A chloramphenicol-resistant strain of the avirulent C. burnetii Nine Mile II isolate (NMII; RSA439; clone 4) was propagated in acidified citrate cysteine medium-2 (ACCM-2), as described previously (Omsland et al., 2011; Sanchez, Vallejo-Esquerra and Omsland, 2018). C. burnetii was stored at −80°C in ACCM-2 supplemented with 10% DMSO. For the enumeration of microbial counts, the samples were diluted serially in ACCM-2 inorganic salts before plating on solid (final 0.25% w/vol. agarose) ACCM-2 supplemented with 500 μM tryptophan and 1.5 μg/mL chloramphenicol. The plates were incubated for 9–10 days in a tri-gas incubator at 37°C with 5% CO₂ and 5% O₂ (Sanchez, Vallejo-Esquerra and Omsland, 2018). To correlate the number of viable bacteria as measured by the plate count method to the total number of bacteria used for inoculation, C. burnetii genome equivalents (GEs) were quantified via the detection of the C. burnetii gene CBU1206 using a CFX96 real-time PCR detection system (Bio-Rad Laboratories, Hercules, CA) (Beare et al., 2012; Sanchez, Vallejo-Esquerra and Omsland, 2018). C. burnetii GEs were extrapolated from a standard curve prepared using recombinant CBU1206. A non-pathogenic strain of Escherichia coli ATCC 25922 and Salmonella enterica serovar Muenchen ATCC BAA 1674 were obtained from the American Type Culture Collection (ATCC) and propagated in Tryptic Soy Broth (TSB) and harvested as reported earlier (Pendyala et al., 2021; Vashisht et al., 2021). To enumerate the microbial population, appropriate dilutions in peptone water (in 0.1% PW) were plated in duplicate onto Tryptic Soy Agar (Oxoid Ltd., Basingstoke, United Kingdom) plates and incubated for 24 h at 37°C. UV-C sensitivity of C. burnetii was hypothesized to be in the range similar to Escherichia coli and Salmonella (Gopisetty et al., 2019; Vashisht et al., 2021).

Ultra-high-temperature (UHT) processed SM from Parmalat, Canada; humic acid (adjusted pH to 7.0) from Agricultural Services of America, Inc., Florida, United States; and phosphate buffer saline (PBS, pH 7.0) were used as test fluids. To remove ACCM-2 and DMSO from the test samples, C. burnetii stocks were washed twice with phosphate-buffered saline (Becton Dickinson, New Jersey, United States) using centrifugation at 3000 g for 15 min. The test samples were prepared by spiking SM with microorganisms (C. burnetii, E. coli, or S. Muenchen) at a concentration of >10⁸ CFU/mL for testing by UV-C irradiation or HTST pasteurization. Each strain was separately inoculated in a sample and treated independently.

The optical properties of SM inoculated with test micro-organism were measured using a double beam Cary 100 spectrophotometer (Varian, United States) equipped with a 6-inch single integrating sphere (Labsphere, DRA-CA-30, United States) to calculate scattered light at a 254 nm wavelength (Shenoy and Pal, 2008). Thin quartz cuvettes (0.08 mm path length) were prepared and used to measure total transmittance and total reflectance values. The transmittance and reflectance (diffuse reflectance) of light were collected by the integrating sphere when the sample was placed at the entrance and exit ports, respectively (Gunter-ward et al., 2018). The amount of light transmitted and reflected by the quartz cuvettes was also quantified and considered to estimate the absorption, scattering coefficients, and reflectance. The refractive index (RI) of SM was measured at 20°C, using a digital refractometer from Schmidt + Haensch GmbH & Co., at nine distinct wavelengths ranging from ultraviolet to the visible range, i.e., 365 nm–706 nm. For each measurement, 0.3 mL of SM was placed in the measurement compartment. Refractive index data were compared with Milli-Q water (control). A fifth-order polynomial fit was then used to calculate the refractive index at 254 nm. A higher-order polynomial model was selected as it fitted the experimental data well with a R ² value of 0.99. For a SM sample, a control was run to check the system calibration. The refractive index of Milli-Q water was found to be 1.376, which is in excellent agreement with the scientific literature, indicating that the system was well-calibrated (Schiebener et al., 1990). The accuracy of the measurements is within 0.002% at each wavelength (data not shown). Data were reported as means ± standard deviation. All measurements were carried out in duplicates (technical replicates) with three biological replicates. The inverse adding doubling (IAD) program (GitHub, CA, United States, 94107) was used to quantify absorption and reduced scattering coefficients by applying total transmittance, reflectance, and the refractive index as input values (Prahl, 1999). IAD is a command-line program which uses the inverse adding doubling method and includes the Fresnel reflection at the surface and corrections along with integrating sphere experiments. The inverse adding-doubling method is an accurate solution of the radiative transport equation for all albedos, all optical depths, and all phase functions, this technique can be applied to any fluid medium for which the radiative transport equation is valid. This method is applicable to homogeneous turbid slabs with any optical thickness, albedo, or phase function. The optical properties are obtained by iterating an adding-doubling solution of the radiative transport equation (Eq 2) until the calculated values of the reflection and transmission match the measured ones. From the obtained optical properties, Ultraviolet transmittance (UVT, %/cm), which indicates the fraction of the incident light transmitted through a fluid over a 1 cm path-length, was calculated as per Eq 1. UVT % / cm = 10 − A × 100 , (1) where A represents the absorbance (base₁₀) of the test fluid at 254 nm for a 1-cm path.

The collimated beam apparatus (Pendyala et al., 2019) was used to irradiate test fluids (Figure 1). The system design followed the recommendations of Bolton and Linden (2003) and utilized a low-pressure mercury vapor arc lamp, primarily emitting at 253.7 nm (positioned at 18.05 cm from the surface). Irradiance at the position of the fluid surface was measured using a calibrated radiometer ILT1700 with SED240 detectors, each equipped with a quartz W diffuser and an NS254 spectral filter to ensure that only 254 nm radiation was measured (International Light Technologies, Peabody, MA, United States). The delivered UV-C dose (fluence) was calculated as the product of the volume average of the fluence rate in the sample and the exposure time by assuming a perfect mixing of the sample by the stir bar (12.7 x 3.2 mm). A measure of 2 mL of the microbial suspension (tested in SM, humic acid, or PBS) was treated in 10-mL beakers (optical path length of 6 mm) at UV dose ranges from 0 to 17 mJ/cm² (n = 3). Since SM scatters UV-C light, the scattering factor ( S ) was estimated by the radiative transport equation (Eq 2) using computational fluid dynamics (CFD). d I r → , s → d s + a + σ S I r → , s → = σ S 4 π ∫ 0 4 π I r → , s ′ → Φ s → ⋅ s ′ → d Ω ′ , (2)

where I represents the radiation intensity, r → represents the position vector, s ′ → represents the scattering direction vector, s represents the path length, a represents the absorption coefficient, n represents the refractive index, σ S represents the scattering coefficient, σ represents the Stefan–Boltzmann constant, Φ represents the phase function, and Ω ′ represents the solid angle.

Then, the average UV fluence rate ( E ′ a v g and, thereby, the delivered UV dose D in the stirred sample was calculated by Eq 3 (Pendyala et al., 2022): E ′ a v g = E 0 × P f × 1 − R × L d + L × 1 − 1 0 − A d A d ln 10 × S . (3) E 0 represents the radiometer meter incident irradiance reading at the center top surface of the water in the beaker; P f represents the Petri factor; 1 − 1 0 − A d A d ln 10 represents the water factor; 1 − R represents the reflectance factor; L d + L represents the light divergence factor. D e l i v e r e d   U V   d o s e   D = E ′ a v g × t . (4) t represents the exposure time in seconds.

CFD simulation was carried out with the commercial code Ansys 2021 R1. Fluent software was used for drawing geometry (2D) and generating the mesh, while Ansys Fluent v12.1 (Ansys Inc., Canonsburg, PA, United States) was used for solving the radiation fields. The mesh resolution nodes and elements were 3643 and 36000, respectively. Figure 1 shows the schematic representation of the geometry of the system adopted for the numerical simulations. This technique used the discrete ordinates irradiation model; the fluid (2 mL) in a 10-mL beaker was exposed to UV incident radiation coming from a near collimated beam apparatus. In the near collimated beam apparatus, the local disinfection depends on the local fluence rate and kinetic rate constants. Modeling the collimated beam photo-reactor, therefore, involves two sub-models, the intensity of the radiation field and UV reaction kinetics.

The thermal challenge studies were conducted using a laboratory-scale HTST pasteurizer (Kontopodi et al., 2022). The pasteurizer consists of three main sections (as shown in Figure 2), a pre-heating section (77°C), a holding section (72°C for 15 s), and a cooling unit. The temperature was monitored using a probe. SM inoculated with C. burnetii was directly fed into the system using a peristaltic pump (Watson–Marlow). The pump was calibrated before the thermal challenge studies. The clean-in-place (CIP) procedure was conducted before any thermal challenge studies. The cleanliness and sterility of the HTST system were ensured by performing a CIP procedure before and after use. This CIP consisted of flushing the system with sterile water, followed by 0.1 N HCl, sterile water again, 0.1 N NaOH, and a final rinse with sterile water. Flushing was performed at 180 mL/min. SM was then pumped through the heating time, where it achieved the pasteurization temperature of 72°C at a flow rate of 90 mL/min at a 15-s holding time. At the end of the holding section, SM passed through a cooling coil. In this phase, SM was cooled to a final temperature of 4°C ± 1. A 10 mL of sample was then collected for enumeration and plating.

To assess the inactivation of E. coli or C. burnetii, the log-linear model available in the GInaFiT tool (a freeware add-in for Microsoft Excel) (Geeraerd, Valdramidis, and Van Impe, 2005) was used to fit the experimental data, and the goodness-of-fit parameters, including R ², root-mean-square error, and rate constants, were evaluated. Inactivation kinetics is expressed as follows: log ⁡ N = Log   N 0 − K max × D u v ln ⁡ 10 , (5) where N and N 0 represent the final and the initial cell numbers, respectively; K max represents the inactivation rate at the highest dose; the UV dose is represented by D u v . D₁₀ represents the decimal reduction value (10% survival in the microbial population), which was expressed as follows: D 10 = LN 10 K max . (6)

A balanced design with four replicates randomized in the experimental order was performed for each treatment. Data were reported as means ± one standard deviation from the mean and significance level set to 0.05 (5%).

Data on optical properties indicate that SM strongly absorbs and scatters UV-C light (Table 1). From the measured optical data, it is apparent that UV-C light has minimal transmission through SM due to presence of aromatic amino acids and other UV-C absorbing organic solutes (Gunter-ward et al., 2018). Light scattering by casein micelles causes skim milk to appear turbid and opaque. The average fluence rate or incident irradiance through the test fluid suspension was calculated by substituting the experimental Petri factor, reflection factor, water factor, divergence factor, and the scattering factor in Eq.3. P f refers to the ratio of the average incident irradiance over the Petri dish area to the irradiance at the dish’s center. The water factor used here accounted for the absorption of UV photons by test fluid. A in the equation stands for the absorption coefficient, and d stands for the sample depth. The reflectance factor demonstrates the part of incident UV photons that enter the test fluid. For a finite distance of the UV lamp, the beam cannot be perfectly collimated; hence, the divergence factor was used for its accountancy. These factors were explained by Bolten and Linden (2003). In addition, the scattering factor was accounted in the dose calculations. In our studies, the values of P_f, water, reflectance, divergence, and scattering factors were 1, 0.038, 0.894, 0.968, and 0.92, respectively (Table 1). These values were similar to the previous studies (Pendyala et al., 2022; Vashisht et al., 2021; Vashisht et al., 2022). The numerical results of UV intensity contours are shown in Figure 1. The results clearly show the absorption of UV photons by the fluid. Without proper mixing, the fluid further from the lamp will receive a lower dose than that which is closer to the free surface. This is especially true for a liquid of low transmittance (SM), which results in a strong intensity gradient in the fluid. The volume average incident radiation was 0.0153 mW/cm². The delivered UV-C dose (fluence) in a stirred near-collimated beam apparatus was calculated as the product of the volume average of the fluence rate in the sample and the exposure time in seconds (Eq 4) by assuming the perfect mixing of the sample by the stir bar. The rotation speed (angular velocity) of the stirrer was optimized to enhance mixing.

To verify the estimated scattering factor from CFD, comparative microbial (E. coli) inactivation studies in SM and humic acid with the same UV-C absorbance as SM without scattering were conducted. Humic acid is a stronger absorber of UV light and has been used as a surrogate fluid (Pendyala et al., 2019). The data show that there is no significant difference in the microbial inactivation kinetics in both SM and diluted humic acid (Figure 3) and confirms the accuracy of the calculated scattering factor. The dose distribution throughout the fluid domain is a crucial parameter to estimate the UV-C sensitivity of microorganisms. The efficient dose distribution conditions, provided to all microbial particles, result in the log-linear inactivation kinetics in mono-microbial populations. To check the UV dose distribution in SM under standard experimental stir bar mixing conditions, a comparative study with a high UV-C transparent fluid phosphate-buffered saline (PBS) was conducted at the same experimental UV-C doses. The experimental data show that there is no significant difference (p >0.05) between the microbial inactivation kinetics with D₁₀ values ranging from 3.20 to 3.32 mJ/cm² and the shown log-linear inactivation kinetics in the three different test fluids (Table 2). Therefore, these results indicated that the experimental mixing conditions distributed the UV-C dose efficiently in opaque test fluids.

SM inoculated with C. burnetii at approximately 10⁸ CFU/mL was exposed to known UV-C doses of 0, 3, 6, and 12 mJ/cm² in a stirred collimated beam UV system. The data revealed a >3 log₁₀ reduction of C. burnetii at a maximum dose of 12 mJ/cm² (Figure 4). The inactivation kinetics of C. burnetii were fitted into a log-linear model with a low (0.35) root-mean-square error (RMSE) and a higher (0.92) R ² value (Table 3). The D₁₀ value and kinetic constant (k _max ) of C. burnetii in SM were estimated as 4.1 ± 0.04 mJ/cm² and 0.56 cm²/mJ, respectively (Table 3). Little, Kishimoto, and Canonico (1980) demonstrated the inactivation of C. burnetii in suspension and within guinea pig peritoneal macrophages by UV-C irradiation. The authors reported that C. burnetii was inactivated at a UV irradiance of 600 μW/cm² for 15 s at a distance of 10 cm in the suspension and macrophages. According to our presented data on the D₁₀ value of C. burnetii, the performance criterion of 5 log₁₀ reductions, as demanded by the Codex Alimentarius, can be achieved at a UV-C dose of 20.5 mJ/cm².

While the inactivation of C. burnetii is a key consideration for milk processing, the cultivation of C. burnetii requires specialized equipment and expertise. Therefore, to facilitate the validation of pilot-scale UV-C systems for the inactivation of bacterial pathogens in milk, we sought to identify a bacterial surrogate for C. burnetii. Based on the D₁₀ value of C. burnetii (4.1 ± 0.04 mJ/cm²), the Salmonella strain Muenchen ATCC BAA 1674 reported with D₁₀ values ranging from 3.9 to 4.3 mJ/cm² in the phosphate-buffered saline (Gopisetty et al., 2019; Pendyala et al., 2021; Vashisht et al., 2021) was selected for UV-C inactivation studies with these organisms in SM (Figure 4). The results show log linear inactivation kinetics with RMSE 0.1913 and R ² >0.98. Interestingly, the data show K _max (0.56) and D₁₀ values (4.1 mJ/cm²) similar to C. burnetii, indicative of their suitability as surrogates for C. burnetii for UV-C inactivation studies in SM. It should be noted that this experiment was designed to evaluate the UV-C sensitivity of C. burnetii using a collimated beam system. Considering the UV-C light distribution in SM, the attenuation of UV-C intensity with the fluid depth, the collimated beam system will not adequately achieve a 99.999% reduction in C. burnetii. In contrast, the system can be used to quantify the D10 value (dose required for 1 log reduction of test micro-organism. Further studies will be conducted using a UV-C continuous flow system to achieve atleast 8 log₁₀ reduction of C. burnetii in milk. For UV-C inactivation studies, Salmonella can be a suitable surrogate for C. burnetii which does not need specialized equipment. Salmonella can be easily propagated in TSB, enumerated, and subsequently plated on TSA plates. These plates are typically incubated at 37°C for a duration of 18–24 h. Based on these characteristics, Salmonella can be an ideal surrogate, particularly when validating UV systems at commercial flow-rates. In our forthcoming field testing, we plan to utilize Salmonella as a surrogate for C. burnetii.

The physical properties of SM and heat exchanger parameters are shown in Table 4. Starting with a C. burnetii population of 3.81 x 10⁸ ± 0.21 CFU/mL (5.38 x 10⁸ ± 0.31 GE/mL), the heat-dependent log₁₀ reduction during a holding time of 15 s at 72°C resulted in 8 log₁₀ reduction of C. burnetii in SM, with a D-value of 1.75 s. It was lower as compared to the value reported in the studies of Enright et al., 1957a and Enright et al., 1957b, where the D value was 1.88 s at 72°C. A recent study by Wittwer et al. (2022) on Coxiella isolates M, WDK299, and WDK1188 indicated that, in general, all isolates were more susceptible to heat over a temperature ranging from 60°C to 65°C with holding times from 15 to 25 s. For the highest temperature of 65°C, the D-value was reported as 5.1–7.6 s, with a predicted reduction of approximately 10.5 log₁₀ at 72.4°C for 15 s.