Alcohol and Acetaldehyde in African Fermented Milk Mursik—A Possible Etiologic Factor for High Incidence of Esophageal Cancer in Western Kenya

In developed countries, alcohol drinking and smoking are the principal known risk factors for upper alimentary tract cancer. Evidence derived from epidemiologic, genetic, biochemical, and microbiological studies strongly suggests that acetaldehyde (ACH), derived either from ethanol or tobacco, is a common denominator in the pathogenesis of these cancers (1–5). Accordingly, the International Agency for Research on Cancer (World Health Organization) recently concluded that ACH present in alcoholic beverages, as a congener or formed endogenously from ethanol, is a Group 1 carcinogen for humans (6).

Esophageal squamous cell cancer (OSCC) is unusually frequent in southern and eastern Africa, including in Western Kenya, where it is the most common malignancy in both men and women (7, 8). Of special interest is the common occurrence of OSCC in young people (<30 year of age) in this area (9, 10), especially among members of the Kalenjin tribe, despite the low prevalence of classical risk factors such as heavy drinking and tobacco smoking (11).

In general, fermented food products constitute a major portion of the daily diet in Africa (12, 13). The use and production of fermented milk is an old tradition among Kenyans. Every tribe has their own unique methods of producing fermented milk, and the differences between fermented milks are based on the location of tribal communities and the different production processes. In general, whole milk is first boiled and poured in a gourd, and leftover fermented milk acting as a starter culture for the new fermentation process is added. These components are mixed well before the gourd is sealed and left to ferment for 2 to 7 days in room temperature.

In the Kalenjin tribe the fermented milk, mursik, is produced in specific calabash gourds, also known as sotet. Some days before the milk is treated, a small branch of an Ite tree (Senna didymobotray) is debarked and allowed to dry. One end of this stick is first burned in a fire and then rubbed on the inner surface of a cleaned gourd. This is repeated several times until the gourd is fully coated inside with charcoal dust. This process reduces the porosity of the gourd and improves the flavor. Some Kalenjin groups also add small quantities of blood obtained from prickling a vein in the neck region of a healthy bull, and from which fibrin has been removed by gentle stirring. Addition of blood can impact the microbial metabolism, as iron is an important cofactor for a number of essential cellular processes. A thick bluish layer forms on the surface when mursik is ready. It is shaken well before drinking, to ensure that a uniformly thick emulsion is formed. In some Kalenjin households, fermented milk is consumed several times daily.

Fermentation and all of these additions to the process are used to improve the odor, the taste, and the flavor of fresh milk. In an interview with farmers, Mureithi and colleagues noted that farmers were of the opinion that fresh milk smells and tastes like cow urine and had to be improved before it can be consumed (14). Also, lack of refrigeration, and the need to store milk for the dry season (when milk production decreases due to a lack of pasture) required that excess milk be stored for a longer time. For example, the Pokot developed chekha mwaka, a specially treated milk that could be stored for more than a year without getting spoiled (14).

Yeast and lactic acid bacteria (LAB) have been reported to be the dominant microbial species in African fermented milk (15). The most commonly isolated yeasts are Candida and Saccharomyces spp. (16). Lactobacillus, Lactococcus, Leuconostoc, and Enterococcus spp. are the most frequently isolated bacteria (15). Interaction of the microbial flora affects the properties of the fermented milk, such as its consistency and flavor. ACH is one of the most important flavor compounds in milk products, especially yogurts (15).

Fermented milk products may contain up to 3.8% w/v (0.8 M) ethanol and mutagenic (>100 μmol/L) concentrations of ACH (17–20). In addition, oral microbes are able to produce marked amounts of ACH even from very low (0.01–0.02 M) amounts of ethanol present in saliva after alcohol intake (21). ACH production locally in the mouth is enhanced when ethanol concentrations increase (22, 23). Also, ACH present in alcoholic beverages blends readily into the saliva (22, 23). Presently, only beverages containing more than 2.8% ethyl alcohol are classified as Group 1 carcinogenic alcoholic beverages. However, fermented foods may expose the upper digestive tract mucosa to ethanol and ACH levels equivalent to those of alcoholic beverages (1, 5).

Our hypothesis is that regular consumption of fermented mursik milk contributes to the particularly high incidence of esophageal cancer in the Kalenjin tribe through its ethanol and ACH content. As an initial test of this hypothesis, we first determined the levels of ethanol and ACH produced in the fermentation process. Second, we identified the microbial flora in mursik milk responsible for the fermentative process to determine whether it was compatible with production of ACH.

Eight samples of mursik milk starter cultures were collected from different Kalenjin families in the Bomet area, Rift Valley province, Western Kenya by investigators from Tenwek Hospital (Bomet, Kenya). The samples were stored in blue-capped plastic tubes for transportation. The samples were first transported via baggage at room temperature to the Division of Cancer Epidemiology and Genetics, National Cancer Institute, NIH, Bethesda, Maryland, USA and then to the Research Unit on Acetaldehyde and Cancer, University of Helsinki, Finland in dark glass tubes by courier; the total transportation time was 4 weeks. In the laboratory, the samples were anonymized and stored at 4–6°C until used for milk fermentation and analysis.

A protocol provided by the Kalenjin families was used for milk fermentation. Finnish whole milk (Arla-Ingman) was used in the process. It was first briefly boiled and allowed to cool before mixing 0.1 mL of each starter culture with 9.9 mL of the milk. This suspension was prepared in sextuplicate, each to be analyzed at different time points. The suspensions were sealed and left to ferment at room temperature (22–25°C) for 0, 24, and 48hours, and 4, 7, and 14 days. The pH of the fermenting milks was measured at each time point. The milk fermentation was conducted twice.

The ACH and ethanol concentrations in the fermented milks were measured at 3 time points (0, 24, and 48 hours) by gas chromatography (Perkin Elmer Headspace sampler HS 40XL, Perkin Elmer Autosystem Gas Chromatograph equipped with Ionization Detector FID) as described previously (24). These time points were selected based on the microbial growth curves, which reached a plateau by 48 hours. Aliquots of 500 μL were pipetted into 20 mL gas chromatography vials and closed airtight. The fermentation process was stopped by injecting 50 μL of perchloric acid through the rubber septum of the vial. To measure artifactual ACH, boiled whole milk without starter culture kept at room temperature in sealed vials alongside those with starter culture, were analyzed. ACH and ethanol levels measured in these control samples were subtracted as background. The measurements were done in triplicate twice. The means of 2 triplicate analyses are presented.

Microbial growth was enumerated [colony-forming unit (CFU)/mL] at 0, 24, and 48 hours, and 4, 7, and 14 days time points by dilution plating on lysed blood agar (BBL 211047) and incubation for 48 hours at 37°C. The microbial flora in starter cultures was determined by using selective and nonselective media in aerobic and anaerobic conditions. Fastidious anaerobe agar [FAA; (LAB-M, LAB 90)] supplemented with 5% v/v horse blood was used to enumerate the total cultivable bacteria. Lysed blood agar [BA; Trypticase soy agar (BBL 211047)] supplemented with 5% v/v horse blood was used for the enumeration of total aerobic bacteria. Neomycin-vancomycin blood agar [NV; blood agar and neomycin sulphate (Sigma N-1876, Sigma-Aldrich)] supplemented with vancomycin (7.5 μg/mL), menadione (0.5 μg/mL), and 5% v/v sheep blood was used to enumerate anaerobic gram-negative bacteria. Cysteine-, lactose-, and electrolyte-deficient agar [CLED; C.L.E.D medium (BBL 212218, BD, USA)] was used to select for aerobic gram-negative fermentative rods. To detect yeasts, Sabouraud Dextrose agar [SP (Lab M)] supplemented with penicillin (100,000 iu/mL) and streptomycin was used. The BA, CLED and SP plates were incubated at 37°C for 48 h and the FAA and NV plates were incubated under anaerobic conditions at 37°C for 5 days. From these primary plates multiple subcultures of bacterial and yeast colonies were prepared after incubation. The pure cultures were Gram-stained and identified by 16S or 18S PCR gene sequencing. In addition, yeast isolates were identified by API 32C auxanographic strips (bioMerieux) following the manufacturer's instructions.

Yeast isolates were identified by sequencing using ITS1 and NL4 primers spanning the ITS1-5.8S, rRNA-ITS2, and 26S rRNA regions (25, 26). The MycXtra DNA extraction kit (Myconostica) was used to isolate DNA. A modified cetyltrimethylammonium bromide (CTAB) method was used (27) for those isolates from which DNA was not extracted successfully using the MycXtra kit. Bacterial isolates were sequenced using PA and PH primers spanning the 16S rRNA region (28). Initially DNA extraction was conducted using Chelex 100 beads (29). If this was unsuccessful, a modified CTAB method was used (27). Sequencing was conducted at the University of Manchester DNA Sequencing Facility by using an ABI Prism 3100 Genetic Analyzer (16 capillary instrument) and by Beckman Coulter Genomics sequencing service (Essex, United Kingdom).

Sequences were compared and aligned with sequences available in GenBank using the NCBI-BLASTn matching tool optimized for highly similar sequences (30). Definitive identification based on sequence similarity in the ITS-5.8S-ITS2 region was based on a thorough BLAST match as defined by Nilsson and colleagues (31).

Data was analyzed using Graph Pad Prism version 5.00 (GraphPad Inc.). To analyze the correlation between ACH and ethanol levels, the means of triplicates measured for each milk at each time point were calculated and plotted. Spearman's rho (r_s) with a 95% confidence interval (CI) was used for the correlation analysis. P values of less than 0.05 were considered statistically significant.

During the fermentation process, the microbial counts increased exponentially up to 48 hours (Fig. 1A). The highest counts were detected at this time point. The counts ranged from 6.63 ± 0.34 log₁₀ CFU/mL (at 0 h) to 10.42 ± 0.21 log₁₀ CFU/mL (at 48 h). After 48 hours, the counts declined slowly to 9.54 ± 0.43 log₁₀ CFU/mL at the 14-day time point. The pH of the fermented milks declined linearly from 6.59 ± 0.01 at 0 hours to 3.46 ± 0.04 at 14 days. There were no significant differences in the total microbial counts or the pH between the different starter cultures used (data not shown).

The mean number of microbial species identified in the mursik milk starter cultures was 5 (range 2–8; Fig. 2). Yeasts were identified in all starter cultures (mean 1.5 species/milk) but their proportion of the total microbial count varied markedly (mean 35%, range 7%–90%, Fig. 1B). A total of 3 Candida species and 1 Saccharomyces species were identified (Fig. 2). The most common yeast species isolated from the mursik milk samples was Candida krusei (5/8 milks, Fig. 2). Candida kefyr and Candida sphaerica were the second most common yeasts found (3/8 milks). Saccharomyces fermentati was identified in 1 milk.

The dominant bacterial findings in the fermented milks were Lactobacilli present in 6 of 8 milks (mean 37%, range <1%–70% of total colonies; Fig. 1B). A total of 8 Lactobacillus species were identified (Fig. 2). The most frequent species of Lactobacillus was Lactobacillus kefiri, which was isolated from 6 milks. In 6 of the 8 milks Lactobacilli were found in combination with yeasts. In 5 milks L. kefiri was found in combination with Candida krusei. The next most common Lactobacillus spp. were Lactobacillus casei, Lactobacillus paracasei and Lactobacillus rhamnosus. Other species were less prevalent (mean 27%, range <1%–93%, of total colonies). Of these, Bacillus spp. were most common (5/8 milks, Fig. 2).

Seven of the 8 starter cultures produced mutagenic (>100 μmol/L) levels of ACH into the milk by the 24 hours time point (Fig. 1C). The highest level of ACH (1,808.7 ± 20.1 μmol/L) was detected in milk 1 at 48 hours. Four of the 8 starter cultures produced more than 1,000 μmol/L ACH, ranging from 1,146.2 to 1,808.7 μmol/L, at the 48 hours time point. Four starter cultures produced lower although potentially mutagenic levels of ACH by the 24 and 48 hours time points (74.8–547.4 μmol/L). Minimal levels of ACH and ethanol were detected in samples 7 and 8 at 48 hours (6.84 and 7.01 μmol/L; and 0.44 and 1.04 mmol/L, respectively). The highest levels of ethanol were detected in the 4 fermented milks with the highest ACH production (mean 79.4 mmol/L, range 61.8–106.2 mmol/L). Other milks contained 0–50 mmol/L of ethanol (Fig. 1D). The levels of ethanol produced during the fermentation process correlated well with the levels of ACH produced (r_s = 0.88; CI, 0.7331–0.9485; P < 0.0001).

The high levels of ACH were detected in samples where a combination of 3 to 5 species of yeast and Lactobacilli were identified (Fig. 1C and Fig. 2). Presence of a higher number of other species, especially Bacillus and Staphylococcus spp., resulted in lower levels of ACH and ethanol. Also, low levels were measured in samples where Lactobacilli were absent (Fig. 1 and Fig. 2).

Esophageal squamous cell carcinoma is a particularly frequent cancer in both males and females in southern and eastern Africa, including Western Kenya (7, 11). Formal epidemiologic studies have not yet been conducted to determine the dominant risk factors for OSCC in Western Kenya, but classical risk factors such as heavy alcohol drinking and tobacco smoking do not appear to explain the high rates of OSCC in this area.

ACH has been shown to be a mutagenic carcinogen in both animal and in vitro models at concentrations as low as 100 μmol/L (32–34). The mutagenic effects of ACH are mediated via binding to DNA, adduct formation, cross-linking and chromosomal aberrations (32). During and after alcohol drinking the majority of ACH is formed in the oral cavity by oral microbes (35). Certain yeast and bacterial species representing normal oral flora, such as Candida, Streptococcus, and Neisseria spp., are capable of producing significant amounts of ACH from ethanol in vitro (24, 36, 37).

The current study provides evidence that high levels of dietary ethanol and ACH are probably present in the traditional diet of many people in Western Kenya. Marked levels (>100 mmol/L) of ethanol and exceptionally high levels (>1,800 μmol/L) of ACH were detected during the fermentation process of the traditional mursik milk. This is in line with earlier studies reporting that fermented milk products may contain up to 3.8% w/v (0.8 M) ethanol and mutagenic concentrations of ACH (17–20). The ACH levels detected in mursik were as high as those found in many alcoholic beverages (34, 38, 39). Marked differences between different starter cultures were seen (6.84–1,808.7 μmol/L at 48 hours). This is in line with the wide variations reported between different alcoholic beverages, where levels from 0 to 25,000 μmol/L of ACH have been reported to be dependent on the microbes and technology used for fermentation. ACH levels detected in the present study were more than 4 times those detected in commercially available yogurts (20). In particular, monoculture of Candida kefyr has been shown to produce high amounts of ACH in fermented milk, and when incubated together with LAB the amount of ACH produced in the fermentation process can be significantly increased (16). Unlike yeasts, LAB are able to break down lactose into glucose and galactose, essential sources of energy for yeasts. And under the hypoxic conditions present in a bowl of milk (≤0.001% O₂), yeast can produce energy through ethanol fermentation, thus increasing the production of ACH (40).

In the Kalenjin tribe, mursik milk is ingested daily by most family members irrespective of age. The proportion of esophageal cancer cases occurring in patients 30 years of age or younger is exceptionally high amongst the Kalenjins, being 45-fold that seen in the United States (11). Repeated exposure of the esophageal mucosa to high levels of ACH may lead to mutagenic changes and finally to carcinogenesis. Regular consumption of home-fermented mursik milk from an early age could provide 1 explanation for the exceptionally high incidence of esophageal cancer in this tribe, and amongst the younger age groups in particular.

In the present study, ACH production in the milks significantly correlated with ethanol levels, reflecting the dynamic balance maintained by the alcohol dehydrogenase (ADH) enzymes of the fermenting microbes. Up to 106 mmol/L ethanol was detected in the mursik milks analyzed in our study. Previous in vivo studies have shown that as low as 10–20 mmol/L ethanol concentration in saliva results in more than 100 μmol/L salivary ACH levels (21). This is in line with in vitro findings, which show that many oral commensal Candida species are able to produce carcinogenic amounts of ACH (>100 μmol/L) in vitro when incubated in clinically relevant levels of ethanol (11 mmol/L; ref. 24).

The microbiota in mursik milks consisted mainly of lactobacilli and yeasts. This is in accordance with previous studies (15). The most common combination of species was C. krusei together with L. kefiri. C. krusei has been previously identified in African fermented milks (15, 16). However, other yeast species, such as C. kefyr and Saccharomyces sp., were also identified in these studies. In another study, 5 species of Lactobacilli were reported from a slightly larger set of fermented milks (kule naoto) produced by Maasai tribe members in Kenya (41). In our material, 8 different species of Lactobacilli were identified in the mursik milks and only 2 species were the same as in the Mathara study (38). However, molecular methods were not used for identification in their study. In addition, a number of other species such as Bacillus and Staphylococcus spp. were isolated in our study. The presence of these probable contaminants is not surprising as the fermentation process is neither industrial nor controlled. These species have also been previously reported from Kenyan and Sudanese milk products (42, 43).

In our material, a combination of yeast and lactobacilli, especially C. krusei with L. kefiri, with the exclusion of other species, seemed to correlate with higher ACH and ethanol levels. We have previously shown that human commensal C. krusei isolates are poor producers of ACH from ethanol in pure cultures (24). C. krusei cannot metabolise lactose, but is able to use the breakdown products made available by lactose metabolism by Lactobacilli. On the other hand, C. kefyr can use lactose for fermentation independently. In a previous study C. kefyr was reported to produce high levels of ACH and ethanol in coculture with lactobacilli (44). Generally, where Lactobacilli were absent ACH production was minimal or very short lived.

Charcoal treatment of the fermentation gourds may modify the odor, taste, and flavor of mursik by decreasing the pH of the milk and impacting on the growth and metabolism of the microbes. We used starter cultures, which had been prepared this way, but we were unable to obtain similar gourds to replicate this special treatment for our laboratory experiments. Nevertheless, we measured highly mutagenic levels of ACH from the milks even in the absence of charcoal treatment. Charcoal-derived carcinogens have been considered to be a contributing factor for the high esophageal cancer incidence in Iran (45). Charcoal can be a source of polycyclic aromatic hydrocarbons, some of which have been identified as carcinogens and could contribute to the carcinogenicity of traditionally prepared mursik (46). On the other hand, charcoal could also act in a protective manner by binding toxins and carcinogens in the fermenting milk. In the present study, the gas chromatograph analyses were limited to the 48 hours time point, the time when mursik is ready for consumption. It is highly likely that fermentation continues after this whereby production of ACH and ethanol in longer fermentations needs to be addressed in future work. However, the impact of longer fermentations has been analyzed for kefir starter cultures where ACH and ethanol levels were shown to remain high beyond 7 days of fermentation (47). The transportation of the mursik milk samples from Kenya took 4 weeks that may have influenced their viability. However, the storage conditions during transportation resembled those used by Kalenjins, who often store their starter cultures at room temperature for long periods of time. Also, it is possible that Kenyan milk may contain fewer carbohydrates than Finnish milk but any such differences are likely minimal.

In conclusion, our study shows that significant levels of ethanol and ACH are produced during mursik fermentation. When ingested several times daily, the esophageal mucosa is inevitably exposed to high levels of carcinogenic ACH, which may lead to carcinogenesis. The incidence of esophageal cancer in Kenya is especially high among certain tribes in Western Kenya, including the Kalenjin tribe, and the mursik milk samples analyzed in this study were all collected from Kalenjin families. It is also possible that predisposing genetic factors, such as ADH or aldehyde dehydrogenase (ALDH) polymorphisms, among closed tribal communities may have an additional role in the high incidence of esophageal cancer. The prevalence of these polymorphisms in different tribal communities in Kenya remains largely unknown, although some initial studies have been done (48). Thus, future studies on genetic factors and molecular mechanisms of ACH-induced carcinogenesis are warranted. Meanwhile, finding alternative fermentation processes and starter cultures leading to minimal ACH production could be a major public health measure in the prevention of esophageal cancer in Kenya.

P. Bowyer is an employee in Astra Zeneca and has honoraria from Speakers Bureau in Pfizer as well as has ownership interest in Alergenetica SL. M. Salaspuro is a board member and medical advisor as well as a member of the advisory board of Biohit OYJ. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M.T. Nieminen, C.C. Abnet, R.E. White, N.D. Freedman, P. Bowyer, M. Salaspuro, R. Rautemaa.

Development of methodology: M.T. Nieminen, L. Novak-Frazer, R.E. White, P. Bowyer, M. Salaspuro, R. Rautemaa

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.T. Nieminen, L. Novak-Frazer, R. Collins, S.P. Dawsey, S.M. Dawsey, C.C. Abnet, N.D. Freedman, R. Rautemaa

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.T. Nieminen, L. Novak-Frazer, C.C. Abnet, N.D. Freedman, P. Bowyer, M. Salaspuro, R. Rautemaa

Writing, review, and/or revision of the manuscript: M.T. Nieminen, L. Novak-Frazer, R. Collins, S.P. Dawsey, S.M. Dawsey, C.C. Abnet, R.E. White, N.D. Freedman, M.M. Mwachiro, P. Bowyer, M. Salaspuro, R. Rautemaa

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.T. Nieminen, L. Novak-Frazer, P. Bowyer, M. Salaspuro

Study supervision: M. Salaspuro, R. Rautemaa

The authors thank Pertti Kaihovaara for his skillful technical assistance with the gas chromatography. The authors also thank the investigators from Tenwek Hospital who provided them with samples.

This study project was supported by Grants from Finnish Dental Society Apollonia (UNID:031510), Yrjö Jahnsson Foundation (#6355), Sigrid Juselius Foundation (2009/642/Salaspuro), and Helsinki University Hospital research grant EVO T1020V0017. This study was also supported in part by the Intramural Research Program of the Division of Cancer Epidemiology and Genetics of the National Cancer Institute, NIH.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.