A three-years survey of microbial contaminants in industrial hemp inflorescences from two Italian cultivation sites | Journal of Cannabis Research
Plant material, cultivation, harvesting, exsiccation of inflorescences, and sampling
Six hemp (Cannabis sativa L.) chemotype III varieties (Carmagnola, Carmaleonte, Codimono, CS, Eletta Campana, and Fibranova) and the chemotype IV (CBG rich) Bernabeo have been cultivated for three consecutive years (2019–2021) in two experimental farms (with three replicated parcels of 20–25 m2 each): “Busa Carrare”, located in Rovigo (RO; 45°04′45.4″N 11°45′57.3″E) and “ Libertinia”, located in Catania (CT; 37°32′25″N 14°34′41.0″E). They were all Italian varieties of the Italian/EU register of plant varieties, characterized by a prevalence of CBD (chemotype III) or CBG (chemotype IV) and a THC level below 0.2% (Table 1). The field experiment was carried out between April and October (according to the length of the life cycle of the hemp plants), with variations in sowing and harvesting dates differing between sites and among genotypes (Table 2). Nitrogen fertilization (40/60 units ha) was applied before sowing and irrigation was done only during seedling emergence as needed. Meteorological information (temperature and precipitation) was also collected, and the monthly averages are reported in Table 2. Female or monoecious inflorescences were taken at seed harvest time from three different parcels/repetitions for each genotype, dried at ambient temperature and with natural ventilation in local greenhouses for 48–72 h until 144 h for the later harvests of Rovigo. In both cultivation areas, for each harvesting season, the exsiccated biomass was trimmed and two samples of approximately 50 g (with inflorescences, floral bracts, and leaves) of all varieties were sealed in sterile bags, resulting in 42 duplicated samples, collected for microbiological contamination and phytocannabinoids content analyses.
Microbiological analyses
The composition of all media utilized for microbiological analysis is reported in Supplementary Table 1. All the materials, unless otherwise indicated, were purchased by Merck Life Science S.r.l. (Milan, Italy) or by Biolife Italia (Milan, Italy). In all the procedures microbial growth was obtained in static incubators (Binder GmbH, Tuttlingen, Germany) kept in aerobic conditions.
The microbial contamination of dried inflorescences was evaluated by culture-dependent methods in compliance with the quality parameters for herbal medicinal products required by the European Pharmacopoeia 9th Edition (Council of Europe 2019), that provide a detailed description of parameter, protocol, and material for the microbial analysis. In particular, the total aerobic microbial count (TAMC), the total combined yeasts/moulds count (TYMC), and the semiquantitative estimation (order of magnitude) of bile-tolerant Gram-negative Bacteria (BTGNB) were determined and expressed as colony forming units (CFU) per gram. The absence of Escherichia coli and Salmonella spp. (in 1 and 25 g, respectively) was assessed.
Twenty-five grams of sample were homogenized in 225 ml of casein soya-bean-digest broth (CSBDB) supplemented with 1 g/L of polysorbate for two minutes in a filter bag by using a blender. tenfold serial dilutions in CSBDB were prepared for TAMAC and TYMC estimation. Petri dishes of agarized CSBDB and Sabouraud-dextrose agar, the latter supplemented with chloramphenicol 50 mg/L, were seeded by surface-spread method, and incubated at 35 °C for 3 days and 25 °C for 5 days, respectively.
For semi-quantitative estimation of BTGNB, after a 2–3 h of incubation at 25 °C for bacterial resuscitation, the initial sample suspension was serially diluted in enterobacteria enrichment broth-Mossel and incubated at 35 °C for 24 h. Each dilution was plated in violet-red bile glucose agar and plates were incubated at 35 °C for 48 h. The growth of colonies represented a positive result for the corresponding dilution.
To test E. coli absence, 10 ml of suspension in CSBDB containing 1 g of sample was incubated at 35 °C for 24 h, then 1 mL of the enriched was seeded in 100 mL of MacConkey broth, incubated at 42 °C for 24 h, and the suspension plated on MacConkey agar, with plates incubated for 48 h at 35 °C.
To verify the absence of Salmonella, the sample suspension in CSBDB was incubated at 30 °C for 24 h, then 0.1 mL was diluted with 10 mL of Rappaport Vassiliadis Salmonella enrichment broth and the sample was incubated at 35 °C for 24 h. The enriched broth was then inoculated on a plate of xylose, lysine, and deoxycholate agar (XLDA). After incubation at 35 °C for 48 h, Salmonella contamination was revealed by the presence of well-grown red colonies, with or without a black center.
Phytocannabinoids analyses
Phytocannabinoids were extracted from hemp biomass samples following the protocol reported in the monograph of cannabis flos included in the German Pharmacopoeia and adapted in our previous works (Tolomeo et al. 2022). Ethanol extraction was carried out on 500 mg of finely grounded hemp biomass in three cycles (20 mL, 12.5 mL, and 12.5 mL). The combined extracts were brought to 50 mL final volume with fresh EtOH in a volumetric flask. A 1 mL aliquot of the extract was centrifuged at 4000 × g for 5 min and filtered through a 0.45 µm regenerated cellulose filter, then diluted 10 times with mobile phase (acetonitrile/H2O, 60:40, v/v, 0.1% v/v formic acid).
Five µL of each sample were injected into the analytical apparatus Vanquish Core System (Thermo Fisher Scientific, Bremen, Germany) equipped with a binary pump, a vacuum degasser, a thermostated autosampler (4 °C) and column compartment (30 °C), a Poroshell 120 EC C18 column (100 × 3.0 mm I.D., 2.7 µm particle size, Agilent Technologies, Santa Clara, USA) and a diode array detector (DAD). The chromatographic parameters were adapted from a previously validated method with slight modifications (Tolomeo et al. 2022). The phytocannabinoids were separated with a constant flow rate of 0.5 mL/min, applying a gradient of acetonitrile from 60 to 95% in 15 min and an isocratic step at 95% acetonitrile held for 3 min, followed by a washing step of 4 min at 98% acetonitrile and a re-equilibration step at 60% acetonitrile for further 4 min. The analyses were acquired in the whole UV spectrum (190–400 nm) and chromatographic traces processed after filtration of the wavelength at 228 nm.
Quantification of phytocannabinoids was achieved using certified reference standard solution of cannabidiolic acid (CBDA), tetrahydrocannabinolic acid (THCA), cannabigerolic acid (CBGA), CBD, Δ9-THC, Δ8-THC, and CBG (1 mg/mL, Cerilliant, Merck Life Science S.r.l., Milan, Italy) diluted with mobile phase to get six non-zero calibration points at 0.1, 0.5, 1.0, 2.5, 5.0, and 10 µg/mL and analyzed with the same conditions used for the samples. Limit of detection (LOD) and limit of quantification were established at 0.03 and 0.1 µg/mL respectively. Linearity was assessed by the coefficient of determination (R2), which was greater than 0.997 for all cannabinoids (detailed information on calibration data in Supplementary Table 2). The back-calculated concentration was considered acceptable if it did not exceed 15% of the nominal value (and 20% for the LOQ). The raw data were processed with Chromeleon 7 (Thermo Fisher Scientific, Bremen, Germany) and the area of the peaks of the cannabinoids under investigations were used to calculate their concentration based on the respective calibration curves. The total content of phytocannabinoids was calculated using the formula: Total = 0.877*C + D, where C represents the amount of carboxylated species (either CBDA, Δ9– and Δ8-THC, or CBGA), and D represents the amount of decarboxylated species (either CBD, THC, or CBG).
Statistical analysis
The variation of microbial contamination (TAMC, TYMC and BTGNB), as well as the presence or absence of E. coli and Salmonella spp., were inspected by considering three key factors: i) the different varieties, ii) the cultivation years, and iii) the geographical sites. This analysis was carried out using a Full Factorial Design of Experiments (DoE) (Leardi 2009), which estimates linear and interaction effects of the factors, considering all sources of information simultaneously. Forty-two samples, including all the possible combination between the different levels of each factor, i.e. seven varieties, three cultivation years (2019. 2020 and 2021) and two sites of cultivation (Rovigo and Catania), were randomly analyzed in duplicate to assess the significance of the factors with genuine replication (Box et al. 2005). The responses (TAM and TYM counts, BTGNB magnitude, and E. coli or Salmonella spp. presence/absence) were modeled using a Partial Least Squares (PLS) algorithm to simultaneously analyze all variations. The regression coefficients from PLS were used to estimate the effects of each factor on the response, identifying the influential parameters at high or low levels. Logistic regression (Christensen 2006). was employed within the DoE method to analyze the binary responses of E. coli and Salmonella spp.. Together, these statistical tools offer a methodology to identify optimal combinations of factors within a set of experiments, aiming to maximize classification accuracy (Lòpez et al., 2008).
The correlation between microbial contaminants and phytocannabinoid content was assessed using the Spearman coefficient, which is based on ranks. This robust measure does not rely on the assumption of a normal distribution of data.
Measures were reported as mean values ± standard deviation. The statistical analyses were performed using IBM SPSS Statistics 21.0 (IBM Corp., Armonk, NY, USA) and Design Expert v12 (Stat-Ease, Inc., Minneapolis, MN, USA).
