Aflatoxin M₁ Enrichment in Buffalo Mozzarella and Ricotta: Case Study on Naturally Contaminated Milk and Exploratory...

Mycotoxins, including sterigmatocystin (STC), aflatoxicol (AFL), aflatoxin B₁ (AFB₁), and its metabolite aflatoxin M₁ (AFM₁), represent a major food safety concern in the dairy industry due to their high toxicity and thermal stability. While AFM₁ in milk is strictly regulated, there is a substantial lack of specific guidelines and scientific data regarding its behavior and enrichment factors (EFs) in buffalo-derived products. Furthermore, knowledge gaps persist regarding the transfer and stability of co-occurring molecules like STC and AFL. This study evaluated the distribution and fate of AFM₁, STC, and AFL during traditional, laboratory-scale cheesemaking of buffalo mozzarella and ricotta. Fourteen trials were conducted using bulk tank milk naturally contaminated with AFM₁, alongside an exploratory trial utilizing toxin-free milk artificially spiked with all three targets. Mass balance and specific EFs were calculated across all processing streams. AFM₁ was detected in all raw milk samples, showing a strong affinity for caseins. Approximately half of the initial milk AFM₁ content was retained in mozzarella (50.3 ± 3.3%), whereas ricotta retained only a minor fraction (6.9 ± 0.7%), with the majority lost in the second whey (30.7 ± 1.3%). The mean EFs were 2.3 ± 0.3 for mozzarella and 1.1 ± 0.2 for ricotta. AFL was never detected in milk samples while STC was mainly present at concentrations < 1 ng/kg. In the spiked trial, AFL demonstrated stability with an EF of 1.5 in mozzarella, while STC displayed potential degradation within the milk matrix and during sample storage. The experimental EFs determined for buffalo mozzarella and ricotta are lower than the standard factors currently applied by Italian regulations for non-bovine dairy products. These accurate, matrix-specific estimations are essential to prevent the under-reporting of non-compliant samples and to refine international consumer risk assessment models.