Inactivation of Avian Influenza Virus in Raw Milk Kefir

TL;DR

Fermenting raw milk into kefir significantly reduces avian influenza virus levels, lowering the risk of consuming contaminated milk.

Contribution

This study demonstrates that kefir fermentation effectively inactivates avian influenza virus in raw milk.

Findings

Fermentation reduced H5N1 and H5N9 virus titers by 4.1 and 3.1 log10 EID50/mL, respectively.

pH reduction during fermentation to ≤5.3 significantly decreased viral levels.

Both spontaneous and starter-culture fermentation achieved similar virus inactivation.

Abstract

Both highly pathogenic avian influenza virus (HPAIV) and its viral RNA have been detected in raw milk during the ongoing outbreak in dairy cows. While pasteurization effectively inactivates the virus in milk, comparatively little has been published on AIV inactivation in fermented dairy products made from raw milk. We evaluated the viability of two isolates of low pathogenic avian influenza virus (LPAIV) during fermentation of kefir to assess the potential for viral persistence. Raw (unpasteurized) milk (ca. 3.0% fat) was inoculated with ca. 7.5 log10 CFU per mL of active commercial kefir starter culture. Next, raw milk (125 mL) was inoculated with 1.0 mL of either strain A/rgGyrfalconHAxPR8/2014 H5N1 or A/turkey/Wisconsin/1968 H5N9 to achieve an average initial level of ca. 5.0 log10 50% egg infectious doses (EID50) per mL. The inoculated raw milk (ca. pH 6.55 ± 0.05) was fermented at 25 °C for ca. 36 h to an average endpoint pH of pH 4.44 ± 0.05. Results confirmed that fermentation of raw milk to produce kefir delivered an appreciable reduction of both LPAIV strains, that being a ca. 4.1 or 3.1 log10 EID50 per mL decrease of strains H5N1 or H5N9, respectively, from their initial levels in raw milk. In the absence of a starter culture, due to fermentation by the indigenous flora the pH of the raw milk after incubation at 25 °C for ca. 36 h also significantly decreased on average to ca. pH 5.3:, and levels of strains H5N1 and H5N9 were reduced by 4.1 and 3.5 log10 EID50 per mL, respectively. Our findings established that fermentation of raw milk (with or without added kefir starter culture) to ≤ pH 5.3 over ca. 36 h at 25 °C appreciably decreases titers (≥ 3.1 log10 EID50 per mL) of AIV strains H5N1 and H5N9 and, therefore, lowers risk associated with its consumption.

Figures

Process used to monitor viability of two strains of low pathogenic avian influenza virus (H5N1 and H5N9) during fermentation of raw milk kefir^A^. Control treatments included: (i) raw milk (125 mL) without the kefir starter culture, but separately inoculated with strain rgGYR/14 or strain A/turkey/WI^B^, (ii) raw milk (125 mL) inoculated with the kefir starter culture, but not inoculated with either of the LPAIV strains^C^, and (iii) raw milk (125 mL) without the kefir starter culture and without either of the LPAIV strains^D^

Full text

Introduction

Despite food safety hazards associated with ingesting raw milk and unpasteurized dairy products, the market for such products continues to expand. Increased consumption of raw milk and fermented raw milk products such as yogurt, kefir, and cheese is likely fueled by its purported health benefits, anecdotal enhanced nutritional quality, and perceived improved taste. Some consumers also value “buying local products” and many consumers prefer “unprocessed, natural foods” (Lucey, 2015; Maughan et al., 2025). From 2021 to 2023, sales of raw milk in the U.S. increased by ca. 27%, going from $12 to$19.4 M (Ventiera, 2023). Moreover, ca. 11 M adults self-reported that they consume raw milk at least once per year, with ca. 1.5 M of them consuming raw milk 1 to 3 times a week, and another ca. 1 M individuals consuming it daily (Lando et al., 2023). Germane to the present study, this relatively small, but resolute, niche of consumers often neglects the fact that unpasteurized milk poses a higher risk to human health because of zoonotic bacterial and viral infections and because consumption of contaminated milk has resulted in numerous illnesses and some deaths (Claeys et al., 2013; Koski et al., 2022; Lucey, 2015; Oliver et al., 2005). Between 1998 and 2018, consumption of unpasteurized milk alone was responsible for 202 outbreaks that resulted in 2,645 illnesses, 228 hospitalizations, and 3 deaths in the U.S., with most outbreaks attributed to Campylobacter spp., Salmonella spp., Shiga toxin-producing Escherichia coli, or Listeria monocytogenes (FDA, 2024; Koski et al., 2022).

Although pasteurization of milk was put into practice in the late 1800’s to eliminate bacterial pathogens such as Brucella spp., Coxiella burnetii, Listeria monocytogenes,* Mycobacterium tuberculosis*, and Salmonella spp. (Currier, 2023; Potter et al., 1984; Schafers et al., 2025), in March of 2024 a potential new public health threat from dairy products arose in the form of highly pathogenic avian influenza virus (HPAIV). More specifically, clade 2.3.4.4b (B3.13 and D1.1 genotypes) H5N1 HPAIV was detected in U.S. dairy cattle (Caserta et al., 2024; FDA, 2025b) and its RNA was detected via RT-qPCR in raw milk and various raw and pasteurized retail dairy products (Alkie et al., 2025; Caserta et al., 2024; Cui et al., 2024; Nooruzzaman et al., 2025; Schafers et al., 2025; Spackman et al., 2024a, b; Suarez et al., 2025). As such, research was conducted by several investigative teams to validate that pasteurization time and temperature parameters used by the dairy industry were adequate to inactivate HPAIV in raw milk. For instance, Spackman et al., (2024a) validated that continuous-flow, high-temperature, short-time (HTST) pasteurization (72 °C for 15 s) of inoculated whole raw milk (ca. 4.5% fat) inactivated ≥ 5.8 log_10_ 50% egg infectious doses (EID_50_) per mL of clade 2.3.4.4b HPAIV isolate Tk/IN/22. As another example, Alkie et al., (2025) demonstrated that both HTST pasteurization (72 °C for 15 s) and low-temperature, long time (LTLT) pasteurization (63 °C for 30 min) delivered a ≥ 6.3-log_10_ EID_50_ per mL reduction of clade 2.3.4.4b HPAIV isolate A/Turkey Vulture/Ontario/FAV473-3/2022 within inoculated whole raw milk. Although pasteurization ensures the safety of milk, the expanding interest among consumers for raw milk and associated (unpasteurized) dairy products, coupled with the recovery of HPAIV and its nucleic acid from raw milk from infected dairy cattle, continues to raise cause for concern from a public health perspective among regulators, dairymen, and consumers alike.

In addition to its recovery from raw milk obtained directly from infected dairy cattle, Spackman et al., (2024a) detected HPAIV in raw milk samples collected from bulk tanks on farms in four states harboring infected and non-infected animals over a 2-week period in April of 2024, with 158 of 275 samples testing positive by RT-qPCR and 24.8% (39 of 158) of these raw milk samples harboring live, infectious HPAIV at levels ranging from 1.3 to 6.3 log_10_ EID_50_ per mL. This same laboratory (Spackman et al., 2024b) also collected 297 samples of Grade A pasteurized milk products (e.g., whole milk, yogurt, cottage cheese, whipping cream, sour cream, and cream) in 2024 from retail stores across 17 U.S. states comprising products from 132 processors in 38 states and reported that clade 2.3.4.4b HPAIV viral RNA was detected via RT-qPCR in 20.2% (60 of 297) of these samples, although infectious virus was not detected in any samples tested. In a complementary study conducted by Suarez et al., (2025), infectious HPAIV was not recovered from 167 retail dairy products that represented processors from 27 U.S. states, including aged raw milk cheese and pasteurized milk, or from products made from pasteurized milk such as butter, ice cream, and various cheese (e.g., mozzarella, Cheddar, cream, and processed cheese), but 17.4% (29 of 167) of these samples tested positive for viral RNA. As a final example, Tarbuck and colleagues (2024) detected HPAIV viral nucleic acid in 36.3% (61 of 168) of retail pasteurized milk samples purchased across 18 U.S. states between April and May of 2024, but none of these samples tested positive for viable, infectious virus. Collectively, these findings substantiate the presence of viral RNA and infectious AIV in raw milk and the presence of viral RNA in retail milk or dairy products. However, experimental studies support effective inactivation of virus by pasteurization as corroborated by retail dairy product testing and as evidenced by detection of HPAIV RNA but not infectious virus in retail milk or dairy products. Apart from pasteurization, these data also provide the impetus for research to validate the efficacy of other processing technologies (e.g., fermentation, drying/aging, and high-pressure processing) to mitigate the potential risk of AIV in dairy products such as yogurt and kefir that on occasion may be prepared from raw milk.

While pasteurization will eliminate infectious HPAIV from raw milk (Alkie et al., 2025; Nooruzzaman et al., 2025; Spackman et al., 2024a), relatively little has been published on interventions to inactivate this virus if present in dairy products made from unpasteurized milk. In this regard, fermentation has been practiced for centuries to prepare/preserve dairy products by lowering levels of pathogenic and spoilage microorganisms during processing and preventing their amplification during subsequent storage (Ray & Joshi, 2014; Sionek et al., 2023). In general, the stabilization/shelf-life of yogurt and kefir as established by fermentation of (raw) milk is achieved via the lowering of pH to ca. pH 4.2 to pH 4.6 (Ahmed et al., 2013; Arslan, 2014; Tamime, 2002). In a companion study conducted to mimic the cheese-making process and assess the impact of pH and time on AIV strain rGyrfal/WA/41088-b/14xPR8/2014 (H5N1) in raw milk (Harrell et al., 2026b), we observed that acidification of raw milk with organic acids to pH 4.0 or pH 5.0 and incubation at room temperature delivered appreciable reductions of up to 5.7 log_10_ EID_50_ per mL of this LPAIV strain in 1–24 h, respectively, whereas its levels in otherwise similar samples adjusted to pH 6.0 remained relatively unchanged after 24 h at room temperature. Irrespective of the organic acid used to adjust pH (i.e., acetic, citric, lactic, or propionic), and as anticipated, the observed reductions in viral titers were associated specifically with pH and time. With respect to its efficacy to control HPAIV in dairy products, Lang and colleagues (2024) reported that fermentation of whole milk at 42 °C for 8 h to pH 4.4 using a commercial yogurt starter culture delivered a ca. 4.0-log_10_ plaque forming units (PFU) per mL reduction of HPAIV clade 2.2.2 H5N1-06 and HPAIV clade 2.3.4.4b H5N1-22 strains. For a home-made-type yogurt, Harrell and colleagues (2026a) also reported that use of a commercial starter culture to ferment raw milk inoculated with low pathogenic avian influenza virus (LPAIV) for ca. 7 h at 42 °C to < pH 4.4 resulted in a > 4.1-log_10_ EID_50_ per mL reduction of H5N1 strain rgGYR/14. Like yogurt, kefir is a fermented milk with probiotic properties, but it is noticeably “foamy”, not as viscous, and according to some consumers is slightly less acidic in taste than yogurt. Moreover, kefir is made using a symbiotic starter culture comprised of a diverse blend of yeast and mesophilic lactic acid bacteria that requires between 24 and 36 h of fermentation at ambient temperature (i.e., 20° to 25 °C) to achieve an endpoint pH ranging from pH 4.0 to pH 4.6 (Ahmed et al., 2013; Alves et al., 2021; Leite et al., 2013; Maughan et al., 2025; Prado et al., 2015; Rattray & O’Connell, 2011). The lower temperature and longer time of the fermentation process for kefir compared with yogurt may not be as detrimental to HPAIV H5N1 viability if this virus were present in raw milk. Given the recovery of genotype B3.13 (Caserta et al., 2024) and D1.1 (USDA APHIS, 2025) strains from dairy cattle in 2024 and 2025, respectively, we monitored viability of two strains [H5N1 (clade 2.3.4.4.) and H5N9 (North American lineage LPAIV)] of LPAIV during fermentation of kefir in response to the time, temperature, and pH conditions for fermentation of raw milk kefir to address the potential for strain-to-strain variation and to assess the attendant impact, if any, on public health.

Materials and Methods

Inoculation of Raw Milk

Three batches (ca. 3.8 L each) of fresh “Grass Fed A2 Raw Milk” (average = 2.95% fat; range = 2.6 to 3.3% fat; labeled “Keep Refrigerated”) were purchased on three separate visits to a permitted local farm and held at 4 °C for up to 8 h prior to use. Approximately 500 mL of each batch of raw milk were transferred into wide-mouth plastic jars (Uline, Pleasant Prairie, WI) that were appropriately sealed and stored at − 20 °C for subsequent proximate chemical analyses as detailed below. For each of three trials, a separate 1-L portion of raw milk was inoculated with a commercial kefir starter culture [1 g per L milk to deliver ca. 7.5 log_10_ CFU per g of active starter culture [(Coolinario Kefir Starter; Imported Goods for Coolinario, Calgary, AB, Canada); 20 billion CFU per g of freeze-dried Lactobacillus bulgaricus, Streptococcus thermophilus, Lactobacillus acidophilus, Lactococcus lactis, Leuconostoc mesenteroides, Lactobacillus kumiss, and Saccharomyces cerevisiae (starter culture also contains skim milk powder and maltodextrin used in the freeze-drying process)]. After swirling by hand for 30 to 60 s to uniformly distribute the starter culture, 125-mL portions of the raw milk containing the starter culture were transferred into sterile screw cap, media glass bottles (CLS1395; Corning Inc., Corning, NY) and subsequently inoculated with either of two isolates of low pathogenic avian influenza virus [LPAIV; A/rgGyrfalconHAxPR8/2014 H5N1, clade 2.3.4.4.; aka strain rgGYR/14 (Spackman & Killian, 2020; Stephens & Spackman, 2017) or A/turkey/Wisconsin/1968 H5N9, a North American lineage LPAIV; aka strain A/turkey/WI (Smithies et al., 1969)] as follows: 1 mL of LPAIV strain rgGYR/14 or 1 mL of LPAIV strain A/turkey/WI were added to 125 mL of milk to achieve an average initial level of ca. 5.0 log_10_ 50% egg infectious doses (EID_50_) per mL of milk (Fig. 1).

Fig. 1. Process used to monitor viability of two strains of low pathogenic avian influenza virus (H5N1 and H5N9) during fermentation of raw milk kefir^A^. Control treatments included: (i) raw milk (125 mL) without the kefir starter culture, but separately inoculated with strain rgGYR/14 or strain A/turkey/WI^B^, (ii) raw milk (125 mL) inoculated with the kefir starter culture, but not inoculated with either of the LPAIV strains^C^, and (iii) raw milk (125 mL) without the kefir starter culture and without either of the LPAIV strains^D^

Control treatments included: (i) raw milk (125 mL) without the kefir starter culture, but separately inoculated with strain rgGYR/14 or strain A/turkey/WI, (ii) raw milk (125 mL) inoculated with the kefir starter culture, but not inoculated with either of the LPAIV strains, and (iii) raw milk (125 mL) without the kefir starter culture and without either of the LPAIV strains. Both control and experimental treatments were fermented/incubated in a circulating water bath (Model 2874; Thermo Fisher Scientific, Marietta, OH) maintained at 25 °C. Both LPAIV strains were prepared at the USDA Agricultural Research Service (ARS) U.S. National Poultry Research Center (USNPRC; Athens, GA) and shipped to the USDA ARS Eastern Regional Research Center (ERRC; Wyndmoor, PA) with approval from the USNPRC Institutional Biosafety Committee. Protocols for using LPAIV at ERRC to inoculate and ferment raw milk to produce kefir were approved by the USDA ARS Northeast Area Institutional Biosafety Committee.

Neutralization of Milk and Kefir

All milk samples were held at 25 °C until an endpoint pH of ca. pH 4.4 was achieved in the samples that received the starter culture (ca. 36 h). The pH of a 5-mL aliquot of the milk or kefir at the end of incubation or at the end of fermentation, respectively, was determined using a pH/temperature electrode (Model ST310; Ohaus, Parsippany, NJ) and pH meter (Model Aquasearcher a-AB33PH-F; Ohaus). All samples (ca. 30 mL each) were neutralized to ca. pH 7.0 using 7 to 135 mL of a two-fold (2X) strength sterile solution of Dey Engley (DE) neutralizing broth (Sigma-Aldrich, St. Louis, MO). Next, 0.4 mL of a penicillin-streptomycin solution (final antibiotic concentrations in neutralized milk or kefir samples were 10,000 UI per mL of penicillin and 10,000 µL per mL of streptomycin; Sigma-Aldrich) and a 3.6-mL portion of each neutralized raw milk or kefir sample to be assayed for infectious LPAIV were aseptically transferred into cryovial tubes (SPL Lifesciences, Gyeonggi-do, Korea), held at ambient temperature for ca. 60 min to reduce the indigenous bacterial flora, and then transported overnight on ice to USNPRC by a commercial courier.

LPAIV Detection Via Real-Time Reverse Transcriptase Quantitative PCR (RT-qPCR)

Extraction of total RNA from raw milk and kefir samples and the RT-qPCR assay were performed as previously described (Harrell et al., 2026a). In short, total RNA was extracted from kefir with the MagMax viral RNA isolation kit (Applied Biosystems, Waltham, MA) via KingFisher (Thermo Fisher Scientific, Waltham, MA) robotic extraction. Viral lysis solution (106.2 µL = 50 µL viral/lysis binding solution, 1 µL carrier RNA, 0.3 µL bovine serum albumin (BSA), 5 µL sodium sulfite (0.8%), and 50 µL 100% isopropanol), bead solution (20 µL = 6 µL bead suspension solution, 4 µL nuclease-free H_2_O, 4 µL binding beads, and 6 µL 100% ethanol), and a kefir or milk sample (50 µL) were used as direct input. Samples were washed in 150 µL NaCl (Sigma) with EDTA 2X (Sigma), as well as kit wash solution 1 and 2 prior to being eluted in 50 µL of elution buffer. Total RNA was used as input for RT-qPCR using AgPath Master mix (Applied Biosystems) according to the manufacturer’s protocol. For each reaction, 2 µL H_2_O, 12.5 µL Ag Path buffer, 1 µL AgPath enzyme, 0.5 µL probe (5′-FAM-TCA GGC CCC CTC AAA GCC GA-BHQ-1-3′), 0.5 µL forward primer (5′-AgA TGA GTC TTC TAA CCG AGG TCg-3′), 0.5 µL reverse primer (5′-CCT GCA AAA ACA TCT TCA AGT CTC TGC GC -3′), and 8.0 µL of sample RNA were used. All tested samples were run in duplicate for the entire study. Quantification of nucleic acid was estimated by RT-qPCR cycle threshold (Ct) values.

Inoculation of Embryonated Chicken Eggs (ECE)

ECE assays were conducted essentially as described (Harrell et al., 2026a). Briefly, 10-day-old embryonated specific pathogen free (SPF) eggs were inoculated with 100-µL of raw milk or kefir injected with a 25-gauge 5/8 needle (Becton, Dickinson and Company, Franklin Lakes, NJ) into each egg through the chorioallantoic membrane and into the allantoic fluid. Each egg was subsequently sealed with glue (Elmer’s Products, Westerville, OH) and then incubated at 37 °C in a circulated air egg incubator (Model 1550, G.Q.F. Manufacturing CO., Savannah, GA). Eggs were visually inspected daily for vascular integrity and embryonic movement as indicators of embryo viability. When deceased embryos were detected, such eggs were removed from the incubator and stored at 4 °C until the end of the experiment (96 h).

Hemagglutination Assays (HA)

HA were performed as previously described (Harrell et al., 2026a; Killian 2014). In short, 50 µL of phosphate buffer solution (PBS; Thermo Fisher) were added to a 96-well clear round bottom plate (Corning) followed by the addition of 50 µL of allantoic fluid. The 1:1 dilution of allantoic fluid was subsequently diluted 1:1 again in PBS to result in a 1:4 dilution of allantoic fluid. Lastly, 50 µL of fresh chicken red blood cells (CRBC) were added to all wells. After gentle vortexing and incubation (30 to 45 min, room temperature), plates were assessed for the formation of a CRBC button (negative) or shield (positive) result. Virus titers were expressed as log_10_ EID₅₀ per mL as calculated based on the number of allantoic fluid samples testing positive via HA assays using the Reed-Muench method (limit of detection of 0.166 log_10_ EID₅₀; Spackman et al., 2019).

Microbiological Analyses

Total aerobic plate counts (TPC) and total lactic acid bacteria (LAB) levels were determined by separately analyzing a single 50-mL portion of each batch of raw milk (N = 3, n = 1) as described (Porto-Fett et al., 2008). Results were expressed as log_10_ CFU per mL. Raw milk was also tested for the presence/absence (limit of detection of 0.04 CFU per mL) of Listeria monocytogenes (AOAC 2004.02) and Salmonella spp. (AOAC 2004.03) by an independent commercial testing facility using methods approved by the Association of Official Analytical Chemists (AOAC, 2023).

Chemical Analyses

Proximate chemical analyses (Table 1) were conducted on a single ca. 500-mL portion from each of the three batches of either raw milk or kefir (N = 3 trials/batches; n = 1 sample per each batch) by an independent commercial testing facility using approved/standard methods (AOAC, 2023). Milk samples (N = 2, n = 1) were also tested by a commercial laboratory for the following 41 antimicrobials (or residues therefrom) according to Jiménez et al., (2011): amoxicillin, ampicillin, cloxacillin, pen V, dicloxacillin, nafcillin, oxacillin, pen-G, chlortetracycline, doxycycline, oxytetracycline, tetracycline, enrofloxacin, ciprofloxacin, danofloxacin, sarafloxacin, difloxacin, flumequine, oxolonic acid, marbofloxacin, norfloxacin, sulfapyridine, sulfadiazine, sulfachlorpyridazine, sulfadoxine, sulfadimethoxine, sulfamethazine, sulfamethizole, sulfisoxazole, sulfamethoxypyridazine, sulfaquinoxaline, sulfathiazole, sulfamethoxazole, sulfamerazine, sulfamonomethoxine, erythromycin, josamycin, spiramycin, lincomycin, tilimicosin, and tylosin. Note, the detection limit for these antimicrobials was < 0.01 ppm (w/w).

Table 1. Proximate composition of raw milk and raw milk kefir (N = 3, n = 1)AnalyteRaw milk (% w/w)Raw milk kefir (% w/w)Ash0.73 ± 0.02^1,a^0.56 ± 0.04^b^Carbohydrates (calculation)6.41 ± 1.67^a^4.15 ± 1.04^a^Fat2.95 ± 0.36^a^3.99 ± 0.44^b^Moisture—forced air oven86.52 ± 1.79^a^88.40 ± 0.43^a^Protein3.38 ± 0.22^a^2.89 ± 0.32^a^GalactoseND^2^0.58 ± 0.01Fructose0.25 ± 0.0^a^0.10 ± 0.0^a^Glucose0.25 ± 0.0^a^0.11 ± 0.02^b^Sucrose0.25 ± 0.0^a^0.10 ± 0.0^a^Maltose0.25 ± 0.0^a^0.10 ± 0.0^a^Lactose3.54 ± 0.07^a^2.44 ± 0.03^b^Acidity, as lactic acid0.09 ± 0.01^a^0.96 ± 0.01^b^Acidity, as acetic acidND0.62 ± 0.01pH6.45 ± 0.13^a^4.42 ± 0.03^b^^1^For a given analyte, means within a row that have no lowercase letter in common denote statistical differences (p < 0.05) between raw milk and kefir^2^ND not determined

Statistical Analyses

A two-way ANOVA followed by Fisher’s LSD post hoc test was used to compare Ct values from raw milk samples with and without the kefir starter culture, both prior to and after fermentation/incubation. Variation in the composition of different raw milk batches and among kefir batches was evaluated using one-way ANOVA. Bonferroni-adjusted significance thresholds were applied, and results were considered significant at a Bonferroni-corrected p value < 0.05. EID₅₀ values were log₁₀-transformed to normalize their distribution, and group comparisons were conducted using paired two-tailed t-tests. Effect sizes were quantified using Cohen’s d, with conventional thresholds of 0.2, 0.5, and 0.8 representing small, medium, and large effects. Differences between raw milk and kefir proximate composition for individual analytes (i.e., fat, protein, lactose, etc.) were analyzed using paired, two-tailed Student’s t-tests. All statistical analyses were carried out in GraphPad Prism v10.3.1 (La Jolla, CA).

Results and Discussion

In early 2024 a potentially serious public health concern emerged for the U.S. due to dairy cattle testing positive for H5N1 AIV and subsequent high viral titers (i.e., ca. 9.0 log_10_ EID_50_ per mL) being detected in raw milk from infected cows (Caserta et al., 2024; Hu et al., 2024; Palme et al., 2024). Accordingly, the U.S. dairy industry has experienced an appreciable economic burden and an overall adverse impact attributed directly to the ongoing H5N1 outbreak in cattle (Peña-Mosca et al., 2025). Detection of AIV on dairy farms and in raw milk from infected cows has also been linked to cross-species transmission of the virus to other animals on the farm such as cats and raccoons (Butt et al., 2024; Caserta et al., 2024). In fact, several domestic/indoor cats from H5N1-positive dairy farms became ill or died from AIV infection, most likely due to consumption of contaminated unpasteurized milk (Banyard et al., 2025; Burrough et al., 2024; Caserta et al., 2024; Mainenti et al., 2025; Palme et al., 2024). Illnesses and deaths of indoor cats caused by H5N1 in the U.S. have also been associated with consumption of raw pet food made with contaminated meat from AIV-infected poultry (FDA, 2025a). However, it is more alarming that, to date, transmission of “bird flu” to humans has caused 71 illnesses, with most cases associated to handling infected animals on the farm (CDC, 2025a). There have also been two human fatalities: one caused from exposure to a non-commercial backyard chicken flock, the environment, and/or wild birds and the other caused by an unknown source (CDC, 2025a). These findings provide justification for additional research to validate interventions to mitigate the potential risk of zoonotic infection of humans with AIV from contaminated raw milk and dairy products made from raw milk.

Consumption of raw milk and raw milk products, as practiced by many consumers, may pose a greater risk for zoonosis because of the presence and persistence of various pathogens in raw milk (Currier, 2023; Schafers et al., 2025), including AIV (Alkie et al., 2025; Schafers et al., 2025; Spackman et al., 2024a, b; Suarez et al., 2025). In December of 2024, raw milk and raw cream products were recalled from a dairy farm in California due to potential contamination with AIV H5N1 (CDPH, 2024). Note, H5N1 virus was also detected in milk and dairy products at retail, as well as in the bulk storage tank at the processing plant and in products at the bottling facility, and all of these samples/products were associated with the incriminated farm (CDPH, 2024). As such, in addition to the voluntary recall of raw milk and raw cream products, the California Department of Public Health (CDPH) warned consumers to avoid other raw products from the implicated farm, including kefir made from raw milk intended for human or animal consumption (CDPH, 2024). Considering that HPAIV may persist for up to 56 days in raw milk stored at 4 °C (Nooruzzaman et al., 2025), that ca. 4.4% of the adult population in the U.S. drink raw milk regularly (Lando et al., 2023), and that AIV has been detected in raw milk samples and can be transmitted to humans through occupational exposure, there is justifiable concern for human illness arising from consuming raw milk or dairy products produced from raw milk due to its potential for contamination with AIV.

Composition of Raw Milk and Kefir

Although there were subtle differences in formulation and fermentation parameters unique to the present study, our results for chemical analyses of raw milk (Table 1) compare favorably with proximate chemical analyses of milk from related studies in which levels of moisture, fat, protein, carbohydrate, and ash were reported to range from ca. 86.3 to 90%, ca. 0.03% to 3.5%, ca. 2.9 to 6.4%, ca. 3.8 to 6.0%, and ca. 0.7 to 1.2%, respectively (Arslan, 2014; Liutkevicius & Sarkinas, 2004; Magalhaes et al., 2011; Rosa et al., 2017; Wszolek et al., 2001). Overall, our results for chemical analyses of kefir (Table 1) also compare favorably with proximate chemical analyses of kefir from related studies published by other investigators (Abou-Ayana et al., 2025; Ceylan & Öncül, 2025; Farag et al., 2020; Garofalo et al., 2020; Hew et al., 2023; Satir & Guzel-Seydin, 2016; Yerlikaya et al., 2025) wherein levels of moisture, fat, protein, carbohydrate, and ash ranged from ca. 86.3 to 90%, ca. 2.2 to 3.5%, ca. 2.7 to 5.8%, ca. 4.8 to 6.2%, and ca. 0.6 to 0.7%, respectively. In the present study, except for levels of carbohydrate, moisture, and protein (p ≥ 0.05), significant differences in the levels of fat (p = 0.0375), ash (p = 0.0291), and pH (p = 0.0018) were observed between raw milk and kefir. Ash values may change “significantly” during fermentation due to changes of the other components like moisture or protein, but in practical terms, such differences were not that significant in the final product. Regarding fat, potential differences could be due to the milk not being homogenized prior to fermentation or that the coagulation of milk by the action of fermentation produced a non-homogenous kefir sample. It should also be noted that lactic acid levels increased substantially (p = 0.0073) in kefir (0.96%) from its initial levels (0.09%) in raw milk, reflecting the conversion of lactose into lactic acid [likely from the naturally occurring or introduced lactic acid bacteria (LAB)] during fermentation and, in turn, would contributed significantly to the characteristic acidity (and presumably flavor) of the final product. Acetic acid was not assayed in the raw milk since it is not naturally found in milk; however, its presence in kefir at 0.62% is attributable to the metabolic activity of acetic acid bacteria, including Acetobacter spp. and certain heterofermentative LAB and yeasts in the kefir “grains” or starter culture, that collectively contribute to the wholesomeness and complex flavor profile of kefir (Leite et al., 2013; Prado et al., 2015; Öner et al., 2010). As expected, there were also significantly lower levels of lactose (p = 0.0147) and glucose (p = 0.0051) in kefir compared to raw milk. In addition, levels of sucrose, fructose, and maltose were also lower in kefir than in raw milk. These results are most likely due to conversion of these sugars into acetic and lactic acids by LAB and lactose-fermenting and non-fermenting yeast that lower pH (Hikmetoglu et al., 2020; Leite at al., 2013, Saleem et al., 2023). Variations in proximate composition of milk and kefir among studies may impact fermentation and virus titers and recovery and, in turn, form the basis for additional studies. Such compositional differences may be attributed to several factors, at least to some extent, including the milk species (i.e., bovine, ovine, or caprine) and type (i.e., whole, skimmed, or partially-skimmed pasteurized or raw milk), type and concentration of starter culture (e.g., kefir “grains” versus commercial starter cultures), composition and levels of the indigenous microflora, and time and temperature of fermentation (Kök-Taş et al., 2013; La Torre et al., 2024; Öner et al., 2010; Wojtowski et al., 2003).

Regarding the microbial profile of the raw milk used herein, the average initial levels of TPC and LAB in the raw milk were ca. 3.4 log_10_ CFU per mL (ranging from ca. 3.0 to 4.0 log_10_ CFU per mL) and ca. 2.8 log_10_ CFU per mL (ranging from ca. 2.2 to 3.2 log_10_ CFU per mL), respectively. Although standard levels for TPC are not stipulated by Federal regulatory agencies for raw milk destined for human consumption in the U.S., levels of TPC observed herein agree with levels recommended (≤ 3.7 log_10_ CFU/mL) by the Raw Milk Institute, a non-profit organization for safe production of raw milk in North America (RAWMI, 2020). Our results for TPC levels are also in agreement with levels reported in raw cow’s milk (ranging from ca. 3.6 to 4.9 log_10_ CFU/mL) in EU countries (EFSA, 2015). Likewise, levels of LAB reported herein compare favorably with published studies wherein the levels of LAB in raw milk ranged from ca. 1.0 to 4.0 log_10_ CFU per mL (Gagnon et al., 2020; Harrell et al., 2026a; Quigley et al., 2013). Regarding pathogenic bacteria, no cells of L. monocytogenes or Salmonella were recovered by enrichment (detection limit of ≤ 0.04 CFU per mL; data not shown). In addition, no H5N1 viral RNA was detected by RT-qPCR in the original (non-inoculated) raw milk samples either before or after incubation. Lastly, none of the 41 target antimicrobials were detected (≤ 0.01 ppm w/w) in the three batches of raw milk used herein. The presence of antimicrobials in raw milk could adversely affect the starter culture and, in turn, negatively impact fermentation, resulting in an insufficient or delayed decrease in pH and a final product with poor quality and/or questionable shelf life and safety.

Kefir-Making Process

The initial pH of raw milk decreased significantly (H5N1 p = 0.0006; H5N9 p < 0.0001) from ca. pH 6.55 ± 0.05 to ca. pH 4.44 ± 0.05 in kefir inoculated with either LPAIV strain after ca. 36 h of fermentation at 25 °C using a commercial starter culture (Table 2). This indicates that the fermentation process successfully occurred in these samples. The initial pH of the raw milk that was not inoculated with the kefir starter culture, but that was inoculated with either of the LPAIV strains, also significantly decreased (H5N1 p = 0.0043; H5N9 p < 0.0001) from ca. pH 6.55 ± 0.05 to on average pH 5.28 ± 0.34 after 36 h at 25 °C (Table 2). Likewise, the initial pH of the raw milk that was not inoculated with either the starter culture or the LPAIV strains decreased significantly (p = 0.0002) from pH 6.55 ± 0.05 to pH 4.99 ± 0.28 after ca. 36 h at 25 °C. The observed decrease in pH of raw milk in the absence of a kefir starter culture is likely attributed to acid production by the proliferation of naturally occurring LAB and other spoilage microorganisms. Disparities in pH values of milk fermented by the indigenous flora rather than by a commercial starter culture may be attributed, at least in part, to the extended time and ambient temperature of incubation. Note, the fermentation parameters for yogurt practiced by Lang et al., (2024) and Harrell et al., (2026a) were measurably different [higher temperature (42 °C) and shorter time (ca. 8 h)] than the fermentation parameters practiced herein for kefir [lower temperature (25 °C) and longer time (ca. 36 h)]. Still, as shown in Table 2, after ca. 36 h at 25 °C significant differences (H5N1 p = 0.0043; H5N9 p < 0.0001) were observed between the pH of raw milk at the end of incubation versus the pH of kefir at the end of fermentation; however, a significant reduction in pH from the initial pH was observed for both samples.

Table 2pH of raw milk and raw milk with or without added starter culture and/or strains rgGYR/14 and A/turkey/WI following incubation at 25 °C for ca. 36 h (N = 3, n = 1)SampleIncubation time (h)pHRaw milk (negative control)^1^06.55 ± 0.05^5,6,a^Raw milk (negative control)^1^364.99 ± 0.28^b^Raw milk, not inoculated with starter culture, but inoculated with strain H5N1^2^06.57 ± 0.04^aA^Raw milk, not inoculated with starter culture, but inoculated with strain H5N1^2^365.31 ± 0.50^bA^Raw milk, not inoculated with starter culture, but inoculated with strain H5N9^2^06.57 ± 0.04^aA^Raw milk, not inoculated with starter culture, but inoculated with strain H5N9^2^365.24 ± 0.09^bA^Raw milk, inoculated with starter culture and with strain H5N1^3^06.57 ± 0.04^aA^Kefir inoculated with strain H5N1^4^364.40 ± 0.02^bB^Raw milk, inoculated with starter culture and with strain H5N9^3^06.57 ± 0.04^aA^Kefir inoculated with strain H5N9^4^364.47 ± 0.06^bB^^1^Negative control: not inoculated with LPAIV, and not inoculated with starter culture. Prior and after incubation at 25 °C for ca. 36 h^2^Raw milk not inoculated with starter culture, but inoculated with LPAIV. Prior and after incubation at 25 °C for ca. 36 h^3^Raw milk inoculated with starter culture (kefir) and inoculated with LPAIV. Prior to fermentation at 25 °C for ca. 36 h^4^Kefir inoculated with LPAIV. After incubation at 25 °C for ca. 36 h^5^For a given milk or kefir sample, means with different lowercase letters denote statistical differences (p < 0.05) before (at time 0 h) versus after incubation/fermentation for ca. 36 h at 25 °C^6^For a given incubation time and LPAIV strain, means with different uppercase case letters denote statistical differences (p < 0.05) between raw milk that was not inoculated with the starter culture versus raw milk that was inoculated with the starter culture (kefir)

Inactivation of LPAIV in Raw Milk Kefir

Overall, a significant decrease in virus titers for both LPAIV strains was observed when comparing samples before and after the ca. 36 h incubation at 25 °C, regardless of whether the raw milk received the starter culture or not. Specifically, for milk inoculated with H5N1 strain rgGYR/14 (initial titer of ca. 5.5 log_10_ EID_50_), as quantified via ECE assays, virus levels were reduced by 4.1 log_10_ EID_50_ per mL during incubation of milk that did not receive starter culture (1.4 ± 1.2 log_10_ EID_50_ remaining after incubation; p = 0.0027) and for kefir which was inoculated with the starter culture (1.4 ± 0.30 log_10_ EID_50_ remaining after fermentation; p = 0.0025); however, infectious virus was recovered from both of these final products (Table 3). Similarly, for milk that was inoculated with H5N9 strain A/turkey/W14 (initial titer of ca. 4.2 log_10_ EID_50_), although virus levels were reduced by 3.5 and 3.1 log_10_ EID_50_ per mL in both incubated milk (0.7 ± 0.32 log_10_ EID_50_ remaining after incubation; p = 0.0005) and in kefir (1.2 ± 0.65 log_10_ EID_50_ remaining after fermentation; p = 0.0006), respectively (Table 3), low levels of infectious virus were also recovered from the associated final products (Table 3).

Table 3. Viral titer via embryonated chicken eggs (ECE) assay and estimated titer via RT-qPCR for raw milk and raw milk kefir inoculated with strains rgGYR/14 (H5N1) or A/turkey/WI (H5N9) before and after incubation/fermentation at 25 °C for ca. 36 h (N = 3, n = 1)TreatmentIncubation time (h)Titer^1^ (log_10_ EID_50_/mL)PCR based virus titer estimates^2^ (log_10_ EID_50_/mLRaw milk—inoculated with H5N1 and starter culture05.5 ± 0.2^3,aA^5.3 ± 0.6Kefir—inoculated with H5N1 and starter culture361.4 ± 0.3^bA^4.1 ± 0.5Raw milk—inoculated with H5N1 only05.5 ± 0.3^aA^5.4 ± 0.5Raw milk—inoculated with H5N1 only361.4 ± 1.2^bA^4.2 ± 0.4Raw milk—inoculated with H5N9 and starter culture04.3 ± 0.2^aA^5.2 ± 0.4Kefir—inoculated with H5N9 and starter culture361.2 ± 0.6^bA^3.3 ± 1.9Raw milk—inoculated with H5N9 only04.2 ± 0.4^aA^5.0 ± 0.4Raw milk—inoculated with H5N9 only360.7 ± 0.3^bA^4.2 ± 0.3^1^Titer: infectious virus concentration determined by endpoint titration in 10-day-old specific-pathogen-free embryonated chicken eggs (ECE) and expressed as the mean log_10_ 50% egg infectious dose per milliliter (EID_50_/mL) ± standard deviation^2^PCR-based virus titer estimates: quantitative estimate of viral RNA copy number converted to log₁₀ EID₅₀/mL equivalents using a standard curve derived from serial dilutions of the respective virus stock with known infectious titer^3^For a given treatment, within a row, means that have no lowercase letter in common denote statistical differences (p < 0.05) before (at time 0 h) versus after incubation/fermentation for ca. 36 h at 25 °C^4^For a given incubation time and LPAIV strain, means that have no uppercase case letter in common denote statistical differences (p < 0.0001) between raw milk that was not inoculated with starter culture versus raw milk that was inoculated with starter culture (kefir)

There were no significant differences in the levels of strain rgGYR/14 (H5N1) or strain A/turkey/WI (H5N9) recovered from kefir after fermentation compared to levels recovered from milk samples that were incubated at 25 °C for ca. 36 h (H5N1 p = 0.9558; H5N9 p = 0.2345) despite a significant difference in the final pH of these treatments (H5N1 p = 0.0041; H5N9 p < 0.0001). More specifically, the pH of raw milk incubated at 25 °C for ca. 36 h was lowered from an average initial level of pH 6.55 to an average final level of pH 5.28 due solely to the metabolic activity of the indigenous flora, whereas the average pH was lowered from pH 6.55 to pH 4.44 in kefir due specifically to inclusion of the starter culture. It was not possible to directly assess the potential for strain-to-strain variation herein due to measurable differences in the starting levels of these two LPAIV strains at 0 h (ca. 1.2 log_10_ EID_50_ higher titer for strain rgGYR/14 at 0 h compared to strain A/turkey/WI). Future studies may be conducted to determine if virus stability, its physiological attributes, and/or sensitivity to intrinsic factors of raw milk may manifest as strain-to-strain variations in viability of AIV if inoculated in/on foods or when challenged under food relevant conditions.

As expected, fermentation of raw milk decreased levels of both strains of LPAIV by ≥ 3.1 log_10_ EID_50_ per mL. Thus, fermentation of raw milk at 25 °C for ca. 36 h would effectively reduce elevated levels of infectious AIV if initially present in raw milk. Our findings are in general agreement with other studies, including by Crossley and colleagues (2025), who reported that a HPAIV H5N1 strain and a LPAIV H6N2 strain were inactivated after ca. 6 h at an “environmental temperature” in raw whole cow milk acidified with citric acid to ca. pH 4.2. In a related study, we reported that raw milk yogurt fermented via a commercial starter culture to pH 4.4 at 42 °C for 7 to 8 h reduced levels of (inoculated) LPAIV H5N1 strain rgGYR/14 by ≥ 4.1 log_10_ EID_50_ per mL (Harrell et al., 2026a). The pronounced effect of high acid conditions on AIV infectivity in dairy products was also apparent in our companion study (Harrell et al., 2026b) wherein appreciable reductions of LPAIV H5N1 strain rGyrfal/WA/41,088-b/14xPR8/2014 in raw milk incubated at room temperature were achieved in ca. 1 h at pH 4.0 compared to ca. 24 h at pH 5.0, with little to no inactivation of the virus observed at pH 6.0. Lenz-Ajuh et al., (2025) also evaluated the effect of pH and temperature on inactivation of HPAIV H5N1 during fermentation of raw milk yogurt at 42 °C for ca. 5 h: the pH of the resulting yogurt ranged from pH 4.2 to pH 4.6 and no infectious virus were detected [i.e., observed reduction of ca. 7.0 log_10_ focus forming units (FFU) per mL from initial levels]. However, these authors reported that when raw milk without starter culture was incubated under the same time and temperature conditions, the pH of the raw milk remained relatively unchanged (pH 6.5), yet reductions of ca. 2.0-log_10_ FFU per mL in levels of HPAIV H5N1 were still achieved. In a similar study by Lang et al., (2024), HPAIV H5N1 was not detected in yogurt prepared with commercial, organic fresh whole milk (3.8% fat; pH 6.5) that was inoculated with 4.0 to 4.5 log_10_ plaque-forming units (PFU) per mL and then fermented (using a commercial yogurt as a starter culture) at 42 °C for 8 h to pH 4.4. In the control sample (whole milk without the starter culture being added), ca. 4.0 log_10_ PFU per mL of the virus were recovered (i.e., no appreciable reduction in levels of H5N1): the pH of the milk in the absence of a starter culture remained relatively unchanged (pH 6.5) after incubation at 42 °C for 8 h (Lang et al., 2024). Unlike the previously mentioned studies reporting that live virus was not detected in the final product, herein low levels of infectious LPAIV were found in kefir, as well as in milk maintained at 25 °C for ca. 36 h, albeit at much lower levels compared to levels inoculated into raw milk. Thus, further research is warranted to elaborate if the residual LPAIV in kefir were present at sufficient levels to render it unsafe for human or animal consumption. Further research is also warranted to elaborate the effect of pH and temperature, alone or in combination, on infectivity of LPAIV in kefir.

LPAIV RNA Degradation During Fermentation

In the present study, the estimated virus titer obtained via RT-qPCR validated that strains rgGYR/14 and A/turkey/WI were present in all LPAIV-spiked samples, including the milk samples at the end of incubation and the kefir samples at end of fermentation; however, the amount of viral RNA was significantly lower (incubated milk H5N1 p = 0.0249, H5N9 p = 0.2424; kefir H5N1 p = 0.0264, H5N9 p = 0.0371) than levels inoculated into the raw milk (Table 3). Thus, acidification of raw milk via fermentation from the starter culture (e.g., kefir) or the indigenous microflora in the raw milk resulted in measurable viral RNA degradation. Loss of infectivity of H5N1 in fermented milk products may be associated with a perturbation of the membrane fusion mechanism, which at < pH 5.0 generates an irreversible conformational change in a minor fraction of the hemagglutinin proteins on the surface of AIV (Lenz-Ajuh et al., 2025; Remeta et al., 2002; Sato et al., 1983; Yang et al., 2025). This alteration of the membrane fraction mechanism most likely exposes AIV RNA to environmental degradation (Hamilton et al., 2012). Lenz-Ajuh and colleagues (2025) also reported that viral RNA degradation, as estimated via RT-qPCR analyses, occurred when yogurt was fermented at 42 °C for ca. 5 h to pH 4.3. Similarly, we recently reported that Ct values generated by RT-qPCR analyses were indicative of significant degradation of viral RNA after fermentation of raw milk to yogurt over 7 to 8 h at 42 °C (Harrell et al., 2026a). The results of the present study confirmed that fermentation to pH 4.4 will appreciably lower levels of infectious LPAIV virus in fermented dairy products made from raw milk such as kefir. Although it is significant that there was an equivalent and measurable reduction of both LPAIV strains in kefir (ca. pH 4.4) and milk (ca. pH 5.28) after incubation for ca. 36 h at 25 °C, mention should be made that raw milk fermented to ca. pH 5.28 does not meet the standard of identity for kefir and would likely not be saleable or accepted by consumers.

Conclusions

In the conduct of our programmatic research on fermented dairy products made from raw milk, we observed greater reductions in viral titers for yogurt (≥ 4.1 log_10_ EID_50_ per mL decrease of H5N1 strain rgGYR/14; Harrell et al., 2026a) compared to kefir (ca. 4.1 or 3.1 log_10_ EID_50_ per mL decrease of strains H5N1 or H5N9, respectively). As noted, yogurt is fermented at a significantly higher temperature and for a shorter time (ca. pH 4.4 over 7 to 8 h at 42 °C) compared to kefir (ca. pH 4.4 over ca. 36 h at 25 °C). In addition to the low pH environment, the higher fermentation temperature for yogurt presents an added challenge for infectious AIV to persist from a food safety perspective. The latter may explain, at least in part, why reductions of ≥ 4.1 log_10_ EID_50_ per mL were achieved for both yogurt (ca. pH 4.4) and the corresponding control sample of raw milk (ca. pH 6.6), both of which were incubated at 42 °C for 7 to 8 h. By comparison, in the absence of a starter culture reductions of 3.5 to 4.1 log_10_ EID_50_ per mL were achieved after incubation of the raw milk at 25 °C for 36 h. However, the pH was significantly lower (ca. pH 5.3) in the raw milk control treatment for present study compared to its corresponding control samples for the above-mentioned yogurt study (pH 6.6). Fermentation is a controlled and ground-truthed “intervention” that extends shelf life, ensures wholesomeness, and renders products like kefir “safe” when proper conditions of pH, time, and temperature are achieved by the action of a defined starter culture. Thus, it should be emphasized that although the pH of raw milk after ca. 36 h of “incubation” at 25 °C in the absence of a starter culture was appreciably lower (pH 5.3) than the initial pH of the raw milk (ca. pH 6.6), and although notable reductions of strain A/turkey/WI (3.1 log_10_ EID_50_) and strain rgGYR/14 (4.1 log_10_ EID_50_) were observed, raw milk left at room temperature should probably not be consumed, and it certainly would not be fit for market, because such milk would be deemed spoiled. In contrast, “fermentation” of raw milk via the action of a defined starter culture to pH 4.4 after ca. 36 h at 25 °C would result in kefir, a “cultured milk” product available at retailers that contains ≥ 3.25% milkfat and 8.25% milk solids-not-fat (21 CFR 131.112) and is made with a prescribed microflora of lactic acid bacteria and yeasts (Codex Alimentarius Standard, CXS 243-2003), that has been enjoyed by consumers for centuries. Validation that fermentation greatly reduces the amount of live AIV in raw milk during the production of kefir or yogurt forestalls the requirement for additional process/product validation studies which, in turn, conserves resources while ensuring safety and verifying compliance for either intra-state (raw milk kefir) or inter-state (kefir made from pasteurized milk) commerce. These data may also be modeled and subsequently used to support risk analyses and policy development to mitigate any public health consequences from AIV in raw milk, as well as from dairy products prepared from raw milk if contaminated with the virus. Despite being limited in scope and in the numbers and volume of samples analyzed, we are the first to report that fermentation of raw milk using a defined starter culture to produce kefir inactivates AIV. However, despite at least a 3.0-log_10_ EID_50_ per mL reduction in levels of AIV, given the observed low levels of AIV remaining in the final product after fermentation, and in the absence of definitive studies on infectious dose and host susceptibility, we conclude that raw milk kefir contaminated with any avian influenza virus may be unsafe for human consumption.