USA), and negatively stained with 2% phosphotungstic acid for 30 s before air-drying. Imaging was performed at an accelerating voltage of 100 kV with at least five randomly selected fields [16].

Physical stability was tested by centrifugation (1,006 × g, 30 min). Thermodynamic stability was evaluated through heating–cooling cycles (alternating between 4°C and 45°C, 48 h each, over three cycles) and freeze–thaw cycles (alternating between −4°C and 25°C, 48 h each, over three cycles). The temperature range of −4°C to 45°C was chosen to reflect typical farm storage conditions rather than extreme laboratory temperatures. Formulations had to pass all tests in sequence, and the absence of creaming, phase separation, or cracking confirmed stability [6]. This study centered on the development of nanoemulsion concentrates, while long-term storage stability and shelf-life modeling will be explored in future research.

Optimization data were analyzed using Design-Expert software version 13. Model adequacy was assessed through analysis of variance and lack-of-fit tests. Data normality was tested with the Shapiro–Wilk test. Data that were normally distributed are presented as mean ± standard deviation.

Antimicrobial assays were conducted in triplicate and repeated across three independent experiments. Inhibition zone data were analyzed using one-way analysis of variance followed by Duncan’s multiple range test. A paired t-test was employed to compare predicted and experimental values for model validation. MIC and MBC values were reported as the mode from triplicate determinations. Killing-time data were analyzed using regression assuming first-order kinetics.

The initial screening of nine formulations with varying ratios of CLO, Tween 80, and PEG 400 revealed a range of visual appearances and transmittance values (Table 2). The first- to third-formulations appeared milky white to turbid and had low transmittance values (1.84%–54.53%), indicating unstable emulsions. In contrast, formulations four to nine appeared clear with high transmittance values (90.02%–91.69%), confirming the formation of transparent nanoemulsions. Among these, formulation four (ratio 1:4:1) was chosen for further optimization because it achieved the desired clarity with the lowest Tween 80 concentration.

The I-optimal mixture design was used to define the design space for formulation optimization. The lower, target, and upper limits of the independent variables were set based on component ratios of 1:3:1, 1:4:1, and 1:5:1. The lower boundary was established at 14.29%, and the upper boundary at 71.43%. A total of 17 experimental runs were performed (Table 3). Statistical analysis of the model and residual responses (Table 4) showed highly significant effects (p < 0.01) for all responses, including transmittance, pH, and emulsification time. The lack-of-fit test was not significant (p > 0.05), indicating good agreement between predicted and experimental values. Additionally, the predicted R² values closely matched the adjusted R² values for all responses (difference <0.2), confirming the model’s good predictability. The linear model equations revealed that Tween 80 (B) had the highest positive coefficient for transmittance, while CLO (A) had the greatest impact on emulsification time.

Diagnostic plots for transmittance (Figure 1a), pH (Figure 1b), and emulsification time (Figure 1c) confirmed that the fitted linear models met key statistical assumptions. Table 5 shows the results of a single-sample t-test comparing predicted and experimental values to evaluate the predictive accuracy of the optimized formulations. The pH model (predicted: 7.11; actual: 7.13 ± 0.01; p = 0.236) and the emulsification time model (predicted: 35.91 s; actual: 35.46 ± 0.08 s; p = 0.072) demonstrated excellent agreement. Although the transmittance model showed a statistically significant difference (p = 0.021), the experimental value (93.36 ± 0.25%) was higher than the predicted value (87.95%), indicating improved optical clarity. Figure 2 displays the response surface for emulsification time, illustrating the connection between formulation variables and the response.

Among the 17 experimental runs, formulations that met all acceptance criteria (transmittance ≥90%, pH 7.0–7.18, and emulsification time ≤60 seconds) were identified as samples 4, 5, 10, 12, 13, 16, and 17, making up 41.2% of the design space. These acceptable formulations had compositional ranges of 14.29–16.70% CLO, 65.23–70.82% Tween 80, and 14.29–20.00% PEG 400. The remaining 10 formulations (samples 1, 2, 3, 6, 7, 8, 9, 11, 14, and 15) failed to meet one or more specifications, mainly because their transmittance values were below 90%.

The optimized CLON formulation showed an average DS of 25.19 ± 2.31 nm, confirming the formation of a nanoscale delivery system with a high surface area. The polydispersity index (PDI) was 0.30 ± 0.04, indicating a fairly narrow and uniform DS distribution. The ZP was −45.07 ± 8.15 mV, implying strong electrostatic repulsion between droplets and supporting the stability of the nanoemulsion system.

Thermodynamic stability tests showed that the optimized formulation remained stable through all heating–cooling cycles, centrifugation, and freeze–thaw tests. No creaming, phase separation, or cracking was observed, indicating strong resistance to thermal and mechanical stress. These findings suggest that the formulation is suitable for storage in farm conditions where temperature fluctuations and handling are common (Figure 3). Morphological analysis using TEM revealed spherical nanoemulsion droplets with scale bars of 50 and 200 nm (Figure 4).

The antimicrobial activity of CLON against bacteria associated with bovine mastitis showed promising results. CLON concentration had a highly significant impact (p < 0.01) on the inhibition zone diameter for all tested microorganisms (Table 6). Post hoc analysis with Duncan’s multiple range test revealed distinct groups marked by superscript letters.

Lower CLON concentrations resulted in smaller inhibition zones. Concentrations below 60% showed gradually smaller zones, ranging from about 4.02 to 7.63 mm. In contrast, higher concentrations (60%–100%) produced stronger inhibition zones ranging from approximately 9.80 to 16.25 mm. CLON at 100% consistently created the largest inhibition zones and was not significantly different from povidone iodine (PV-I) for E. coli and C. albicans (sharing superscript “a”). Additionally, CLON at 80% showed antimicrobial activity comparable to PV-I against S. aureus (both marked with superscript “b”). The excipient control and CLON at 0% showed no inhibitory activity (0.00 mm), confirming that the antimicrobial effect came from the clove leaf oil component. Based on these results, CLON at 100% was chosen for subsequent MIC and MBC tests.

The MIC and MBC values of CLON were 12.5% and 50% (v/v), respectively, for all tested microorganisms (Table 7). The MIC indicates the lowest concentration that produces ≥90% growth inhibition, while the MBC indicates the concentration that results in ≥99.9% microbial killing. The MBC/MIC ratio of 4:1 classifies CLON as bactericidal. In comparison, PV-I showed lower MIC (0.13–0.25%) and MBC (0.5%) values, corresponding to approximately 50–96-fold higher potency for MIC and 100-fold higher potency for MBC. The consistent MIC and MBC values across Gram-positive (S. aureus), Gram-negative (E. coli), and yeast (C. albicans) species indicate broad-spectrum antimicrobial activity. Based on these results, the MIC value (12.5%) was selected as the reference concentration for the killing-time assay, with treatments set at 0% (negative control), 1× MIC (12.5%), 2× MIC (25%), and 3× MIC (37.5%).

The killing-time assay showed concentration- and time-dependent antimicrobial kinetics for all tested microorganisms (Figure 5, Table 8). The negative control (CLON 0%) exhibited increasing microbial counts over time, confirming active microbial growth without treatment. At 1× MIC (12.5% CLON), moderate antimicrobial activity was observed with log₁₀ reductions of 3.73 ± 0.01, 2.97 ± 0.10, and 2.98 ± 0.10 log₁₀ CFU/mL for S. aureus, E. coli, and C. albicans, respectively, after 60 min. At 2× MIC (25% CLON), enhanced killing was seen with reductions of 4.12 ± 0.17, 3.76 ± 0.20, and 3.73 ± 0.14 log₁₀ CFU/mL, surpassing the ≥3 log₁₀ CFU/mL reduction threshold required for disinfectant effectiveness. At 3× MIC (37.5% CLON), the highest antimicrobial activity was observed, with reductions of 5.20 ± 0.05, 5.14 ± 0.12, and 5.06 ± 0.10 log₁₀ CFU/mL, nearing the efficacy of PV-I 1% (5.80 ± 0.08, 5.61 ± 0.14, and 5.34 ± 0.16 log₁₀ reductions).

S. aureus showed the fastest killing rate among the tested microorganisms, followed by E. coli and C. albicans, consistent with the susceptibility pattern observed in the inhibition zone assay. Analysis of the AUKC, which combines both the magnitude and rate of killing over the 60-min exposure, confirmed this concentration-dependent trend. PV-I 1% exhibited the lowest AUKC values (123.58–157.61 log₁₀ CFU min./mL), followed by CLON 37.5% (162.48–172.70 log₁₀ CFU min./mL), CLON 25% (213.57–233.13 log₁₀ CFU min./mL), and CLON 12.5% (233.12–262.76 log₁₀ CFU min./mL).

The initial screening of the oil and Smix (surfactant and co-surfactant) ratios in the CLON formulation (Table 2) showed that transmittance increased with higher Tween 80 concentrations. The 1:4:1 ratio produced transmittance values above 90% at the lowest Tween 80 level among the acceptable formulations and was therefore chosen for further optimization. Kaur et al. [19] similarly reported that an oil:Tween 80 ratio of 1:4 was optimal for nanoemulsion formation. Lechuga et al. [20] emphasized that lowering surfactant concentration is important for enhancing the safety profile of pharmaceutical formulations. Thus, the selected formulation achieved adequate emulsification while balancing safety and functional performance. Jia et al. [21] further indicated that optimizing oil and Smix ratios should aim to produce stable, environmentally adaptable, and cost-effective emulsions.

The I-optimal mixture design was chosen because it reduces the average prediction variance across the design space and is particularly well-suited for formulation optimization studies that require accurate prediction of response variables [15]. The optimization experiment produced formulations with transmittance values ranging from 55.34% to 93.54% (Table 3), showing that only specific compositional regions yield clear nanoemulsions. Statistical analysis verified that the developed linear model was valid and significant, with a non-significant lack-of-fit test (Table 4). Additionally, the close match between predicted R² and adjusted R² values (difference <0.2) confirmed that the model has adequate predictive capability without overfitting [22].

The model equations (Table 4) revealed that Tween 80 had the most significant impact on transmittance due to its role in lowering interfacial tension between oil and water phases. Conversely, CLO primarily influenced variations in emulsification time because higher oil concentrations require more surfactant coverage for stable emulsification. These results show that the differences among formulations reflect the system’s natural sensitivity to compositional changes rather than a flaw in the formulation strategy. Such predictive robustness is crucial for guiding scale-up and industrial production [23].

From a practical standpoint, formulations with CLO levels above approximately 18% consistently failed the transmittance requirement because the available surfactant was not enough to fully stabilize the oil phase. Similarly, reducing Tween 80 concentrations below about 63% led to incomplete emulsification and longer processing times. These results highlight Tween 80 as the most critical component that needs precise control during scale-up manufacturing.

The pH range of the optimized formulations (7.02–7.16) is especially relevant for mastitis teat disinfection because it closely matches the physiological pH of bovine teat skin (7.00–7.18), as reported by Fox et al. [25]. Pour et al. [24] observed that topical disinfectant formulations should stay compatible with skin pH to reduce irritation. In contrast, conventional iodophor-based teat disinfectants usually have acidic pH levels (3–6), which may cause skin irritation during long-term use [26]. Therefore, the near-neutral pH of CLON offers a significant advantage in preserving skin integrity during repeated applications.

The emulsification time ranged from 35.4 to 37.4 s, indicating rapid and efficient nanoemulsion formation. According to Taha et al. [27], longer emulsification times are generally associated with reduced emulsion stability because they reflect difficulty in achieving uniform phase dispersion. The short emulsification time observed in this study, therefore, indicates favorable emulsification kinetics and formulation stability.

Diagnostic plots further verified the reliability of the statistical models (Figure 1). The standardized residuals exhibited a linear distribution pattern, indicating normality, while residual-versus-run plots displayed a random distribution, confirming independence and homoscedasticity of the residuals [28]. These statistical diagnostics support the robustness of the developed formulation models.

The optimized CLON formulation showed a mean DS of 25.19 ± 2.31 nm, confirming the successful formation of a nanoemulsion system. The polydispersity index (0.30) indicated a relatively narrow DS distribution, while the strongly negative ZP (−45.07 ± 8.15 mV) suggested robust electrostatic repulsion between droplets. These features are consistent with previously reported clove oil nanoemulsions, such as those described by Shahavi et al. [29].

Ultrasonication likely played a significant role in reducing DS. While mechanical stirring alone generally produces droplets in the 100–500 nm range, ultrasonication can markedly decrease DS through acoustic cavitation effects. Previous studies have reported droplet sizes between 68 and 189 nm with ZP values from −44 to −55 mV when sonication is used [30, 31].

Although Tween 80 is a non-ionic surfactant, the strongly negative ZP observed in CLON may result from hydroxyl ion adsorption at the oil–water interface or from anionic impurities like free fatty acids in the oil phase, rather than surfactant ionization itself [32, 33]. PEG 400 further increased system stability by providing steric stabilization, which improved droplet uniformity and prevented coalescence [5, 34].

Thermodynamic stability testing showed that CLON remained stable under freezing (−4°C), heating (45°C), and centrifugation stress conditions. No creaming, cracking, or phase separation was observed, indicating strong physical stability. This stability is especially important for farm-level applications where products might face varying storage and handling conditions.

TEM images showed spherical droplets with a clear shape (Figure 4). Small droplet clusters seen in the micrographs probably resulted from dehydration artifacts during sample preparation, not from instability in the emulsion system. This is supported by the strong ZP and excellent thermodynamic stability observed during characterization [35–37].

CLON concentrations from 60% to 100% caused moderate to strong inhibition of the tested microorganisms. According to Humphries et al. [38], inhibition zones larger than 14 mm indicate sensitivity, while zones between 9 and 14 mm show intermediate susceptibility. The lack of antimicrobial activity in the excipient control confirmed that the inhibitory effect was due to CLO, not the surfactant system.

Differences in antimicrobial susceptibility among the tested microorganisms can be explained by structural variations in their cell envelopes. Gram-positive bacteria, such as Staphylococcus aureus, have thick peptidoglycan layers but lack an outer membrane, making them more vulnerable to lipophilic antimicrobial compounds like eugenol. In contrast, Gram-negative bacteria, such as Escherichia coli, have an additional outer membrane containing lipopolysaccharides that serves as a permeability barrier against hydrophobic antimicrobial agents [39, 40]. The yeast Candida albicans has a complex cell wall made up of chitin, β-glucan, and mannoproteins, which can further restrict antimicrobial penetration [41].

The nanoscale DS probably enhanced antimicrobial activity by increasing surface area, improving water dispersibility of hydrophobic compounds, and facilitating interaction with microbial membranes [41, 42]. Previous studies have shown that nanoemulsification can lower MIC values of essential oils by about 50% compared to non-nanoemulsified oils [43, 44].

The killing-time assay showed clear concentration-dependent antimicrobial kinetics. A 25% CLON concentration (2× MIC) achieved over a 3-log₁₀ CFU/mL reduction within 60 minutes for all tested microorganisms, meeting the efficacy threshold commonly used for disinfectant evaluation [45]. Faster killing of S. aureus compared to E. coli aligned with structural susceptibility differences between Gram-positive and Gram-negative bacteria. The rapid bactericidal activity observed aligns with how eugenol works, which damages cell membranes and causes quick leakage of internal contents instead of slow metabolic inhibition [41, 42, 46]. AUKC analysis also confirmed the antimicrobial effect depends on concentration, with CLON 37.5% nearly matching the activity of 1% povidone iodine. Considering both antimicrobial efficacy and practical formulation considerations, the 25% CLON concentration offers a promising balance between antimicrobial performance and resource efficiency. This finding is especially relevant for teat disinfection practices, where short contact times are commonly used during pre-milking sanitation.

The antimicrobial activity of CLON results from its complex phytochemical composition. GC–MS analysis showed that eugenol (72.90%) was the main compound, followed by β-caryophyllene (12.00%), eugenyl acetate (3.84%), caryophyllene oxide (2.12%), α-copaene (1.88%), and various minor terpenoids (Table 1). Eugenol mainly works by disrupting membranes, inhibiting ATPase, producing reactive oxygen species, and inactivating enzymes (Figure 6).

Additional phytochemicals may work together synergistically. β-Caryophyllene may improve membrane permeability and influence inflammatory responses, while eugenyl acetate and caryophyllene oxide may provide additional antimicrobial actions [46, 47]. This multi-compound makeup offers a potential advantage of plant-derived antimicrobials, as targeting multiple cellular pathways simultaneously decreases the chances of resistance developing [3].

Overall, the optimized CLON formulation showed favorable physicochemical properties, strong thermodynamic stability, and broad-spectrum in vitro antimicrobial activity against major microorganisms associated with bovine mastitis. These findings suggest that CLON is a promising plant-derived alternative for teat disinfection.

However, further research is needed before practical application can be considered. Clinical mastitis isolates may exhibit different susceptibility patterns compared to reference strains, and in vivo studies are essential to determine whether the in vitro activity observed translates into effective mastitis prevention. Future work should also compare the effects with non-nanoemulsified CLO to measure the specific enhancement from nanoemulsification, evaluate shorter contact times relevant to field conditions, and assess the economic feasibility for large-scale dairy farm use.