Analytical Techniques for Phytocannabinoid Profiling of Cannabis and Cannabis-Based Products—A Comprehensive Review

Cannabis is gaining increasing attention due to the high pharmacological potential and updated legislation authorizing multiple uses. The development of time- and cost-efficient analytical methods is of crucial importance for phytocannabinoid profiling. This review aims to capture the versatility of analytical methods for phytocannabinoid profiling of cannabis and cannabis-based products in the past four decades (1980–2021). The thorough overview of more than 220 scientific papers reporting different analytical techniques for phytocannabinoid profiling points out their respective advantages and drawbacks in terms of their complexity, duration, selectivity, sensitivity and robustness for their specific application, along with the most widely used sample preparation strategies. In particular, chromatographic and spectroscopic methods, are presented and discussed. Acquired knowledge of phytocannabinoid profile became extremely relevant and further enhanced chemotaxonomic classification, cultivation set-ups examination, association of medical and adverse health effects with potency and/or interplay of certain phytocannabinoids and other active constituents, quality control (QC), and stability studies, as well as development and harmonization of global quality standards. Further improvement in phytocannabinoid profiling should be focused on untargeted analysis using orthogonal analytical methods, which, joined with cheminformatics approaches for compound identification and MSLs, would lead to the identification of a multitude of new phytocannabinoids.


Introduction
Cannabis sativa L. (C. sativa L.), from the family Cannabaceae, is the most widely cultivated, trafficked, consumed and investigated, yet most notorious and controversial, plant in the world [1,2]. It is one of the oldest known crops to humanity, with first records of use dating to 3000 B.C. [3], and one of the most commonly used plants for industrial and medical purposes, with a global legal market expected to reach 147 billion USD by the end of 2027 but also the world's most widespread drug of abuse [1], comprising around 200 million global users.
1.1. Botany of C. sativa C. sativa is an annual dioecious plant with histaminate male and pistillate female flowers on separate plants. It grows up to 5 m height, with serrated leaves with a distinct vein pattern that extends to their tips [4]. The inflorescences of the female plants produce  Figure 1. Mid−IR spectra from the main cannabinoids; THCA, THC, CBDA, CBD were adapted from [21,22]; CBN, CBGA and CBG were adapted from [18].

Use of C. sativa
For more than 12,000 years, Cannabis spp. is used as a source of textile fiber and food worldwide [23]. The earliest data for medical use of Cannabis by the Assyrians goes back to 3000 B.C. Follow-up records date from around 2700 B.C. in China, where C. sativa was used as a medicine for menstrual fatigue, rheumatism, malaria, constipation and other conditions. It was later used by other ancient civilizations-the Egyptians (ca. 1700 B.C.), the Indians (ca. 1600 B.C.), the Persians (ca. 750 B.C.), the Greeks and the Romans (ca. 450 B.C.). Historically, Cannabis was also used for additional indications such as glaucoma, anal fissures, diarrhoea, as obstetric aid and as anxiety relief [23]. The plant was introduced to the modern western medicine in the early 19th century, and was mainly indicated for treating pain, glaucoma, nausea, depression and neuralgia [24].
Today, medical use of C. sativa includes multiple indications supported by reliable clinical evidence. Such are the treatment of chronic pain, mutiple sclerosis, resistant epilepsy and chemotherapy-associated nausea and vomiting, apetite and weight loss associated with HIV/AIDS, Tourette syndrome, anxiety disorders, sleep disorders, post-traumatic stress disorder and schizophrenia [1,25,26]. Multiple additional health benefits of C. sativa extracts are reported in in vitro and in vivo trials, such as lowering blood cholesterol, triglycerides and blood pressure, antioxidant and antimicrobial activity [27]. C. sativa and its extracts are also used in the treatment of dermatitis and degenerative imunological diseases, both as supplements and as traditional medications [25]. Cannabis oils, oral solutions, oil-like concentrates and tinctures used orally and sublingually, lotions, balms, creams, bath salts, salves, gels, patches and other topical products as well as rectal and vaginal products (suppositories, tablets) are the most common pharmaceutical dosage forms employed [26].
Moreover, hemp is additionally used in food and beverage production, as hempseeds are shown to have great nutritional value: high content of protein, containingnine essential amino acids, dietary fibers and an ideal ratio of ω-6:ω-3 fatty acids (3:1). The same applies to hempseed flour and oils, which have high content of proteins, insoluble fibers and polyunsaturated fatty acids. The "cannabis edibles" are the latest type of cannabis-based products that recently became popular. Chewing gums, lollipops, caramel hard candy, berry gummies, lozenges, candy bars, jam, tea, soda, coffee, water, honey, etc., containing ∆ 9 -THC-and/or CBD-dominant extracts and concentrates are widely marketed [28]. Today, cannabis is the most commonly used illicit drug worldwide, despite the strict international control for more than eight decades [9].
1.4. Legal Aspect of C. sativa C. sativa and/or cannabis-based products have been legalized for medical use in 41 countries (23 in Europe) between 2012-2021. As C. sativa and cannabis-based products are classified based on the ∆ 9 -THC content with psychotropic properties, and most of them contain drug-type C. sativa extracts, various legal limitations in many countries still exist. In total, 50 countries in Europe, Asia, North and South America use the plant for industrial purposes [29]. Cultivation and supply of 69 C. sativa varieties [30] with ∆ 9 -THC content not exceeding 0.2% is legal in EU [31], with some exclusions (Czech Republic and Austria, <0.3%, Switzerland 1.0%).The industrial use of C. sativa is focused on production of >2500 products used in agriculture, textile production, recycling, automotive industry, furniture production, paper industry, production of construction materials, energy production, personal care products and medical supplements [29].
Cultivation and use of marijuana, the crude drug derived from C. sativa for recreational purposes, is not legalized in Europe, but it is decriminalized in 32 countries worldwide (16 in Europe), with various limitations regarding the amount of dry marijuana, number of cultivated plants and punishing public consumption [32].

Incentive for Investigating Phytocannabinoids in C. sativa and Cannabis-Based Products
Under the pressure of its criminal association, the chemical constitution, pharmacological effects, genetic structure, evolutionary and domestication history of remained poorly understood until the last decade of the 20th century. Authorized investigations related to C. sativa were either forensic studies to aid law enforcement or medical and social research specifically intended to document and reduce harmful effects [6]. Since the last decade of the 20th century, a great urge for more thorough investigation of C. sativa appeared, mainly as a result of the resurrection of production of C. sativa for non-narcotic and medical purposes and the growing tolerance of the extremely widespread recreational use. This increased attention C. sativa and cannabis-based products gained due to their high pharmacological potential, updated legislation authorizing many different uses, and, thus, the emerging need to control their quality. This imposes a great challenge for academics, particularly in the field of natural products, from which a contribution to improve and standardize the extraction and characterization of the bioactive compounds from C. sativa species is expected.
Scientific and technological development in regards to C. sativa began, highlighting the need of sensitive, specific and robust analytical methods for identification and quantification of the active constituents of C. sativa. Chemical profiling of the plant became extremely relevant, since the acquired knowledge further enhanced: (1) chemotaxonomic classification; (2) cultivation set-ups examination, and thus adjustment of cultivation conditions and breeding methodologies in order to produce C. sativa varieties with fit-for-purpose physicochemical properties; (3) investigation of potency of seized samples, thus discovering sources of interconnected illegal production and trafficking; (4) association of medical and adverse health effects with potency and/or interplay of certain phytocannabinoids and other active C. sativa constituents and (5) QC of medical cannabis and final medical cannabis-based products and potency examination [1,2].
The multitude of cannabis-based products including cannabis extracts, oils, resins, pharmaceutical dosage forms, cannabis-infused edibles and beverages are obliged to comply with the national regulatories' quality control regulations, especially in terms of phytocannabinoids content. Accurate qualitative and quantitative analyses of the phytocannabinoids content and chemical profile in cannabis plants are extremely relevant, in order to associate medicinal and possible adverse health effects with the potency of certain phytocannabinoids and other compounds, such as terpenoids [2]. One of the most relevant problems in analytical determinations for QC, especially when there are legal problems related with quantitation, such as for cannabis, relates to the proficiency of laboratories. Both qualitative and quantitative determinations require carrying out of standardized assays that meet the analytical criteria approved by the relevant control authorities.
This review aims to capture the versatility of analytical methods for natural phytocannabinoids profiling in cannabis and cannabis-based products in the past four decades . As such, this thorough overview is first of its kind. Other most recent reviews cover either a shorter time period, i.e., 2002-2016 [1], 2010-2016 [2], 2009-2019 [39], focus on both plant materials and biological matrices [1,39], describe multiple [2] or single (GC) [39] instrumental analytical platforms, omit cannabis-based products [4] or, apart from phytocannabinoids, also include profiling of other bioactive C. sativa constituents [1].

Analytical Methods for Phytocannabinoid Profiling
The overview of the analytical methods for phytocannabinoid profiling used in phytocannabinoid profiling of cannabis and cannabis-based products is schematically presented in Figure 2.

Sample Preparation Techniques
Rapid and simple extraction methods are essential for time-and material-efficient highthroughput phytocannabinoid profiling. Optimization of the three key parameters during extraction is crucial for the overall analytical method [40], which are: (1) the granulometry of the solid sample; (2) the system temperature and (3) the affinity of the extraction liquid towards compounds of interest [41]. C. sativa and cannabis-based products are very complex and inhomogeneous matrices, as their different parts may have different cannabinoid profiles due to the variety of phytocannabinoids, terpenes and other volatile compounds and high sugar and fat content. Thus, extraction of phytocannabinoids from plant material and cannabis-based products in an efficient and consistent manner, with acquisition of accurate and reliable potency data can be a challenging task. In addition, there are no standardized preparation procedures for the hemp-based infusions (hemp leaf, hempbased tea mixtures), and the cannabinoid content could be significantly affected by the infusion preparation procedure [42]. Finally, there is large and unpredictable variability of the average composition as a result of genetic and environmental differences, making the efforts for standardization of sample preparation techniques an ultimatum [43].
Initial steps of sample preparation include mechanical preparation aiming to increase the contact surface between the solvent and the active ingredients. It is of special importance for cannabis plant material, since, despite that most of the active resin is claimed to occur in the superficial glandular trichomes, significant amounts are found in non-glandular tissues. Consequently, immersion of unbroken fresh plant material would give unsatisfactory extraction [44]. Cannabis or its resins are reduced to small pieces by a grater [5] or spatula [45], grinded or pulverized, while cannabis oils are directly proceeded to instrumental analysis. Manual pulverization and homogenization of the dried plant material can be performed using mortar and pestle [46][47][48][49][50], metal spoon [51] or glass rod [52], by cutting the plant material [53] or crushing and riddling (0.5 mm) [54] or by manual grinder [50,55]. According to the UNODC [5], dried herbal cannabis material and cannabis resins should be pulverized by a cutter (at high revolution speed, i.e., 100 rps) and sieved (mesh size 1 mm). Plant samples can also be homogenized in a crucible [56], in laboratory blender, usually to 60-80 mesh (177-250 µm) particle size, but sometimes a larger (≤335 µm) or smaller (100-150 µm) particle size is required [57][58][59]. Unlike classical, mechanical hand grinders and electrical grinders, some sample preparation methods employ superfine grinding of cannabis plant material [60][61][62]. However, manual grinding with a handheld herb grinder resulted in higher yield of total phytocannabinoids (17.5 ± 0.5%) than with electric blender (12.0 ± 0.3%). The minimization of analyte loss using manual grinder is attributed to the adhesion of cannabis resin to the blades and plastic housing surface of a plastic blender during the high-speed pulverization [50]. Mechanical grinding-activation in an intensity planetary vibrational mill [62,63], ball mill [55], knife mill [64] or freeze mill [65] are also applied. Instead of drying, fresh cannabis plant material can be frozen with liquid nitrogen and crushed [66,67] or frozen, lyophilised at −50 • C and grounded by hand [68] or in a mill [69]. Comparison of coarse homogenization by sieving through a 1-mm mesh and fine homogenization with a ball mill revealed better extraction efficiency for CBDA and THCA for the finely powdered plant material, and no difference for the neutral phytocannabinoids (CBD, ∆ 9 -THC) [55].
In the next step, thermal processing occurs, which aims to remove moisture, usually to 8-13% residual humidity, as recommended by UNODC [5], achieved by drying at room temperature for several days or at 70 • C until the leaves become brittle or, according to EC, within 48 h using any method below 70 • C [70]. Lower drying temperatures should be avoided, as they result in mould growth [71]. Despite this, drying is frequently performed at variety of temperatures and durations, such as at 135 • [86], for 4 weeks [87] or until a residual humidity ≤12% is achieved [47].
Dried and mechanically processed samples are extracted using maceration, LLE, PLE, HS-SPME, SFE or FUSE. Other extraction techniques, including ultrasonic assisted extraction (UAE), microwave assisted extraction (MAE), dynamic maceration (DM) and accelerated solvent extraction (ASE), are faster and use less extraction fluids than the "classic" maceration. The number of the consecutive extractions did not have significant effects on total phytocannabinoid yield. Yet, phytocannabinoids yield after sonication was found to be slightly lower than the yield obtained by one-day DM [50]. Phytocannabinoid extraction was omitted in only one study, which, consequently, reported low sensitivity (Table S1); thus difficulties occurred during quantification of trace phytocannabinoidsin cannabis plant tissues [88]. The summary of the properties of the most frequently used sample preparation techniques are given in Table 2.
The conventional sample preparation methods for cannabis plant material are maceration and LLE using versatile organic solvents with great affinity towards phytocannabinoids. Although universal and simple, they are time-consuming and not environment-friendly, as they require large quantity of organic solvents.
Maceration and LLE of semisolid fatty/oily matrices of cannabis-based products, such as butter, margarine, chocolate bars, nonpolar topical ointments and balms requires warming of the samples on a hot plate after addition of the extraction solvent in order to melt the matrix. Prior to maceration/LLE, lactose-containing cannabis products (cream cheese, butter, coffee beverages with milk) require lactase pre-treatment/lactose hydrolysis in order to avoid matrix interferences [90]. For high sugar and carbohydrate matrices including hard candies, honey and fruit preserves, a matrix trapping effect occurs, resulting in the formation of glassy, impervious precipitates which are hard to extract. To avoid this, the aqueous portion of the extractant is added first, followed by warming the matrix in order to dissolve the sugars and carbohydrates, and addition of ACN (to 83-91% final proportion) [90]. Maceration and LLE of cannabis-based products is further aggravated by the presence of glycerine and propylene glycol, especially in oral supplements and vape products. They interfere cannabinoid profiling significantly. High levels of co-extracted glycerine or propylene glycol may swamp the silylating derivatization agents, disabling complete derivatization of CBD to CBD-2TMS, along with undesired side conversion of ∆ 8 -THC and ∆ 9 -THC. The problem can be avoided with CAN extraction, as sugars and glycerine have much lower solubility in ACN than in EtOH [90].
While most of the methods include vortexing, ultrasonication in bath and centrifugation immediately after solvent addition, older methods include immersion, e.g., for 1 h [9,75,121,143,144], overnight [121], 10-days soaking [123], several hours' maceration [87] or heating near boiling for few hours in solvent, followed by separation of the liquid extract by filtration [108]. Only one study evaluated the shaking time during extraction with MeOH/CHCl 3 (9:1, v/v) at room temperature in a range from 10 min to overnight shaking and showed that 20 min of shaking gained sufficient extraction efficiency for ∆ 9 -THC, CBD and CBN [101]. Finally, exhaustive extractions in Soxhlet apparatus are rarely performed [95,117]. In terms of temperature conditions, extraction is performed with highest efficiency at room temperature, with rare exemptions (e.g., at 4 • C) [77], in order to avoid conversion of phytocannabinoids from plant material [1]. If additional heating is required (e.g., when no preliminary decarboxylation was performed), hot extraction (e.g., at 70-78 • C) is performed by sonication [92] or in Soxhlet apparatus [95,117].
Solvent exchange may also be included as final step, by diluting extracts with solvents and solvent mixtures more similar to the mobile phases employed, in case of (HP)LC analysis. For example, cannabis-infused chocolate is first soaked in IS solutions and isopropyl alcohol, extracted with MeCN + 1% acetic acid and finally diluted with MeCN prior to LC-MS/MS or MeCN/H 2 O (75:25, v/v) prior to HPLC-UV analysis [65]. Solvent exchange is also performed prior to GC analysis, especially when derivatization is performed [7,80,87,104,121,123,133]. Extracts are reduced to dryness usually under gentle N 2 steam, causing least damage to total extracted amounts of phytocannabinoids and terpenoids than drying in rotary evaporator or in a speedvac, the latter reducing the concentrations of ∆ 9 -THC and CBG for two-thirds [66]. Reduced extracts can be dissolved in solvents (e.g., pyridine and benzene [116], CHCl 3 [5], MeCN [116], dry EtAc [55]), in a mixture of derivatization agent and solvent (e.g., toluene and BSTFA [94], pyridine and BSTFA + 1% TMCS [118], pyridine, isooctane and MSTFA [99], pyridine and BSTFA [90], pyridine and MSTFA + 1% TMCS [84]) or directly in the derivatization agent [80,99,100,133]. As silylation agents are harmful for GC injection port and column, additional evaporation to dryness is frequently performed, followed by dissolution in solvent, e.g., n-hexane [84] or MeCN [95]. If no derivatization is performed, EtAc [120,126,127] and EtOH [66,123] are the most commonly used solvents for reconstitution prior to GC analysis.
Before instrumental analysis, decarboxylation may be also introduced, initiating thermal degradation of phytocannabinoid acids to neutral counterparts for the purpose of accurate phytocannabinoid profiling and potency examination. Dried extracts are most commonly heated at 150-210 • C for 10-30 min and reconstituted in the same solvent or solvent mixtures [7,52,60,101,128], or, alternatively, at 50 • C for 180 min and then 145 • C for 15 min [69]. A total 15 min of decarboxylation at a temperature range of 120-180 • C showed that maximum yield is achieved at 140-160 • C, with no significant within-range differences [115]. Decarboxylation temperatures higher than 160 • C should be avoided, as ∆ 9 -THC is oxidized to CBN [115] and isomerised to ∆ 8 -THC [87]. However, it is almost impossible for decarboxylation to yield 100%, which initiates significant discrepancies in potency data. Laboratories quantifying the total ∆ 9 -THC as the sum of the ∆ 9 -THC already present in the plant and ∆ 9 -THCA get higher values than laboratories that perform decarboxylation prior to instrumental analysis [115].
HS-SPME is a solvent-free sample preparation method used for analysis of phytocannabinoids in the headspace over solutions or solid samples [5]. Multiple factors affect the extraction efficiency during HS-SPME, including SPME fiber coating, exposure temperature, extraction time and desorption time. Evaluation of the effect of the fiber coating on extraction efficiencies of ∆ 9 -THC, CBD and CBN from herbal cannabis samples showed that among polydimethylsiloxane (PDMS) 100 µm, PDMS/divinylbenzene (PDMS/DVB) 65 µm, Carboxen ® /PDMS and divinylbenzene/Carboxen ® /PDMS (DVB/Carboxen ® /PDMS) 50/30 µm, PDMS 100 µm performed optimally in general, although the PDMS/DVB fiber provided higher extraction efficiency for CBD, due to its higher polarity and affinity to PDMS/DVB. Among the three exposure temperatures (80 • C, 90 • C, 150 • C), 150 • C was optimal, simultaneously promoting volatilization and decarboxylation [59,148]. Extraction time depends upon matrix viscosity and lipophilicity that define the speed of diffusion of analytes from the liquid to the gas phase and, as a result, HS-SPME rate and efficiency. HS-SPME is more appropriate for simpler matrices (e.g., cannabis tea), as the extraction recoveries are proportional to the sample amount [99], while complex liquid-and protein-containing matrices cause significant matrix retention and lower recoveries, with higher LODs and lower method precision. For fatty/oily matrices, such as versatile hemp foods, alkaline hydrolysis with NaOH and Na 2 CO 3 is performed prior to HS-SPME in order to saponify the matrix lipids and reduce lipid matrix interferences [99]. Therefore, extraction time varies depending upon the sample matrix, from 10 min for herbal cannabis [59,148] to 25 min for different hemp food products [99]. Finally, desorption time depends upon analytes' lipophilicities [148]. It is superior to LLE (n-hexane/EtAc (9:1, v/v)) in terms of chromatographic peak shape and matrix interferences, despite the good agreement of achieved LODs in food samples [99]. SFE uses supercritical fluids (SCFs) and liquefied gases as green solvents for extraction and fractionation of complex samples. SFE offers low solvent consumption and ensures stability of thermolabile and light-sensitive compounds. For the purpose of phytocannabinoid profiling, SFE is rarely used in sample preparation; its main purpose is to separate the aromatic fraction for further analysis. It is usually performed using supercritical CO 2 (SC-CO 2 ) as attractive SCF with solvent strength tuned by sensitive changes in temperature and pressure above the critical point (31.1 • C, 73.7 bar), conditions that are experimentally easy to reach [4], along with the low cost, short processing time and low environmental impact [149]. However, SC-CO 2 is a low polarity solvent that poorly dissolves phytocannabinoids; therefore, employment of co-solvent, usually H 2 O, alcohols and acids, to improve the overall extraction rate of phytocannabinoids is required. EtOH (5-20% in CO 2 ) is most commonly used co-solvent, added in constant flow [40,43,143,150,151] or in pulses [143]. Higher SC-CO 2 pressures offer lower extraction selectivity, but high initial extraction rate, apparent solubility and total yield, which are also a function of temperature, exposure time and phytocannabinoid content of the plant material [143]. A 90-94% extraction yields for ∆ 9 -THC, CBD and CBN are achieved at 100 bar, 35 • C and 1 mL/min flow during 10 min [43,150]; at 340 bar, 55 • C and 200 g/min maximal yields up to 92% are reported [143], along with satisfactory yields at milder conditions, (37 • C, 250 bar). Extraction efficiency was further improved by washing of the extract with fresh SC-CO 2 and addition of a cold separator (separating chamber) immediately after the sample containing chamber [149].
When compared to DM, UAE and MAE with same extraction solvent (EtOH) in same w/v ratio to sample, no significant difference existed between SFE and UAE, with the lowest extraction yields for CBD, CBDA and CBGA. DM and MAE showed higher yield for CBD and CBGA, but DM was selected as the optimal sample preparation technique (EtOH, room • C, 45 min). In case of MAE, increased CBD yield was accompanied by decreased CBDA yield, suggesting partial decarboxylation due to high extraction temperature [142].
FUSE is used for phytocannabinoid extraction from herbal cannabis, employing cyclohexane/isopropanol (1:1, v/v) in an ice-water media in order to avoid degradation and solvent evaporation, followed by centrifugation and filtration through 0.45 µm nylon filter [43]. FUSE is slightly more efficient than SFE for extraction of ∆ 9 -THC, CBD and CBN, as 80% of the phytocannabinoids are extracted at the first extraction, while <40% are extracted after the third extraction with pure SC-CO 2, and more than 90% are extracted with the first extraction in the presence of co-solvent. Therefore, adding the ability of SFE to separate terpenes from phytocannabinoids, and therefore minimize matrix interferences, SFE was selected as more optimal sample preparation technique [43].
PLE is one of the fastest and most efficient extraction techniques for plant metabolites. High extraction yields are achieved under pressure, using the extractant at a temperature above its normal boiling point, thus increasing its diffusion into the plant matrix. PLE on herbal cannabis is performed using n-hexane as extraction solvent at 100 • C and 40 bar for 15 min. MeOH and n-hexane are found equally efficient for ∆ 9 -THC and CBN, but not for ∆ 9 -THCA, as it is less soluble in n-hexane [95]. PLE with hot water, i.e., pressurized hot water extraction (PHWE) is used to yield CBD-rich extracts while supressing the THC and CBN content. Here, decarboxylation is performed in situ, i.e., in the extraction cell, heated in the oven prior to the dynamic extraction [80]. SPE with QuEchERS is used for purification of honey extracts, with high recoveries for CBDA, CBGA, ∆ 9 -THCA, CBG, CBD and ∆ 9 -THC and low intra-and inter-day variability. The method was more efficient than UAE with H 2 O at 40 • C, yielding homogeneous solution with no phase separation or solid residues, followed by extraction of phytocannabinoids from the aqueous phase through LLE with n-hexane or EtAc, observing higher phytocannabinoid yield in n-hexane extracts [40]. QuEChERS is also used for purification of MeCN extracts of hemp seeds, hempseed oil, hemp proteins, raw and skimmed milk, coffee and chocolate, using reaction mixture of MgSO 4 /NaCl/C 6 H 5 Na 3 O 7 × 2H 2 O/disodium hydrogen citrate sesquihydrate (4:1:1:0.5, w/w); with further supernatant dilution with MeCN/H 2 O (1:1) or with H 2 O. d-SPE with MgSO 4 , different combinations of C 18 , primary secondary amines and zirconia-coated silica sorbents were also evaluated. When PSA is used, cannabinoids were trapped by interactions with amines, resulting in low recoveries; other combinations achieved satisfactory extraction [144].
Cloud point extraction (CPE) involves employment of non-ionic surfactant, salt (Na 2 SO 4 ) and deionized water to extract ∆ 9 -THC from cannabis resin using heating (40-90 • C) and centrifugation. Despite the low extraction efficiency (60%), CPE offers many advantages, such as the possibility of extraction and pre-concentration of analytes in a single, solvent-free step and avoidance of analyte loss during solvent evaporation [145].

Instrumental Analysis
Influenced by the intense scientific and technological development in regard to C. sativa cultivation, analytical platforms for phytocannabinoid profiling in cannabis and cannabis-based products have intensively evolved over the last four decades (Figure 3). GC-and LC-based methods as most commonly used, have achieved comparable accuracy, selectivity, linearity, sensitivity and precision in phytocannabinoid profiling and are both used in routine and investigational analysis of cannabis and cannabis-based products. Despite the lack of standardization process for analysis of phytocannabinoids omitting the comparison of reliability of measurement among analytical platforms, a recent interlaboratory study concluded that GC-MS is the most accurate and robust analytical method for phytocannabinoid profiling, performing much better than GC-FID and UHPLC-MS/MS [55].
The advantages and disadvantages of the most frequently used analytical techniques for analysis of cannabis and different products are given in Table 4.

GC-Based Methods
GC coupled to versatile detectors and mass analyzers is one of the oldest, but still the most preferred and researched analytical platforms for phytocannabinoid profiling in both plant material and biological matrices due to its robustness, reproducibility, sensitivity and speed [1,42,58,153]. As such, GC methods are officially employed by authorities for phytocannabinoid profiling [1], including the predominant phytocannabinoids (∆ 9 -THC, CBD and CBN) and quantification of ∆ 9 -THC/CBD ratio. This analytical platform is also used for terpene profiling, pesticide screening and residual solvents analysis, which affords potential benefits to regulatory bodies and cannabis industry [39].
By combining short columns, fast oven temperature ramps, high carrier gas linear velocities, narrow columns, hydrogen carrier gas and low film thickness, fast and robust GC methods are generated, appropriate for phytocannabinoid profiling in both research and monitoring purposes [39]. The access to the well-established MSLs, such as the National Institute of Standards and Technology (NIST) Mass Spectral Library and the Wiley Registry Mass Spectral Library eases compound identification through GC-MS analysis. Here, phytocannabinoids identification is performed by comparison of acquired MS or MS/MS spectrum to spectra present MSLs, and further confirmed by analysis of analytical standards. Although the main employment of GC-based platforms is for profiling of terpenes, it has been extensively used for phytocannabinoid profiling as well (Table S1).

• GC Detectors
A large proportion of the profiling studies employ FID [10,17,49,52,55,56,58,70,73-76, 79,82,85,93-96,98,101-103,106,108,115,118,119,138,141,157-159] or dual FID [9,75,121,144]. FID offers more accurate quantitative response with respect to MS, but at the price of lower sensitivity and specificity. The need of analogue analytical standards only makes FID low-cost and simpler analyzer when compared to MS, which needs the use of the corresponding deuterated analytical standards [1]. Moreover, GC-FID methods are more robust than GC-MS methods both in full scan and SIM mode [107]. However, due to its inability to discriminate CBD and CBC, FID is often replaced by MS detectors for more thorough cannabinoid profiling.
GC-MS is the most researched analytical platform for cannabinoid profiling to date [1,42,153,167]. Electron impact ionization (EI) is the most commonly used ionization technique in cannabinoid profiling, while others, i.e., chemical ionization (CI) in positive/negative mode and atmospheric pressure ionization (API) are preferred in forensic analysis of phytocannabinoids in biological samples.
A relatively novel GC-based analytical platform employs analyser based on vacuum UV (VUV) operating in the UV/VUV spectral range (120-240 nm). UV/VUV absorption events are very sensitive for differentiating isomers (positional isomers and diastereomers). This, together with the ability to deconvolute overlapping spectra [4], which significantly shortens analysis time, makes GC-UV/VUV potentially favourized analytical platform over GC-MS. Almost all phytocannabinoids exhibit maximum absorbance in the region 170-200 nm, with no overlap among different phytocannabinoids and significant spectral differentiability. This is especially important for CBC and CBD. However, this method reports high LODs and, thus, cannot be employed for phytocannabinoid profiling in biological matrices and cannabis-based products, but is sufficiently sensitive for plant matrices [161].
Two-dimensional (2D) gas chromatography (GC × GC) is reported to offer better chromatographic resolution of phytocannabinoids [11,56,150,160]. Despite this, it is rarely used for phytocannabinoid profiling; chemical fingerprinting and classification are more commonly performed GC x GC is usually coupled to MS [56,160] or to FID/MS [150]. Combination of columns of different polarity is used, as, for example, 100% dimethylpolysiloxane and polyethylene glycol in sol-gel matrix [56], or a non-polar column (e.g., DB-5 or HP-5MS) for first-dimension separation and medium-polarity column (e.g., DB-17) in the second dimension [150,160].

LC-Based Methods
LC-based methods are recently becoming methods of choice for qualitative and quantitative phytocannabinoid profiling. The simplified sample preparation and the low temperatures, high pressure and high flow rates used during TLC, HPTLC, HPLC and UHPLC, and the recently emerging supercritical fluid chromatography (SFC) analysis allow sample preservation without decarboxylation and decomposition, reliable separation of neutral and acidic phytocannabinoid species and, thus, direct identification and quantification of both neutral and acidic forms of phytocannabinoids in the extracted samples [18,165]. The simplified sample preparation methods, along with avoidance of analytes loss favourized LC over GC in cannabinoid profiling [66].

• TLC and HPTLC Methods
TLC is an attractive method for analyses of herbal drug constituents [168], and especially suitable method for the purpose of preliminary semi-quantitative screening of cannabinoid content in routine tests [164]. It is method of choice for identification of cannabis flowers in all Cannabis flos monographs (DAB, AHP, Ph.Helv., Ph.Eur.) [33][34][35][36] and DAB's Cannabis extractum normatum monograph [38]. Using TLC, cannabinoid identification is performed by comparing retardation factors (R F s) of analytes with R F s of standards on a TLC plate developed with appropriate mobile phase, whereas visual evaluation is obtained by dipping or spraying the TLC plate into/with the appropriate visualization reagent under UV light, or under daylight. In DAB [36] and Ph.Eur. [35] analytical monographs of cannabis flower, a MeOH extract of 0.1 g pulverized drug is identified by comparison of R F s of analytes to reference solutions of CBD and ∆ 9 -THCA (5 mg each) in MeOH. Solutions are applied on TLC C 18 silica gel F 254 plate (2 to 10 µm) and developed with H 2 O/glacial acetic acid/MeOH (15:15:70, v/v/v). After air drying, the TLC plate is sprayed with vanillin reagent, dried at 100-105 • C for 15 min and examined on daylight. Identical procedure applies to Cannabis extractum normatum according to DAB [38]. Ph.Helv. includes filtration of MeOH extract of cannabis flower through membrane filter (0.45 µm) as an additional step prior to application to the C 18 silica gel F 254 plate and reference solution of CBDA in MeCN and ∆ 9 -THCA in 2-isopropanol, following the identical procedure for development and detection [34]. AHP's cannabis flower monograph employs TLC C 18 F 254 plate with MeOH/H 2 O with 1% glacial acetic acid (75:25, v/v) as a mobile phase for identification of CBC, ∆ 9 -THC, CBN, CBG, CBD THCV, ∆ 9 -THCA and CBDA in dichloromethane extract of 0.1 g pulverized drug. Visualization is performed using Fast Blue reagent and vanillin/H 2 SO 4 under UV (254 nm) [33]. UNODC suggests maceration in ultrasound bath with 10 mL of solvent (MeOH, petroleum ether, n-hexane, toluene, CHCl 3 or solvent combinations-MeOH:CHCl 3 (9:1, v/v) for 15 min at room • C, using three systems for elution of HPTLC silica gel plates (A: petroleum ether 60/90/diethyl ether (80:20, v/v); B: cyclohexane/di-isopropyl ether/diethylamine (52:40:8, v/v) and C (for cannabinoid acids): n-hexane:dioxane/MeOH (70:20:10, v/v)). Fast Blue reagent BB or RR in MeOH or MeOH:H 2 O is used as spaying reagent using visualization method 1 or method 2 [5].
The accuracy, repeatability and the acceptable LODs and LOQs in the linear dynamic range of this methodology makes TLC methods attractive for fingerprinting cannabis [4]; however, such parameters are fairly low compared to the more sophisticated LC analytical platforms. The "classic" TLC became further less utilized due to its inconvenience to document for peer review, poor resolution due to systematic errors rising from handspotting, temperature/humidity control and imprecise R F measurement [167,169].
Recent advances in TLC are in the development of HPTLC methods. Such reliable methods could offer advantages over both HPLC and GC techniques for cannabis profiling, including its ability to analyse multiple samples simultaneously and the consequently lower running costs and analysis runtime. Further, automation of sample application in HPTLC methods eliminate systematic errors, provide better resolution and generate reports for more convenient documentation for peer review of casework [170]. In that spirit, normal-phase HPTLC with an automated spotter is shown to achieve better separation than TLC for the main neutral phytocannabinoids. The method is comparable within a small degree of error (±0.5%) to a validated HPLC method [110].
Visualization of (HP)TLC plates is usually performed using 0.1% aqueous solution of Fast Blue B salt reagent [110,[125][126][127]130,147], alone, under white light (254 nm and 366 nm) [127], under UV (254 nm) [131], as solution in 0.1M NaOH [5] or under UV (206 nm) [110] or as 0.5% aqueous solution, followed by 0.1M NaOH [23,125,173]. Prior to Fast Blue B, diethylamine can be applied (50 mg·L −1 H 2 O + 20 mL MeOH) [5]. RP-TLC plates are visualized using Fast Blue B in 0.1M NaOH or in 50 g·L −1 H 2 O/acetone (9:1, v/v) [129]. Fast Blue RR was better for visualization of ∆ 9 -THC, CBD, CBG and CBN than Fast Blue B salt [172]. As qualitative evaluation for the presence of cannabinoids during (HP)TLC analysis is based on color determination, it is often subject of analysts' erroneous determination. Recent studies made the pioneering efforts in developing a method for standardizing and naming colors using the Sci-Chromus ® software, that significantly reduced the subjectivity of the color names in identifying ∆ 9 -THC, CBD, CBN and CBG in cannabis extracts [172].
Apart from TLC and HPTLC, other planar chromatography methods are seldomly used, such as optimum performance laminar chromatography (OPLC) and automated multiple development (AMD) for phytocannabinoid profiling, despite their greater reproducibility due to complete automation. Moreover, OPLC offers extension as semipreparative technique for sample purification, while AMD offers best resolution. The only reported employment of OPLC in AMD in cannabinoid profiling is in hexane extracts of cannabis resins (dried and reconstituted in toluene) and in hexane extracts of cannabis resin. OPLC was performed for determination of ∆ 9 -THC, CBD and CBN on HTSorb BSLA 011 and HT Sorb BSLA 003 columns using isooctane/diethylether (90:10, v/v) as eluent. Semi-preparative OPLC was performed for isolation of CBD from cannabis resin using hexane/diethylether (80:20, v/v). Using AMD, separation was performed on HPTLC with the elution gradient 1C acetone (100, v/v), diisopropylether (100, v/v), hexane (100, v/v), hexane (100, v/v) and hexane (100, v/v) during 20 steps. For both OPLC and AMD, visualisation is performed with Fast Blue B salt reagent [125].

• HPLC Methods
The HPLC technique is gaining popularity as the main choice for fingerprinting study for the quality control of herbal drugs [174], thus enabling chemical characterization of herbal medicines [175]. HPLC methods offer larger linear ranges and more consistent calibration curves for all phytocannabinoids in regards to GC-based methods [163]. In terms of reliability, reproducibility and sensitivity, it was shown that high-resolution GC/FID and HPLC-UV methods for quantification of ∆ 9 -THC, CBD and CBN are comparable [98].
Decrease in buffer concentration from 50 mM to 25 mM eliminated baseline drifting, thus avoiding decrease in UV absorption. As ammonium formate causes co-elution of ∆ 9 -THCA with CBG, it is preferably replaced by ammonium acetate (25 mM, pH 4.75), which offers more reproducible, reliable and rugged chromatographic separation, especially between CBG and ∆ 9 -THCA with improved peak shape and, thus, sensitivity. Better reproducibility is achieved using H 2 O/MeCN (15:85, v/v) + 50 mM phosphoric acid than with 0.1% formic acid as buffering solution [185]. Fast separation of 10 phytocannabi-noids in less than 8.5 min using binary system H 2 O + 0.085% phosphoric acid/MeCN + 0.085% phosphoric acid as mobile phase was achieved, that, together with employment of RP-C 18 column prevented co-elution of CBD and THCV and both isomers, ∆ 9 -THC and ∆ 8 -THC [165]. Another, less frequently used buffer is triethylammonium phosphate (TMAP) in MilliQ, with MeCN in isocratic elution programme [191].
A better chromatographic performance (in terms of both resolution and sensitivity), a shorter analysis time (10 min vs. 12 min) and a considerable saving of solvent consumed while working at a flow rate of 0.5 mL/min instead of 1.5 mL/min, was observed while working on a fuse-core stable bond (SB) RP-C 18 column rather than fully porous RP-C 18 column [182]. Another study evaluated three different columns, RP-C 18 , fused-core stable bond (SB) RP-C 18 and fused-core end-capped (EC) RP-C 18 with numerous mobile phases and gradient conditions in an attempt to shorten the run time and to increase the separation of 8 phytocannabinoids (CBDA, CBGA, CBD, CBG, CBN, ∆ 9 -THCA, ∆ 9 -THC and ∆ 8 -THCA). The SB RP-C 18 core shell column provided the best performance due to significant improvement in separation and symmetry of chromatographic peaks with a baseline separation between CBD and CBG within 20 min shorter run time [176].

• HPLC Detectors
For the purpose of phytocannabinoid profiling, (U)HPLC analytical platforms are coupled to UV, DAD, PAD or MS. Phytocannabinoids have low molar absorptivity, which results in relatively low sensitivity of LC methods employing UV and DAD. This restricts employment of DAD detection to low wavelengths where there is often strong background absorbance from the eluent components, especially during gradient elution experiments [176]. Additionally, UV/DAD methods often have low specificity for some phytocannabinoids, e.g., CBDA and CBGA, due to similar UV/DAD spectra [178].
Phytocannabinoids have different UV behaviour on the basis of their chemical structure. Cannabinoid acids (CBDA and CBGA) are characterized by three absorption maxima (λ max ), one stronger at 220-223 nm, the second at 266-270 nm and the third one around 305 nm, while neutral phytocannabinoids (CBD and CBG) show a first λ max at 210-215 nm and an additional one at 270 nm. Generally, the ranges 190-600 nm and 200-650 nm are most commonly used for UV acquisition, while two wavelengths are selected-210 nm for neutral phytocannabinoids and 220 nm for cananbinolic acids [43,72,185]. For DAD, a narrower range is selected (i.e., 200-400 nm [83], 190-500 nm [182]). Single wavelength can be selected for evaluation of multiple phytocannabinoids in hemp seed oils [179], in plant material (230 nm) [68], 214 nm [188,189] or 220 nm in plant material and resins [69,165] or cannabis extracts [27], in commercial veterinary supplements (225 nm) [192] or for quantification of ∆ 9 -THC, ∆ 9 -THCA, CBN and CBD (210 nm) [191] or ∆ 9 -THC and ∆ 9 -THCA in plant material (211 nm and 220 nm) [185]. For wide range methods, such as cannabisbased medical extracts [182], cannabis-infused cholcolate [65] and 17 phytocannabinoids in cannabis inflorescences and oils [100], 228 nm has been shown to be the most suitable. Multiple detection wavelengths, e.g., 220 nm, 240 nm, 270 nm and 307 nm are also used for phytocannabinoids profiling [111,190]. Evaluation of these detection wavelengths for five phytocannabinoids (CBD, CBDA, CBN, ∆ 9 -THC and ∆ 9 -THCA) in versatile cannabis-based products was performed. While none of the phytocannabinoids showed λ max at 240 nm, this wavelength tended to equalize the response (slope) across the five phytocannabinoids, except for CBN, which has the highest response. The highest response for all five components was found with 220 nm and was useful for low level quantification (Table S2). 270 nm and 307 nm provided high selectivity for CBDA and ∆ 9 -THCA, which was used to minimize or eliminate detection of matrix interferences, as needed. Detection wavelengths in the range of 270-280 nm are inappropriate for CBD, CBN, and ∆ 9 -THC due to retention time (R t ) interferences [90].
Although many studies employ UV/DAD for separation of the major phytocannabinoids (CBDA, CBD, CBN, ∆ 9 -THC and ∆ 9 -THCA), most do not account for interference from minor phytocannabinoids (e.g, CBNA). Such interference is of special importance during profiling of concentrates where minor phytocannabinoids can be enriched to detectable levels. Additionally, some terpenes absorb UV light at the same wavelength as phytocannabinoids. All these issues decrease sensitivity, specificity and selectivity of UV/DAD methods, which are easily overcome by employment of MS detection. Detection of neutral phytocannabinoids based on their ability to absorb fluorescence under the acidic conditions used in RP-LC is only recently reported [111] as a fast, low-cost and selective alternative, but without the ability to detect phytocannabinoid acids and a somewhat narrower linear range than DAD due to saturation at high concentrations.
HPLC-UV high-resolution MS (HRMS) is employed for simultaneous quantification of the two main impurities in "pure" commercial CBG samples (cannabigerovarin (CBGV) and cannabigerobutol (CBGB)) with subsequent confirmation by comparison with synthesized compounds [193]. •

Matrix Effect
The sample, i.e., matrix type greatly affects selection of extraction technique, extraction solvent(s), HPLC column, mobile phase and detection method, which further enhances method sensitivity, selectivity and specificity. All three validation parameters are affected by the presence of matrix constituents that co-extracts with phytocannabinoids, causing signal alteration (suppression or enhancement). In that spirit, matrix effect is frequently reported during LC-MS-based phytocannabinoid profiling. Cannabis plant material is complex matrix with high fat, pigment and polar compounds content, being flavonoids and terpenes most prominent. Cannabis-based products are much more versatile in terms of fat-, sugar-and polar-interferences content, thus being more prominent to expressing significant matrix suppression during instrumental analysis.

SFC Methods
Despite being efficient, SFC is an analytical technique that has still not been fully exploited for the analysis, separation and quantification of cannabis plants and cannabisbased products, compared to GC and LC. The limited number of studies available report SFC as a fast (8-10 min runtime), cost-effective method with high specificity and separation power for phytocannabinoid profiling [195,199,200]. Prior to SFC analysis, derivatization and/or decarboxylation of phytocannabinoids is not required, thus reducing the risk of sample contamination (unlike GC). Additionally, SFC allows separation of the neutral from the acidic phytocannabinoids, simultaneously due to the properties of the supercritical fluids, offers shorter analysis time, better resolution and definitive identification in a single chromatogram of cannabis products when compared to both GC-MS and HPLC methods [199]. UHPSFC is shown to offer greater selectivity than UHPLC, but at the price of lower sensitivity, as a result of considerable variation of the refractive index of CO 2 , resulting in greater baseline noise [188].
Recently, the development of ultra-high performance SFC (UHPSFC) improved resolving power and efficiency, such that SFC has regained its popularity as an alternative phytocannabinoid profiling of cannabis plants and cannabis-based products. SFC combined with a 2 µm particle size column offers rapid separation and when coupled to UV or MS detection, offers highly efficient analysis of the main phytocannabinoids, using inexpensive and environmentally friendly SC-CO 2 as solvent. On the other hand, UHPSFC is considered as highly orthogonal technique which provides different elution order and relative retention of the investigated components compared to UHPLC; therefore, in combination with MS, it could increase the discrimination power of phytocannabinoids in complex matrices.
For the purpose of phytocannabinoid profiling, SFC using cyanopropyl silica packed column was employed [199], while the reported UHPLSFC methods used BEH 2-EP (2ethylpyridine) column [195,200]. Evaluation of columns with different stationary phases, Torus 1-AA (1-aminoanthracene), Viridis BEH-2EP (ethyl-pyridine) and Torus Diol (OH) revealed that the latter column achieved the highest number of (although not completely separated) peaks; thus, it is most appropriate for routine phytocannabinoid profiling [188]. SC-CO 2 is used with MeOH as co-solvent, with a constant (2%) [188] or gradually increasing concentration from 2% to 7% [199], often with the addition of 0.1% formic acid to improve peak shape of cannabinoid acids [188]. UHPSFC methods employ SC-CO 2 with isopropanol/MeCN (80:20, v/v) with 1% H 2 O and linear gradient [195,200]. SFC and UHPSFC are usually coupled either to PDA detectors [188] or to mass analyzers, such as APCI-QQQ [199] and ESI-Q mass analyser [195,200]. The reported linearity, sensitivity and specificity confirm the potential of SFC and UHPSFC to become the main profiling analytical platforms instead of GC-and LC-based methods in near future.

Vibrational Spectroscopy Methods
In the past decade, vibrational spectroscopy techniques (IR, NIR, MIR, FTIR and Raman) emerged as process analytical tool in the pharmaceutical industry for monitoring various quality attributes, and as such were recognized by United States Food and Drug Administration (FDA) [201]. Their ability for high-throughput screening of large volume of sample for a short period of time can exert sampling-based errors and provide rapid, versatile and non-invasive approach in qualitative and quantitative profiling and growth staging of cannabis plant and extracts (Table S3).
The vibrational IR and Raman spectroscopy are considered complementary techniques; even though both are relying on different physical processes, their main observations are focused on light-induced molecular excitation [202]. Vibrational spectroscopy is based on sample's absorption of light at a defined wavelength range, which occurs as a consequence of the vibrational features of the sample that result in the formation of overtones and combination bands that form the spectrum. The relationship between the spectrum and the physicochemical properties of the sample are mathematically modelled using various multivariate regression methods (principal component analysis (PCA) and increment PCA (iPCA) [203], alone [204] or together with partial least-square (PLS) regression analysis [205,206], based on which the compound concentration in the sample is predicted.
NIR spectroscopy, in conjunction with multivariate data analysis, is a widely accepted approach for the abovementioned analysis. Quantitative data generated by NIR strongly agree with UHPLC-UV data, confirming the potential of employment of NIR spectroscopy in routine monitoring of cannabis plant material and cannabis resins [189]. Dispersive NIR and FT-NIR methods were developed for quantification of eight different phytocannabinoids (CBDV, ∆ 9 -THCV, CBD, CBC, ∆ 8 -THC, ∆ 9 -THC, CBG and CBN) in ground leaves and inflorescences from C. sativa [205] and for discriminating illegal and legal cannabis varieties [207]. A similar NIR method demonstrated sensitivity and specificity for CBD quantification in different liquid pharmaceutical products, thus showcasing its potential as a fast method for monitoring of CBD in the production process [206]. Moreover, an NIR method for the growth staging of Cannabis plants was reported as being sensitive to concentrations of phytocannabinoids and volatile substances in the samples, which are also correlated to the plant age, thus justifying the feasibility of the method for growth staging of cannabis [203]. Dispersive NIR using a scanning monochromator [205], FT-NIR spectrometer with integrated Michelson interferometer and a highly sensitive PbS detector [205], FT-NIR spectrometer based on measurement by diffuse reflectance [206], NIR spectrometer with In-Ga-As detector [203], and a handheld NIR with In-Ga-As array detector [189] were so far employed in phytocannabinoid profiling in cannabis ground leaves and inflorescences [205], in cannabis seeds [203] and in liquid pharmaceutical products (medium-chain triglyceride and propylene glycol-based formulations) [206]. Two handheld NIR devices (NIR-S-GI and MicroNIR) are used for in-field determination of ∆ 9 -THC content in cannabis inflorescences and cannabis resins. For this purpose, spectrophotometers with a larger sample analysis window are more appropriate for highly heterogenous samples, such as whole cannabis inflorescences [189].
The literature data regarding phytocanabinnoids' structural and molecular analysis, to date, is very scarce, lacking band assignation and in-depth structural analysis of the molecules [23,208,209] despite the documented ability to provide chemical fingerprinting and qualitative profiling of phytocannabinoids, especially of FTIR in quantification of biological compounds in complex matrices [210]. However, the interest in qualitative profiling of cannabis (identification/classification) and quantitative profiling of the main phytocannabinoids is slowly gaining momentum in the last few years (Supporting information, Table S3). A recent study for ATR mid-IR quantification of ∆ 9 -THC and CBD in cannabis flowers and extracts described a stepwise approach in developing multivariate quantification models accompanied by detailed band assignment of the mentioned phytocannabinoids, both for pure compounds and analysed samples (complex matrices) [21]. In a further study by the same group, the potential of ATR mid-IR as a process analytical tool (PAT) for continuous monitoring of ∆ 9 -THCA decarboxylation was showcased [22].
The new portable vibrational spectroscopy apparatus versions are very applicable, especially for continuous monitoring of the main phytocannabinoids in all growth stages of cannabis and cannabis-based products manufacturing. Thus, a handheld Raman spectrometer in conjunction with orthogonal PLS-DA was utilized to construct a classification model for discriminating ∆ 9 -THCA rich, CBD rich Cannabis plants and hemp (low ∆ 9 -THCA, CBDA and CBD) [211,212]. Reference standards from the main phytocannabinoids were used to perform a detailed band assignation of the spectra that were further correlated with the loading plot of the multivariate models. In both cases, favourable accuracy descriptors for the classification models were reported, thus showcasing the applicability of the portable Raman device for accurate and fast Cannabis plant classification. Raman spectroscopy is a highly sensitive analytical technique, that, due to the variety of monochromatic light sources, and the emergence of surface-enhanced Raman scattering (SERS), stimulated resonance and coherent anti-Stokes Raman scattering (CARS) offers greatly enhanced capability and resolution, especially in its imaging mode. CARS imaging at different Raman vibrations, known as hyperspectral CARS imaging, is a spectroscopic imaging technique with high-resolution capabilities for a chemical distinction that employs sophisticated data processing methods [208]. This analytical method was used to analyze the secondary metabolites (∆ 9 -THCA and CBDA) in glandular cannabis trichomes with distinct spatial resolution, without the need to extract the resin [213]. To get additional morphological data, the authors superimposed the image over a single photon fluorescence and SEM image of the trichomes. The similarity of the chemical fingerprints of the distinct regions with the secondary metabolites was determined with hierarchical clustering analysis (HCA). The proposed methodology enables an easy discrimination between trichomes with high-∆ 9 -THCA and high-CBDA content.

Other Analytical Techniques
• CE CE is analytical platform used in cannabinoid profiling in only one study [68]. MeCNbased background electrolyte (with 6.5 mM NaOH) in the presence of β-cyclodextrin (βCD), improving orthogonal separation media by transiently interacting with compounds based on their geometry and polarity, was used to separate 14 phytocannabinoids (CBG, CBGA, CBD, CBDA, CBN, ∆ 9 -THC, CBC, CBCA, ∆ 9 -THCA, THCV, CBDV and CBGVA) by constant transition between the background electrolyte and βCD. CE performed better than HPLC-DAD in terms of selectivity and runtime, but with significantly lower sensitivity.
CE's variety, capillary electromatography (CEC), coupling the benefits of CE and HPLC methods, has been used in phytocannabinoid profiling coupled with UV PDA detector in only one study [214]. Baseline separation of seven phytocannabinoids (CBG, CBD, CBN, ∆ 9 -THC, ∆ 8 -THC, CBC, ∆ 9 -THCA) was achieved by using C 18 column and MeCN/25 mM phosphate buffer (75:25, v/v) as mobile phase for analysis of MeOH/CHCl 3 (9:1, v/v) cannabis plant material extracts, while improved sensitivity is achieved using UV cell with extended path length and injection size.
• NMR Spectroscopy NMR-based methods are superior for the purpose of 3D-structure elucidation, especially ζ-resolving, 1 H-1 H COSY and 1 H-13 C heteronuclear multiple quantum coherence (HMQC) and heteronuclear multiple bond correlation (HMBC) spectroscopy. The complete 1 H-and 13 C-NMR assignments of the major Cannabis constituents, ∆ 9 -THC, ∆ 9 -THCA, ∆ 8 -THC, CBG, CBN, CBD, CBDA, cannflavin A and cannflavin B have been determined on the basis of one-and two-dimensional NMR spectra, including 1 H-and 13 C-NMR, 1 H-1 H-COSY, HMQC and HMBC [20]. However, they are seldomly used for quantification purposes, due to the laborious separation and isolation steps required, where significant loss of mass can occur and where there is low sensitivity [4,215]. Additional issues are the high instrumental costs and necessity of highly specialized personnel. Despite that, NMR is considered as a highly accurate, reproducible and fast technique [1].
For the purpose of phytocannabinoid profiling, NMR spectroscopy is used as (semi)quantitative method alone [20,209] or as an orthogonal technique to LC [216][217][218] or GC [219] for the purpose of qualitative peak assignment of major phytocannabinoids [20], chemical and morphological examination [220], chemotaxonomic classification [20,219,220], metabolomics-based chemovar distinction [51] or quantitative analysis of cannabis plant material without the need of pre-purification step [221], chromatographic separation or use of certified reference standards [219]. Cryogenic NMR spectroscopy combines improved sensitivity and noise reduction with a cryogenic cooling system for the receiver coil and preamplifiers. Its improved spectral quality is employed in compound identification from mass limited samples and as orthogonal analytical technique to HPLC in phytocannabinoid profiling in laser-micro dissected samples of capitate-stalked and capitate-sessile trichomes [220].

Conclusions and Future Directions
Scientific and technological advancements in cultivation, manufacturing, recreational, industrial and medical use of cannabis, as well as updated legislation, led to the development of multitude of analytical methods for phytocannabinoid profiling. Matrix nature greatly affects selection of extraction technique, sample preparation and analytical method due to the fact that significant matrix interferences can occur and aggravate the overall analysis of target phytocannabinoids. Sample preparation for phytocannabinoid profiling in the past four decades is mainly based on versatile types of accelerated maceration, such as SLE, LLE, PLE, SPE, USE, FUSE and MAHD. Recently, new trends have enlightened environmental-friendly techniques, such as easily-automatable HS-SPME and SFE, which adds speed, repeatability and reproducibility to the analyses. From the multitude of analytical platforms, TLC and HPTLC, HPLC-DAD, GC and LC coupled with mass spectrometry (MS or MS/MS), are most commonly used; however, recently emerging techniques are NMR and vibrational spectroscopy methods, such as IR, NIR, FTIR and FT-NIR. TLC, together with HPTLC, which is a suitable method for screening of samples and is included in the pharmacopoeias in the identification methods. Cannabinoid profiling for research, industrial and QC purposes is based mostly on two analytical platforms: GC and LC. GC coupled to versatile detectors and mass analyzers is one of the oldest, the most preferred and researched analytical platforms for phytocannabinoid profiling due to its robustness, reproducibility, sensitivity and speed.
As a result of the advancement of computational tools, mass spectral libraries (MSLs), public compound repositories and compound databases (DBs), as well as various advanced detection techniques, GC has become the analytical platform in forensic, pharmacokinetic and phytochemical analysis of natural phytocannabinoids. As such, GC methods are also officially employed by authorities for terpene profiling, pesticide screening and residual solvents analysis, which affords potential benefits to regulatory bodies and cannabis industry. High pressure and high flow rates used during TLC, HPTLC, HPLC and UHPLC, and the recently emerging SFC technique allow sample preservation without decarboxylation and decomposition, reliable separation of neutral and acidic cannabinoid species and, thus, direct identification and quantification of both neutral and acidic forms of phytocannabinoids in the extracted samples. In the past decades, LC has become an analytical platform of choice (HPLC-DAD and LC-MS) in first line for potency studies but also for untargeted analysis of cannabis and cannabis-based products. Despite the fact that vibrational spectroscopy methods, such as NIR, FTIR, FT-NIR and Raman spectroscopy are reserved for structural elucidation, in the last few years there is an evident trend of their utilization for rapid quantitative phytocannabinoid profiling. Although these vibrational spectroscopy techniques can provide rapid, versatile and non-invasive approach in qualitative and quantitative profiling and growth staging of cannabis plant and extracts, they demonstrate far higher LOD and LOQ than the described chromatography-'wet' methods, they are fast, inexpensive, non-destructive and require minimum (e.g., drying, grinding) or no sample preparation. Regardless of high instrumental costs and necessity of highly specialized personnel, NMR is considered as a highly accurate, reproducible and fast technique that offers quantitative analysis of cannabis without the need of pre-purification step, chromatographic separation or use of certified reference standards. Even though there are currently various well-established methods available for chemical analyses of phytocannabinoids, there is still a need for adaptations and enhancement of these methods in the light of new scientific evidence regarding the plant and its plant metabolites, especially taking into account the pharmacological activity and its medical use, association of medical and adverse health effects with potency and/or interplay of certain phytocannabinoids and other active constituents, quality control and stability studies of cannabis and cannabis-based products. Further advancements in phytocannabinoid profiling should move towards untargeted analysis of cannabis plant material and cannabis-based products using orthogonal analytical methods. By employment of cheminformatics approaches for small molecule identification and MSLs, a multitude of new phytocannabinoids and other compounds is expected to be identified in the near future, thus allowing access to complete and accurate phytocannabinoid and terpene profiles.
Funding: This research received no external funding.

Conflicts of Interest:
The authors declare no conflict of interest.
Sample Availability: Samples of the compounds are available from the authors.