Cyanogenesis in Macadamia and Direct Analysis of Hydrogen Cyanide in Macadamia Flowers, Leaves, Husks, and Nuts Using Selected Ion Flow Tube–Mass Spectrometry

Macadamia has increasing commercial importance in the food, cosmetics, and pharmaceutical industries. However, the toxic compound hydrogen cyanide (HCN) released from the hydrolysis of cyanogenic compounds in Macadamia causes a safety risk. In this study, optimum conditions for the maximum release of HCN from Macadamia were evaluated. Direct headspace analysis of HCN above Macadamia plant parts (flower, leaves, nuts, and husks) was carried out using selected ion flow tube–mass spectrometry (SIFT-MS). The cyanogenic glycoside dhurrin and total cyanide in the extracts were analyzed using HPLC-MS and UV–vis spectrophotometer, respectively. HCN released in the headspace was at a maximum when Macadamia samples were treated with pH 7 buffer solution and heated at 50 °C for 60 min. Correspondingly, treatment of Macadamia samples under these conditions resulted in 93–100% removal of dhurrin and 81–91% removal of total cyanide in the sample extracts. Hydrolysis of cyanogenic glucosides followed a first-order reaction with respect to HCN production where cyanogenesis is principally induced by pH changes initiating enzymatic hydrolysis rather than thermally induced reactions. The effective processing of different Macadamia plant parts is important and beneficial for the safe production and utilization of Macadamia-based products.


Introduction
Macadamia-based commercial products have rapidly increased in recent years. In addition to Macadamia nuts, Macadamia flowers, husks, leaves, and shells are now widely used as a source of functional foods, beverages, and raw materials in cosmetics, feed, and other applications. Abundant antioxidant substances, such as polyphenols, can be extracted from Macadamia skin and husks for utilization in the food and pharmaceutical industries [1][2][3]. Bioactive constituents in Macadamia are believed to provide health benefits such as improved blood lipid profiles, decreased inflammation, oxidative stress, and reduced cardiovascular disease risk factors [4,5].
Acute and chronic toxicities of hydrogen cyanide from plant-derived food have been reported [16,19,20]. Ingestion of 0.5-3.5 mg cyanide/kg body weight results in acute toxicity. Sublethal doses could lead to headache, hyperventilation, vomiting, weakness, abdominal cramps, and partial circulatory and respiratory systems failure. Moreover, cyanide can inhibit cellular respiration, which could result in fatal poisoning [7,16,21]. The concentration of cyanogenic glycosides, such as dhurrin and proteacin, varies among plant species of Macadamia (i.e., M. ternifolia, M. integrifolia, and M. tetraphylla) [6,22]. These compounds are also unevenly distributed within the different parts of a plant (i.e., nuts, seeds, and roots), and their concentrations change at different developmental stages from seed germination to plant maturation [23,24].
In this study, different conditions causing the hydrolysis of cyanogenic compounds in Macadamia that subsequently produce hydrogen cyanide gas were evaluated. Most characterization studies in Macadamia and other plants only involved analysis of cyanogenic glycosides using timeconsuming assays coupled with HPLC analysis [6,25,26]. Furthermore, typical analysis of releasable cyanide uses tedious assays and subsequent spectrophotometry or LC-or GC-MS analysis [14,27,28]. In this study, hydrogen cyanide was measured directly above the headspace of the different parts of the Macadamia plant including the flowers, leaves, husks, and nuts using selected ion flow tube-mass spectrometry (SIFT-MS). To our knowledge, this is the first study to measure hydrogen cyanide in real time and directly above the headspace of Macadamia samples using SIFT-MS. The rapid and realtime analysis of hydrogen cyanide is particularly important in the processing of the various parts of Macadamia that are known to contain cyanogenic glycosides and can subsequently hydrolyze and undergo cyanogenesis. The optimum conditions (heating temperature, heating time, and pH) for the hydrolysis of cyanogenic glycosides via cyanogenesis toward the maximum generation of hydrogen cyanide were determined. Identifying these conditions would be useful in the pre-processing of Macadamia to ensure maximum hydrolysis of cyanogenic glucoside, leading to maximum release and  [17,18].
Acute and chronic toxicities of hydrogen cyanide from plant-derived food have been reported [16,19,20]. Ingestion of 0.5-3.5 mg cyanide/kg body weight results in acute toxicity. Sublethal doses could lead to headache, hyperventilation, vomiting, weakness, abdominal cramps, and partial circulatory and respiratory systems failure. Moreover, cyanide can inhibit cellular respiration, which could result in fatal poisoning [7,16,21]. The concentration of cyanogenic glycosides, such as dhurrin and proteacin, varies among plant species of Macadamia (i.e., M. ternifolia, M. integrifolia, and M. tetraphylla) [6,22]. These compounds are also unevenly distributed within the different parts of a plant (i.e., nuts, seeds, and roots), and their concentrations change at different developmental stages from seed germination to plant maturation [23,24].
In this study, different conditions causing the hydrolysis of cyanogenic compounds in Macadamia that subsequently produce hydrogen cyanide gas were evaluated. Most characterization studies in Macadamia and other plants only involved analysis of cyanogenic glycosides using time-consuming assays coupled with HPLC analysis [6,25,26]. Furthermore, typical analysis of releasable cyanide uses tedious assays and subsequent spectrophotometry or LC-or GC-MS analysis [14,27,28]. In this study, hydrogen cyanide was measured directly above the headspace of the different parts of the Macadamia plant including the flowers, leaves, husks, and nuts using selected ion flow tube-mass spectrometry (SIFT-MS). To our knowledge, this is the first study to measure hydrogen cyanide in real time and directly above the headspace of Macadamia samples using SIFT-MS. The rapid and real-time analysis of hydrogen cyanide is particularly important in the processing of the various parts of Macadamia that are known to contain cyanogenic glycosides and can subsequently hydrolyze and undergo cyanogenesis. The optimum conditions (heating temperature, heating time, and pH) for the hydrolysis of cyanogenic glycosides via cyanogenesis toward the maximum generation of hydrogen cyanide were determined. Identifying these conditions would be useful in the pre-processing of Macadamia to ensure maximum hydrolysis of cyanogenic glucoside, leading to maximum release and volatilization of hydrogen cyanide and ultimately toward the safe production and utilization of Macadamia-based products especially as ingredients in food and beverages.

Sample Preparation
Macadamia (M. integrifolia) flowers, leaves, nuts, and three variety of husks (A16, Oc, and 695) were donated by Shouxiang Township Organic Agricultural Products Development Co., Ltd. (Guangxi, China). The three varieties of M. integrifolia husks were introduced and propagated in China from Australia and are hybrid cultivars selected from different plantations or open-pollinated progeny (variety A16). HPLC-grade water, hexane, Na 2 HPO 4 , and citric acid were purchased from Fisher Scientific (Fisher Chemical, Fair Lawn, NJ, USA).
Macadamia flowers and leaves were freeze-dried, ground, and sifted. Macadamia husks were air-dried, crushed, and sifted. Macadamia nuts were crushed, defatted using hexane, and air-dried. All samples were stored in sealed bottles under freezing temperature (−20 • C).

Buffer Preparation
Na 2 HPO 4 solution (0.2 mol/L) was prepared by dissolving 14.2 g Na 2 HPO 4 with 500 mL carbon dioxide-free HPLC water. Citric acid solution (0.1 mol/L) was prepared by dissolving 10.51 g citric acid with 500 mL HPLC water.
Different volumes of 0.2 mol/L Na 2 HPO 4 solution and 0.1 mol/L citric acid were mixed to prepare various buffer solutions with pH 2, 3, 4, 5, 6, 7, 8, and 9. The pH of each solution or sample mixture was measured using a Model 10 pH meter (Denver Instrument Company, Arvada, CO, USA).

Optimization of Heating Temperature and Heating Time
Macadamia samples (0.100 g) were subjected to different heating times and temperatures to evaluate the optimum conditions for the maximum hydrolysis of cyanogenic compounds and maximum production of hydrogen cyanide. Samples were heated at 30, 40, 50, 60, 70, or 100 • C. At each temperature, samples were heated for 20, 30, 45, 60, 80, 100, or 120 min. Immediately after heating, the headspace concentration of hydrogen cyanide was measured using SIFT-MS.

Optimization of pH-Buffering Solution
To evaluate the optimum pH for maximum enzymatic activity and the hydrolysis reaction, 0.100 g powdered Macadamia flower sample was dissolved in 0.75 mL Na 2 HPO 4 -citric acid buffered solutions with different pH (2, 3, 4, 5, 6, 7, 8, or 9). The solutions were heated at 50 • C for 15, 30, 60, 90, and 120 min. The headspace concentration of hydrogen cyanide was immediately measured using SIFT-MS.

Headspace Cyanide Analysis Using SIFT-MS
Headspace hydrogen cyanide (HCN) was analyzed using a V200 selected ion flow tube-mass spectrometry, SIFT-MS (Syft TM Technologies, Middleton, Christchurch, New Zealand). Using selected ion scan mode, HCN was measured using the H 3 O + precursor ion to detect a protonated HCNH + at m/z 28 with a reaction rate, k, of 3.8 × 10 −9 cm 3 s −1 . SIFT-MS has recently been used for the headspace analysis of various compounds in different food (oil, cheese, and garlic) and breath matrices [29][30][31][32][33][34]. For the headspace detection of HCN using SIFT-MS, 0.100 g of Macadamia flowers, leaves, husks, or defatted nuts sample was weighed into individual 500 mL Schott bottles. Then, 0.75 mL HPLC water or Na 2 HPO 4 -citric acid buffer was added, and the solution was mixed and heated (50 • C) in a water bath (Precision Inc., Winchester, VA, USA).
A stock cyanide standard solution (1002 ± 5 mg/L KCN in 0.1% NaOH, Specpure, Alfa Aesar, Tewksbury, MA, USA) was used to prepare the working standard aqueous solutions (0, 20, 40, 80, 160, 200, 425, and 1000 µg/L). After this, 1 mL of the working standard solution or matrix blank (HPLC water) was transferred to a 100 mL Schott bottle sealed with a septum-lined screw cap. The working standards were heated at 50 • C for 30 min to allow for headspace equilibrium prior to SIFT-MS analysis. Figure 2A shows the concentration of cyanide in the headspace (ppb v ) as a function of cyanide concentration in aqueous solution (µg/L) generated by a linear regression model. The correlation coefficient (R 2 ) for the calibration curve was 0.9993 which signifies that the linear regression model fits the data having <0.0001 significance probability associated with the F statistic (Pr > F) at 95% confidence intervals.
MS analysis. Figure 2A shows the concentration of cyanide in the headspace (ppbv) as a function of cyanide concentration in aqueous solution (µg/L) generated by a linear regression model. The correlation coefficient (R 2 ) for the calibration curve was 0.9993 which signifies that the linear regression model fits the data having <0.0001 significance probability associated with the F statistic (Pr > F) at 95% confidence intervals.
Immediately after achieving headspace equilibrium by heating, headspace sampling was carried out by inserting a passivated sampling needle (~3.5 cm) through the bottle's septum. The sample inlet flow rate was optimized to 0.35 ± 0.01 Torr·L s −1 (26 ± 1 cm 3 min −1 under standard ambient temperature (298 K). The scan duration was 120 s. HPLC water or Na2HPO4-citric acid buffer was used as a blank solution. Lab air was analyzed in between samples to minimize carry-over effects and potential crosscontamination. Five replicates were performed in all analyses.

Dhurrin Analysis in Plant Extracts Using HPLC
Dhurrin extraction and analysis were performed based on the procedure by De Nicola and coworkers [35]. Briefly, 0.2 g of freeze-dried, powdered plant sample was weighed into a 25 mL centrifuge tube and 0.1 g of activated carbon (Fisher Chemical, Fair Lawn, NJ, USA) and 10 mL methanol (Fisher Chemical, Fair Lawn, NJ, USA) were added. The mixture was sonicated for 25 min at room temperature in a 435 W ultrasonic water bath (Model FS28H, Fisher Scientific, Fair Lawn, NJ, USA) and was left overnight in the tube. After 12-14 h, the mixture was centrifuged (Model Sorvall Legend XFR Centrifuge, Thermo Fisher Scientific, Waltham, MA, USA) for 30 min at 17,000× g and 10 °C and was filtered through a Whatman no. 4 filter paper (GE Healthcare, Buckinghamshire, UK). Immediately after achieving headspace equilibrium by heating, headspace sampling was carried out by inserting a passivated sampling needle (~3.5 cm) through the bottle's septum. The sample inlet flow rate was optimized to 0.35 ± 0.01 Torr·L s −1 (26 ± 1 cm 3 min −1 under standard ambient temperature (298 K). The scan duration was 120 s. HPLC water or Na 2 HPO 4 -citric acid buffer was used as a blank solution. Lab air was analyzed in between samples to minimize carry-over effects and potential cross-contamination. Five replicates were performed in all analyses.

Dhurrin Analysis in Plant Extracts Using HPLC
Dhurrin extraction and analysis were performed based on the procedure by De Nicola and co-workers [35]. Briefly, 0.2 g of freeze-dried, powdered plant sample was weighed into a 25 mL centrifuge tube and 0.1 g of activated carbon (Fisher Chemical, Fair Lawn, NJ, USA) and 10 mL methanol (Fisher Chemical, Fair Lawn, NJ, USA) were added. The mixture was sonicated for 25 min at room temperature in a 435 W ultrasonic water bath (Model FS28H, Fisher Scientific, Fair Lawn, NJ, USA) and was left overnight in the tube. After 12-14 h, the mixture was centrifuged (Model Sorvall Legend XFR Centrifuge, Thermo Fisher Scientific, Waltham, MA, USA) for 30 min at 17,000× g and 10 • C and was filtered through a Whatman no. 4 filter paper (GE Healthcare, Buckinghamshire, UK). The supernatant was collected and 1:1 (v/v) HPLC-grade water was added to the resulting solution. Prior to HPLC analysis, the diluted supernatant solution was filtered through a 0.2 µm RC membrane filter (Phenomenex, Torrance, CA, USA) using a luer-type syringe (Henke-SASS Wolf GmbH, Tuttlingen, Germany) and was transferred into 1.5 mL amber vials for HPLC analysis.
Dhurrin stock standard solution was prepared by dissolving 1 mg of pure dhurrin standard (Sigma Aldrich, St. Louis, MO, USA) with 1 mL of HPLC-grade water. Working standard solutions (0, 5, 10, 25, 50, and 100 mg dhurrin/L solution) were prepared using aliquots of the stock standard solution and diluted with 1:1 H 2 O/methanol (v/v) solution. Dhurrin standard solutions were transferred to 1.5 mL amber vials, correspondingly, for HPLC analysis. A solution of 1:1 H 2 O/methanol (v/v) was used as matrix blank. Figure 2B shows the peak area of dhurrin as a function of dhurrin concentration in aqueous solution (mg/L) generated by the linear regression model. The correlation coefficient (R 2 ) for the calibration curve was 0.9999 which signifies that the linear regression model fits the data having a 0.0037 significance probability associated with the F statistic (Pr > F) at 95% confidence intervals.
Analysis of dhurrin from the sample extracts and standards were carried out using an HPLC (1100 Series, Agilent Technologies, Santa Clara, CA, USA) equipped with a G1311A quaternary pump, a G1322A degasser, a G1313 ALS autosampler, and a G1316A thermostated column compartment with a C-18 column. The chromatographic conditions involved a flow rate of 1 mL/min by eluting with a gradient of water (A) and acetonitrile (B). The gradient program was set as follows: isocratic 10% B for 1 min, linear gradient to 30% B for 7 min, and linear gradient to 10% B for 2 min. Dhurrin was detected using a G1315B diode array detector (DAD) detector (Agilent Technologies, Santa Clara, CA, USA), and its absorbance was monitored at 232 nm. Dhurrin's spectral peak was identified by comparing the retention time to that of pure dhurrin from the standard solutions. The resulting chromatograms ( Figure 3) were automatically integrated using ChemStation software (Agilent Technologies Inc., Santa Clara, CA, USA). Five replicates per standard or sample extracts were performed in all analyses. The supernatant was collected and 1:1 (v/v) HPLC-grade water was added to the resulting solution.
Prior to HPLC analysis, the diluted supernatant solution was filtered through a 0.2 µm RC membrane filter (Phenomenex, Torrance, CA, USA) using a luer-type syringe (Henke-SASS Wolf GmbH, Tuttlingen, Germany) and was transferred into 1.5 mL amber vials for HPLC analysis. Dhurrin stock standard solution was prepared by dissolving 1 mg of pure dhurrin standard (Sigma Aldrich, St. Louis, MO, USA) with 1 mL of HPLC-grade water. Working standard solutions (0, 5, 10, 25, 50, and 100 mg dhurrin/L solution) were prepared using aliquots of the stock standard solution and diluted with 1:1 H2O/methanol (v/v) solution. Dhurrin standard solutions were transferred to 1.5 mL amber vials, correspondingly, for HPLC analysis. A solution of 1:1 H2O/methanol (v/v) was used as matrix blank. Figure 2B shows the peak area of dhurrin as a function of dhurrin concentration in aqueous solution (mg/L) generated by the linear regression model. The correlation coefficient (R 2 ) for the calibration curve was 0.9999 which signifies that the linear regression model fits the data having a 0.0037 significance probability associated with the F statistic (Pr > F) at 95% confidence intervals.
Analysis of dhurrin from the sample extracts and standards were carried out using an HPLC (1100 Series, Agilent Technologies, Santa Clara, CA, USA) equipped with a G1311A quaternary pump, a G1322A degasser, a G1313 ALS autosampler, and a G1316A thermostated column compartment with a C-18 column. The chromatographic conditions involved a flow rate of 1 mL/min by eluting with a gradient of water (A) and acetonitrile (B). The gradient program was set as follows: isocratic 10% B for 1 min, linear gradient to 30% B for 7 min, and linear gradient to 10% B for 2 min. Dhurrin was detected using a G1315B diode array detector (DAD) detector (Agilent Technologies, Santa Clara, CA, USA), and its absorbance was monitored at 232 nm. Dhurrin's spectral peak was identified by comparing the retention time to that of pure dhurrin from the standard solutions. The resulting chromatograms ( Figure 3) were automatically integrated using ChemStation software (Agilent Technologies Inc., Santa Clara, CA, USA). Five replicates per standard or sample extracts were performed in all analyses.

Total Cyanide Analysis in Plant Extracts Using UV-Vis Spectrophotometer
The alkaline picrate method was used for the extraction and analysis of total cyanide as outlined by Sarkiyayi and Agar [36] and Omar and co-workers [37]. Five grams (5 g) dried samples and 50 mL HPLC water were placed in a conical flask which was soaked overnight and then filtered using Whatman no. 4 filter paper. One mL of the filtrate was transferred to a test tube, and 4 mL alkaline picric acid solution was added. The mixture was incubated for 5 min in a 95 • C H 2 O bath. After color development, the absorbance of the mixture was measured at 490 nm using a Varian UV-vis spectrophotometer (Agilent, Cary 50 Bio UV/Visible, Santa Clara, CA, USA). Alkaline picric acid solution was prepared by mixing 1 g picric acid (2,4,6-trinitrophenol crystal, Electron Microscopy Sciences, Hatfield, PA, USA), 5 g Na 2 CO 3 (Fisher Scientific, Fair Lawn, NJ, USA), and 200 mL HPLC water.
A stock cyanide standard solution (1002 ± 5 mg/L KCN in 0.1% NaOH) was used to prepare the working standard aqueous solutions (0-20 mg/L). One milliliter of the working standard solution or matrix blank (HPLC water) was transferred to a test tube. Four milliliters of alkaline picric acid solution was added, and the mixture was incubated for 5 min in a 95 • C H 2 O bath for color development. The solution absorbance was measured at 490 nm using a UV-vis spectrophotometer. Figure 2C shows the spectral absorbance as a function of the cyanide concentration in aqueous solution (mg/L) generated by the linear regression model. The correlation coefficient (R 2 ) for the calibration curve was 0.9966 which signifies that the linear regression model fits the data having <0.0001 significance probability associated with the F statistic (Pr > F) at 95% confidence intervals. For analysis, 20 replicates per standard and 10 replicates per sample extract were used for UV-vis measurement.

Statistical Analysis
Data fitting, analysis of least square means, and regression analysis of the headspace hydrogen cyanide concentrations were carried out using the PROC REG and PROC MIXED options of Statistical Analysis System (SAS ® Institute Inc., Cary, NC, USA). Analysis of variance (ANOVA) was performed to analyze the statistical differences of cyanide concentration between different samples using least significant difference of the means (LSD) technique using SAS. Significance was defined using p < 0.05 (95% confidence intervals) for least square means comparison. Five replicates were performed in all analyses, except where otherwise specified. Limit of blank (LOB) and limit and detection (LOD) were determined using the methods described by Browne and Whitcomb [38] and Shrivastava and Gupta [39]. The estimated headspace LOB (PHBA) and LOD (PHBA) for HCN using SIFT-MS were 2.462 ppb v and 2.775 ppb v , respectively, which were determined using repeated headspace measurements of blank (n = 60) heated in a water bath at 90 • C for 50 min. Figure 4A shows the concentration of hydrogen cyanide (HCN) from the headspace of Macadamia flower samples heated at 30, 40, 50, 60, 70, and 100 • C for 20, 30, 45, 60, and 80 min. For all heating times, the headspace HCN concentration increased from 30 to 50 • C and decreased linearly beyond 50 • C. Thus, HCN generation in Macadamia was maximum at 50 • C. These results suggest that the optimum temperature for enzymatic activity of endogenous dhurrinase and α-hydroxynitrile lyase is 50 • C (Figure 1). At higher heating temperatures (i.e., above 50 • C), cyanide production decreased, which could be caused by decreased enzyme activity or inactivation and, therefore, reduced subsequent hydrolysis reaction of the main cyanogenic glycoside dhurrin ( Figure 4A). It is interesting to note that the decreasing production of cyanide at higher temperatures (60-100 • C) is gradual rather than an abrupt cyanide reduction that could be expected from thermally induced enzyme inactivation. The measured cyanide at high temperature could be produced from other thermolabile cyanogenic glycosides that are present in minor amounts. At high temperature, the isomers of dhurrin such as taxiphyllin, zierin, and p-glucosyloxy-mandelonitrile can readily dissociate and release cyanide without enzymatic hydrolysis [40][41][42]. Therefore, the cyanide released from the thermally induced decomposition of these minor cyanogenic glycosides could be contributing to the detected cyanide in the headspace of Macadamia flower samples heated at higher temperatures. isomers of dhurrin such as taxiphyllin, zierin, and p-glucosyloxy-mandelonitrile can readily dissociate and release cyanide without enzymatic hydrolysis [40][41][42]. Therefore, the cyanide released from the thermally induced decomposition of these minor cyanogenic glycosides could be contributing to the detected cyanide in the headspace of Macadamia flower samples heated at higher temperatures. In addition, the longer the heating time, the higher the HCN concentration, with the longest heating times (60 and 80 min) generating the highest HCN concentration ( Figure 4A). The maximum HCN concentration was reached when samples were heated at 40-50 °C for 80 min or 50 °C for 60 min. From these results, the optimum heating time and temperature were determined to be 50 °C for 60 min, which were used for succeeding experiments. Figure 4B shows the headspace concentration of hydrogen cyanide above Macadamia flower samples treated with Na2HPO4-citric acid buffered solutions at different pH (2,3,4,5,6,7,8,9) and heated at 50 °C for 15, 30, 60, and 90 min. As the pH increased from pH 2 to 7, the concentration of HCN increased. From pH 7 to 9, the concentration of HCN decreased slightly. The Macadamia flower sample treated with pH 7 buffer and heated at 50 °C for 60 min generated the highest headspace In addition, the longer the heating time, the higher the HCN concentration, with the longest heating times (60 and 80 min) generating the highest HCN concentration ( Figure 4A). The maximum HCN concentration was reached when samples were heated at 40-50 • C for 80 min or 50 • C for 60 min. From these results, the optimum heating time and temperature were determined to be 50 • C for 60 min, which were used for succeeding experiments. Figure 4B shows the headspace concentration of hydrogen cyanide above Macadamia flower samples treated with Na 2 HPO 4 -citric acid buffered solutions at different pH (2,3,4,5,6,7,8,9) and heated at 50 • C for 15, 30, 60, and 90 min. As the pH increased from pH 2 to 7, the concentration of HCN increased. From pH 7 to 9, the concentration of HCN decreased slightly. The Macadamia flower sample treated with pH 7 buffer and heated at 50 • C for 60 min generated the highest headspace concentration of HCN. This result suggests that these conditions are optimum for the underlying enzymatic activities involved in the hydrolysis reaction of cyanogenic glycoside compounds producing hydrogen cyanide gas (Figure 1).

Optimization of Mixture's pH for Maximum Generation of Hydrogen Cyanide
When Macadamia flower was heated at 50 • C for 60 min under its normal physiological pH (pH 4.35), the HCN level was only about 4900-5600 ppb v (Figure 4A,B). Increasing the treatment's pH to pH 7, significantly increased the headspace HCN concentration by 200-250% (~12,500 ppb v ). At a more basic pH (pH 8 or 9), HCN concentration was still significantly higher than the concentration at acidic pH (pH 6 and below), but it was lower than that at pH 7.
Further analysis of data was done by plotting the hydrogen cyanide concentration as a function of pH ( Figure 4C), the hydrolysis reaction of cyanogenic glycosides at 50 • C could be described as a first-order reaction with respect to the production of hydrogen cyanide. A constant pseudo first-order rate value (k = 0.0081 ± 0.0007 M min −1 ) was determined from the linear regression slopes of ln [HCN], mol L −1 versus heating time (min) plots for pH 2, 3, 4, 5, 6, and 7. At pH 8 and 9, the 90 min data point had to be excluded. The calculated empirical rates of hydrogen cyanide production (d [HCN]/dt) at different pH (Table 1) suggest that hydrogen cyanide production is slower at acidic pH values (pH 2, 3, 4, 5, and 6), increases at basic pH, but reaches a peak at pH 7. These findings are similar to the results of the study by Johansen and co-workers [17]. According to their study, hydrolysis of the cyanogenic glycoside, dhurrin, follows a first-order reaction with respect to dhurrin and the rate of dhurrin hydrolysis is very slow at low pH values but strongly increases as the pH is increased. Thus, the first-order rate of hydrolysis of cyanogenic glycoside dhurrin in aqueous solution is supported by the in vitro hydrolysis of cyanogenic glycosides in Macadamia flower as reported by the present study.   [6,43]. Previous reports have mentioned that cyanogenic compounds are highest in growing tissues of plants and that activation of metabolic processes coincide with cyanogenic glycoside production [6,44]. The flower is the main reproductive organ of a plant and has very active and complex morphological and physiological features, which support an abundance of ecological functions related to floral development and plant reproduction [45][46][47]. For instance, de novo synthesis of amino acids, enzymes, and structural proteins, which are precursors of N-containing secondary metabolites (such as cyanogenic glycosides) and signaling molecules, all occur in floral tissues [47]. These complex metabolic processes during floral development and growth could be contributing to the increased biosynthesis of cyanogenic glycosides  The concentration of hydrogen cyanide in Macadamia leaves (513 ± 0.6 ppbv) is within the concentration range of cyanide (364-1403 ppbv) detected in the leaf tissue of M. ternifolia, M. integrifolia, and M. tetraphylla species during their early to mid-developmental stages (3rd-4th week) [6]. Young leaves were observed to contain higher amounts of cyanogenic glycosides, which could be due to the copious amounts of carbon and nitrogen precursors readily available during germination, so there is rapid biosynthesis of cyanogenic compounds. Cyanogenic glycoside in leaf tissue was, however, observed to decrease with plant maturation because these compounds are rapidly metabolized and broken down as the leaves become older [6,[48][49][50].
Macadamia husks are the fleshy green fibrous pericarp covering the conical or spherical hard brown shell enclosure of Macadamia nuts [51]. Similar to Macadamia flowers, there are no available published data reported for the hydrogen cyanide concentration in Macadamia husks for comparison. Moreover, the hydrogen cyanide concentration of husks analyzed from three different Macadamia varieties were significantly different: 21 ± 0.1 ppbv for variety 695; 256 ± 0.4 ppbv for variety Oc; and 476 ± 0.7 ppbv for variety A16 ( Figure 5). It was previously reported that the quantities of cyanogenic glycosides in Macadamia seedlings and other plants vary according to species, developmental stage, and tissue type; however, the cyanogenic glycosides in the varieties of Macadamia husk used in this study have yet to be conclusively identified [6,52].
Seeds of Macadamia species are also capable of accumulating cyanogenic glycoside compounds and the concentration varies depending on the variety [6,43]. In the present study, the hydrogen cyanide concentration of Macadamia nuts (5.8 ± 0.1 ppbv) was lower than the reported cyanide concentrations in commercially used seeds (~74 ppbv) of M. integrifolia or M. tetraphylla and significantly lower than the concentration detected in M. ternifolia (~4800 ppbv), which is considered to be inedible. The concentration of hydrogen cyanide in Macadamia leaves (513 ± 0.6 ppb v ) is within the concentration range of cyanide (364-1403 ppb v ) detected in the leaf tissue of M. ternifolia, M. integrifolia, and M. tetraphylla species during their early to mid-developmental stages (3rd-4th week) [6]. Young leaves were observed to contain higher amounts of cyanogenic glycosides, which could be due to the copious amounts of carbon and nitrogen precursors readily available during germination, so there is rapid biosynthesis of cyanogenic compounds. Cyanogenic glycoside in leaf tissue was, however, observed to decrease with plant maturation because these compounds are rapidly metabolized and broken down as the leaves become older [6,[48][49][50].

Dhurrin and Total Cyanide Concentrations in Untreated and Treated Macadamia Plant Part Extracts
Macadamia husks are the fleshy green fibrous pericarp covering the conical or spherical hard brown shell enclosure of Macadamia nuts [51]. Similar to Macadamia flowers, there are no available published data reported for the hydrogen cyanide concentration in Macadamia husks for comparison. Moreover, the hydrogen cyanide concentration of husks analyzed from three different Macadamia varieties were significantly different: 21 ± 0.1 ppb v for variety 695; 256 ± 0.4 ppb v for variety Oc; and 476 ± 0.7 ppb v for variety A16 ( Figure 5). It was previously reported that the quantities of cyanogenic glycosides in Macadamia seedlings and other plants vary according to species, developmental stage, and tissue type; however, the cyanogenic glycosides in the varieties of Macadamia husk used in this study have yet to be conclusively identified [6,52].
Seeds of Macadamia species are also capable of accumulating cyanogenic glycoside compounds and the concentration varies depending on the variety [6,43]. In the present study, the hydrogen cyanide concentration of Macadamia nuts (5.8 ± 0.1 ppb v ) was lower than the reported cyanide concentrations in commercially used seeds (~74 ppb v ) of M. integrifolia or M. tetraphylla and significantly lower than the concentration detected in M. ternifolia (~4800 ppb v ), which is considered to be inedible.  The cyanide concentration in the extracts have the same trend as the dhurrin concentration. Figure 7 shows the total cyanide concentration of the fresh, untreated Macadamia flower (417.7± 0.8 mg/L), leaves (167 ± 2 mg/L), nuts (67.1 ± 0.6 mg/L), and husks (695: 94.3 ± 0.7 mg/L; A16: 23.9 ± 0.1 mg/L; Oc: 50.6 ± 0.4 mg/L). After full sample treatment using the optimized conditions (i.e., samples treated with pH 7 buffer solution and heated at 50 °C for 60 min), significant amounts of dhurrin and cyanide were removed in the analyzed extracts as shown in Figures 6 and 7, respectively. The cyanide concentration in the extracts have the same trend as the dhurrin concentration. Figure 7 shows the total cyanide concentration of the fresh, untreated Macadamia flower (417.7 ± 0.8 mg/L), leaves (167 ± 2 mg/L), nuts (67.1 ± 0.6 mg/L), and husks (695: 94.3 ± 0.7 mg/L; A16: 23.9 ± 0.1 mg/L; Oc: 50.6 ± 0.4 mg/L). After full sample treatment using the optimized conditions (i.e., samples treated with pH 7 buffer solution and heated at 50 • C for 60 min), significant amounts of dhurrin and cyanide were removed in the analyzed extracts as shown in Figures 6 and 7, respectively.

Dhurrin and Total Cyanide Concentrations in Untreated and Treated Macadamia Plant Part Extracts
It is interesting to note that heating samples at 50 • C for 60 min without pH adjustment (heated-only samples) had little to no effect on the removal of dhurrin ( Figure 6A) or cyanide ( Figure 7A) in Macadamia flower and leaf samples. On the other hand, treating the Macadamia flower and leaf samples with buffered solution at pH 7 without heating (buffered-only), resulted in significant removal of dhurrin ( Figure 6B: flower, 419 ± 1 mg/L; leaves, 98 ± 1 mg/L) and cyanide ( Figure 7A: flower, 370 ± 2 mg/L; leaves, 48 ± 1 mg/L) in the extracts. Analysis of treatment efficiencies (Table 2) showed that the full treatment of samples (i.e., treated samples) by heating (50 • C, 60 min) and pH 7 adjustment results in 93-100% removal of dhurrin and about 81-91% removal of cyanide in the different Macadamia plant parts. Treatment by heating alone was only about 1% effective in the removal of dhurrin and only about 5-12% effective in the removal of cyanide (Table 2). However, treating the Macadamia flower and leaf samples with buffered solution at pH 7 without heating (buffered-only), treatment produced about similar removal effectivity (Table 2) of dhurrin (93-100%) and cyanide (89-91%) as that of the fully treated samples heated at 50 • C for 60 min at pH 7. It is interesting to note that heating samples at 50 °C for 60 min without pH adjustment (heatedonly samples) had little to no effect on the removal of dhurrin ( Figure 6A) or cyanide ( Figure 7A) in Macadamia flower and leaf samples. On the other hand, treating the Macadamia flower and leaf samples with buffered solution at pH 7 without heating (buffered-only), resulted in significant removal of dhurrin ( Figure 6B: flower, 419 ± 1 mg/L; leaves, 98 ± 1 mg/L) and cyanide ( Figure 7A: flower, 370 ± 2 mg/L; leaves, 48 ± 1 mg/L) in the extracts. Analysis of treatment efficiencies (Table 2) showed that the full treatment of samples (i.e., treated samples) by heating (50 °C, 60 min) and pH 7 adjustment results in 93%-100% removal of dhurrin and about 81%-91% removal of cyanide in the different Macadamia plant parts. Treatment by heating alone was only about 1% effective in the removal of dhurrin and only about 5%-12% effective in the removal of cyanide (Table 2). However, treating the Macadamia flower and leaf samples with buffered solution at pH 7 without heating (buffered-only), treatment produced about similar removal effectivity ( Table 2) of dhurrin (93%-100%) and cyanide (89%-91%) as that of the fully treated samples heated at 50 °C for 60 min at pH 7.

Conclusions
The optimum conditions for the maximum release of hydrogen cyanide in Macadamia samples were 50 • C, 60 min at pH 7. Under these treatment conditions, trace amounts of hydrogen cyanide could still be detected in the headspace directly above the different Macadamia plant part samples using SIFT-MS. The measured hydrogen cyanide in the headspace of the treated samples were 12,535 ± 11 ppb v (flower), 513 ± 0.6 ppb v (leaves), 6 ± 0.1 ppb v (nuts), 476 ± 0.7 ppb v (husk A16), 256 ± 0.4 ppb v (husk Oc), and 21 ± 0.1 ppb v (husk 695). Treatment of Macadamia samples under these optimum conditions produced 93-100% removal of dhurrin and 81-91% removal of total cyanide in the sample extracts. Treatment by pH 7 adjustment (buffered-only) without heating also resulted in an effective removal of dhurrin (86-100%) and total cyanide (88-89%) in Macadamia extract similar to the full, optimized treatment conditions. Heating the samples alone at 50 • C for 60 min without pH adjustment was not effective in the hydrolysis and removal of cyanogenic glycoside dhurrin and total cyanide in Macadamia samples. The varying concentration of generated hydrogen cyanide could be correspondingly attributed to the concentration of cyanogenic glycosides (such as dhurrin) from the different parts of the Macadamia plant and their subsequent hydrolysis to hydrogen cyanide. Cyanogenic glycosides were greatest in Macadamia flowers, followed by the leaves and husks (depending on variety), and lowest in nuts. The results indicate that the hydrolysis of cyanogenic glycosides in Macadamia is predominantly induced by pH changes rather than by heat. This further suggests that the enzymatic hydrolysis involved in cyanogenesis is chiefly pH-directed rather than thermally induced. In addition, the hydrolysis reaction of cyanogenic glycosides could be described as a first-order reaction with respect to the in vitro production of hydrogen cyanide.
These results provide further insights into the cyanogenic systems in Macadamia. Moreover, the evaluated optimum conditions for the hydrolysis of dhurrin and removal and release of hydrogen cyanide could be helpful for the effective processing of different parts of Macadamia. Such information provides some guidelines toward the safe production, utilization, and consumption of Macadamia-based products.