The effect of sucralose on flavor sweetness in electronic cigarettes varies between delivery devices

The appeal of sweet electronic cigarette flavors makes it important to identify the chemical compounds that contribute to their sweetness. While volatile chemicals that produce sweet aromas have been identified in e-liquids, there are no published reports of sugars or artificial sweeteners in commercial e-liquids. However, the sweetener sucralose is marketed as an e-liquid additive to commercial flavors. The primary aims of the study were to determine if sucralose is delivered in sufficient concentration in the inhaled aerosol to enhance flavor sweetness, and whether the amount delivered depends on the e-liquid delivery system. Thirty-two adult smokers rated flavor intensity, sweetness, harshness and liking/disliking for 4 commercial flavors with and without sucralose (1%) using 2 e-cigarette delivery systems (cartridge and tank). Participants alternately vaped normally or with the nose pinched closed to block perception of volatile flavor components via olfaction. LC/MS was used to measure the concentration of sucralose in the e-liquid aerosols using a device that mimicked vaping. Sweetness and flavor intensity were perceived much more strongly when olfaction was permitted. The contribution of sucralose to sweetness was significant only for the cartridge system, and the chemical analysis showed that the concentration of sucralose in the aerosol was higher when the cartridge was used. Together these findings indicate that future regulation of sweet flavor additives should focus first on the volatile constituents of e-liquids with the recognition that artificial sweeteners may also contribute to flavor sweetness depending upon e-cigarette design.


Introduction
With the goal of improving public health, the Food and Drug Administration has the authority to evaluate and regulate factors such as the design and constituents of tobacco products that contribute to the appeal of the product. A significant factor in the appeal of electronic cigarettes (e-cigarettes) is the availability of a wide variety of flavors [1][2][3]. Surveys have shown that sweet and fruity flavors are among the most popular, particularly with youth and young adults [4][5][6][7][8], and a recent laboratory study confirmed there is a significant association between sweetness and the degree of liking of e-cigarette flavors [9]. Sweet flavors may therefore heighten the risk of nicotine exposure in this young, at-risk cohort [10]. There is also concern that some sweet flavorants commonly used in foods and beverages have irritant properties when inhaled [11][12][13]. It is therefore important to identify the chemical compounds in commercial flavors (e-liquids) that contribute to the perceived sweetness of the inhaled aerosols and whether device characteristics can influence the delivery of these compounds.
In foods and beverages, sweetness arises primarily from the sense of taste and is served by specialized chemoreceptors in the tongue and palate that are sensitive to sugars and artificial sweeteners [14,15]. However, odors can also have sweet perceptual qualities [16,17], and the senses of olfaction and taste interact to produce overall flavor sweetness [18][19][20]. This perceptual interaction occurs when volatile molecules that are released in the mouth reach the olfactory epithelium via the nasal pharynx ("retronasal olfaction") and produce odors that are perceived in the mouth rather than in the nose [21][22][23][24]. Perceptual "referral" of retronasal olfaction from the nose to the mouth binds tastes and odors into integral perceptions of flavor. Accordingly, it can be difficult to know whether the sweetness of a flavor arises from sweettasting non-volatile ingredients, sweet-smelling volatile ingredients, or both.
Analyses of the constituents of commercially available flavors have identified several wellknown volatile chemicals that have sweet and/or fruity odors [25,26] which we would be expected to be sensed via retronasal olfaction. However, the role of retronasal olfaction in the perception of e-cigarette flavors, including their sweetness, has never been studied. It is less clear whether non-volatile sweeteners that act through the sense of taste might also play a role. We have found no published reports of sugars or artificial sweeteners in chemical analyses of commercial e-liquids. The artificial sweetener sucralose is marketed as an e-cigarette flavor additive and is readily available on the internet. As a high potency sweetener, sucralose is an attractive potential additive, since in theory only very small quantities would be necessary to produce perceptible sweetness. Like sugars and other artificial sweeteners, sucralose is not a volatile molecule and thus is unlikely to be present in large quantities in inhaled e-cigarette aerosols.
The primary objective of the present study was therefore to determine if sucralose is an effective sweetening agent compared to the volatile constituents of sweet and fruity e-liquid flavors. Additionally, because preliminary psychophysical testing indicated that the effect of sucralose on sweetness seemed to depend on the type of e-cigarette used, we tested 2 different e-liquid delivery systems, a cartridge and a tank. To address both questions we employed a psychophysical procedure that enabled separation of the taste and retronasal olfactory components of flavor perception, together with analytical chemistry to quantify the amount of sucralose that was delivered in the e-cigarette aerosols. and New Haven, CT, were paid to participate in the experiment. The research was conducted in accordance with the principles expressed in the Declaration of Helsinki, and the research protocol was approved by the Human Investigations Committee of the Yale University IRB, and all subjects gave written consent before participation. Prior to enrollment, all subjects were screened over the phone to determine eligibility. Only those who spoke English fluently and reported being current daily cigarette smokers for at least 1 year, in overall good health, not pregnant, and with no deficits of taste or smell were scheduled to participate. Subjects were asked to abstain from smoking for at least 10 hours before each session, which was confirmed by alveolar carbon monoxide levels <10 ppm [27] using the MicroCO breath carbon monoxide monitor (Micro Direct, Inc., Lewiston, ME). If the subject's carbon monoxide reading was !10 ppm, the subject was rescheduled for another day.
Equipment and stimuli. Fig 1 shows the 2 e-cigarette delivery systems were used: the V2 blank cartridge (cartomizer) atomizer and the V2 EX blank tank (clearomizer) atomizer (V2™, VMR Products LLC, Miami, FL). Both delivery systems were powered by the standard 79mm V2 e-cigarette batteries (4.2V). Four commercial e-liquid flavors were purchased from an online retailer (AmericaneLiquidStore™, Wauwatosa, WI): Strawberry, Vanilla, Watermelon and Cherry. All e-liquids nominally contained 12 mg/mL of nicotine in a 50/50 PG/VG base. However, gas chromatography (GC) determined the actual nicotine contents were 13.3, 14.1, 13.3 and 14.7mg/mL, and analysis by liquid chromatography followed by detection by The e-liquid stimuli were prepared both with and without added sucralose (Sigma-Aldrich; St. Louis, MO) in a concentration of 1% (wt/vol). Use of this concentration was based on pilot testing which indicated that 1% was sufficient to evoke a weak sweet taste when the e-liquids were vaped using the V2 cartridge with the nose pinched closed, and tests with 5% sucralose did not appear to significantly increase sweetness. The blank cartridges and tanks were filled with 500 μL of e-liquid on the day of testing at least 1h prior to the testing session. The functionality of each cartridge was tested after filling by using a syringe to pull air through the e-cigarette to simulate inhalation. On the rare occasion that a cartridge or tank failed (i.e., no vapor was visible inside the syringe) the cartridge was discarded and a new one was filled and tested. Tanks and cartridges were not used across participants and were disposed of upon test completion.
Experimental design. A within-subjects design was used in which all subjects served in every condition of the experiment. To assess the effects of retronasal olfaction and sucralose on flavor, the design included 2 olfaction and 2 sucralose conditions. The olfaction conditions were nose-open (Olfaction), which allowed the inhaled aerosol to reach the olfactory epithelium through the oro-nasal pharynx during exhalation (i.e., via retronasal olfaction), and noseclosed (No-Olfaction), which prevented the inhaled aerosol from reaching the olfactory epithelium. The 2 sucralose conditions were the e-liquid with sucralose added (Sucralose) and the eliquid without sucralose added (No-Sucralose). The olfactory and sucralose conditions were paired in a crossed design to yield 4 test conditions: (1) Olfaction + Sucralose, (2) Olfaction + No-Sucralose, (3) No-Olfaction + Sucralose, and (4) No-Olfaction + No-Sucralose.
To investigate the potential effects of e-cigarette design on flavor delivery and perception, the 4 commercial flavors (E-liquids) were sampled under all 4 testing conditions using both the V2 cartridge and V2 EX tank (Delivery Systems). The 2 systems were tested in separate sessions and replicate ratings were collected in 2 additional sessions. Subjects therefore served in a total of 4 sessions with each session containing 16 trials (4 test conditions x 4 e-liquids). On each trial data were collected on sensation intensity (overall flavor, sweetness, and harshness/ irritation) and liking/disliking. Within each session the Sucralose and No-Sucralose conditions were blocked to limit potential carryover of sucralose sweetness into No-Sucralose trials. The order in which the sucralose conditions were tested was counterbalanced across subjects and the e-liquids were presented in 4 different pseudo-random orders.
Testing procedure and subjective ratings. In the first testing session subjects were read instructions about how to use the generalized version of the Labeled Magnitude Scale (gLMS) [28][29][30] to rate sensation intensity and the Labeled Hedonic Scale (LHS) [31] to rate liking/disliking. The gLMS and LHS scales were presented sequentially on a computer monitor and the subjects made their ratings by using a mouse to move a cursor to the appropriate locations on the scale. Subjects were given practice using both scales by rating the perceived intensities and liking/disliking of 15 remembered or imagined sensations that were read to them by the experimenter.
Instructions and practice were also given in the use of the 2 e-cigarette delivery systems. In order to fully activate the heating element in V2 e-cigarettes, on each trial subjects took two priming puffs into the mouth only, exhaling them both through the mouth, before fully inhaling a third puff into the airways, which was also exhaled through the mouth. (Note that with the nose open, expired volatiles can nevertheless reach the olfactory epithelium via the retronasal route, including during normal breathing immediately following exhalation). This vaping procedure was used throughout testing. On No-Olfaction trials subjects wore a nose clip which remained on the nose as they vaped and while making ratings of sensation intensity and liking/ disliking. The clip was removed between trials to allow normal breathing through the nose.
On each trial the subject was given a number coded e-cigarette that gave no indication of its flavor, sucralose content, or nicotine concentration. The subjects were instructed to vape each puff as they had during training: 2 priming puffs into the mouth before fully inhaling the third puff, which was always exhaled through the mouth. Subjects rated sensation intensity and liking/disliking immediately after exhalation. The instructions were to base their ratings exclusively on their perception of the third (inhaled) puff. During a 5-min break after the first 8 trials subjects rinsed the mouth with deionized (DI) water to remove any lingering taste before beginning the next block.
Statistical analysis. Analyses were carried out using the repeated-measures module of the advanced linear modeling tool of Statistica™ 13. An overall ANOVA was conducted first with Flavor Characteristic as the dependent variable with 3 levels (overall intensity, sweetness, and harshness/irritation) to evaluate the predicted effects of Sucralose, Olfaction and Delivery System on flavor perception. Specifically, it was expected that ratings of overall flavor and sweetness would be higher in the Olfaction and Sucralose conditions, and that the effect of adding sucralose would be dependent upon Delivery System. After finding significant main effects and interactions among these factors as well as E-liquid, individual repeated-measures ANO-VAs together with Tukey HSD tests were conducted to clarify the sources of the interactions. Ratings of Liking/Disliking were analyzed in a separate ANOVA. Prior to statistical analysis the raw perceived intensity ratings were converted to log 10 to adjust for the log-normal distribution of the data that is characteristic of the gLMS [29,32]. No log conversion is necessary for liking/disliking data collected using the LHS [31].

Analytical chemistry methods
An analytical system was setup that mimicked the parameters of vaping.
The vaping setup consisted of a micro diaphragm pump (KNF Neuberger Inc., Trenton, NJ) powered by a DC power supply (Velleman, Inc., Fort Worth, TX) and regulated with a needle valve, followed by a flow meter (Omega Engineering, Inc., Norwalk, CT), a cold finger trap chilled in liquid nitrogen (Airgas, Radnor, PA), and a reducing connector (McMaster-Carr, Elmhurst, IL) attached directly to the cold-finger trap. Before filling the cartridge or tank with e-liquid, its resistance was measured using an electronic multimeter. Subsequently, approximately 500 mg or 700-800 μL of the e-liquids in question were added to the cartridges and tanks, respectively, and these were allowed to settle for at least 1 h. To carry out the vaping experiment, a flow rate of 1.1 L/min was set using a blank cartridge and the filled cartridge or tank was subsequently attached to the reducing connector and the V2 standard 79mm battery was screwed onto the cartridge or tank. All connections were sealed using threaded seal tape (McMaster-Carr, Elhurst, IL) where necessary, or parafilm (Bemis, Neenah, WI). The power supply was controlled using a programmable Arduino Uno controller (Arduino, Somerville, MA) which was programmed to take 3 primer puffs of 0.25 s followed by one, 4-s puff. This procedure resulted in a puff volume of 73.3 ml and was based on preliminary in-house vaping topography data that were collected from 11 subjects who reported regular e-cigarette use as well as data from the literature [33]. After allowing the e-cigarette to cool down for 30 s, the procedure was repeated for a total of 20 times to collect a sufficient amount of vapor. Once 20 puffs were collected, the setup was disassembled and the cold finger trap was allowed to thaw. The content of the trap was dissolved in DI water and the condensed liquid inside the removable mouthpiece of the cartridge or tank was carefully rinsed into a vial.
Liquid chromatography-Mass spectrometry (LC/MS). The amount of sucralose in both the trap and the mouthpiece was quantified using LC/MS (Varian 500-MS, 212-LC pump system, electrospray ionization, Palo Alto, CA) based on a method established previously [34].
A reverse phase column (Atlantis dc18; Waters Limited, Milford, MA) with 100Å particle pore size, 3 μm particle size, 3.9 mm ID, 150 mm length, and the corresponding inline guard column (Atlantis dc18 VanGuard cartridge; Waters Limited Milford, MA) was used. Sample injection was done with a Varian Prostar 420 autosampler (Palo Alto, CA) into a 20 μL sample loop. HPLC-grade water (solvent A) and acetonitrile (solvent B, both Fisher Scientific, Waltham, MA) were used to run an elution gradient at 35˚C and a flow rate of 200 μL/min for a total runtime of 46 min. B was held constant at 5% for 2 min followed by a linear gradient to reach 95% at 15 min, and held constant again until 30 min. To ensure complete removal of any material, a steep gradient to 5% B was run until 31 min, and B was held again at 5% until 46min. The MS was run in negative mode with a capillary voltage of 65V, RF loading of 110%, and needle voltage of -4000V.

Psychophysical data
The data on the perceived intensity of the e-liquid flavors are summarized in Fig 2. Most evident in the figure is a strong dependence of overall flavor and sweetness intensity on the availability of retronasal olfaction (Olfaction condition). Analysis of the full data set confirmed there was a significant main effect of Olfaction (p<0.00001) and a 3-way interaction among Olfaction, Flavor Characteristic and e-Liquid [p<0.01]. For all 4 e-liquids, overall flavor and sweetness were rated slightly above "barely detectable" with the nose closed, whereas ratings of harshness/irritation were reduced to a lesser degree, particularly for the cherry-flavored e-liquid. The same analysis also showed that flavor perception depended upon Delivery System in complex ways, including a 4-way interaction with Sucralose, E-liquid, and Flavor Characteristic (p<0.01). To investigate this interaction separate ANOVAs were conducted on the data from the cartridge and tanks systems alone. Analysis of the cartridge data confirmed there was a main effect of Sucralose (p<0.0001) and an interaction between Sucralose and Flavor Characteristic (p<0.00001), which indicated that adding sucralose increased overall flavor and sweetness but not harshness/irritation. This advantage can be seen in Fig 2, where sucralose led to higher mean ratings of overall flavor and sweetness for all 4 e-liquids when the V2 cartridge (cartomizer) was used. The same analysis showed that an interaction between Sucralose and Eliquid fell just short of significance. In contrast, analysis of the V2 tank data found no main effect of Sucralose but instead uncovered an interaction between Sucralose and Flavor Characteristic (p<0.05) which reflected a slight decrease in harshness/irritation rather than an increase in overall flavor and sweetness.
To focus on the effect of adding sucralose under conditions of normal vaping with the nose open we conducted an ANOVA on the data from the Olfaction condition alone. The analysis revealed a significant 3-way interaction among Sucralose, Delivery System, and Flavor Characteristic (p<0.05) which was again driven by increased ratings of overall flavor and sweetness when the cartridge delivery system was used. The analysis did not find a significant 4-way interaction that included E-liquid (p = 0.16), confirming that the flavor enhancing effect of sucralose was similar across e-liquid flavors. However, an ANOVA on the sweetness data alone confirmed there was a significantly stronger effect of sucralose for the cartridge compared to the tank (Sucralose x Delivery System; p<0.001) which, however, was not consistent across eliquids (Sucralose x E-liquid interaction; p<0.05). Tukey tests revealed that the latter interaction was driven by a significant increase in the sweetness of the cherry flavored e-liquid; increases in the sweetness of the other e-liquids were not significant. Analysis of the harshness/ irritation data in the Olfaction condition uncovered a main effect of Delivery System (p<0.05), with the tank system producing more sensory irritation than the cartridge system. A 3-way  Table 1 summarizes the results of the vaping experiments with the data arranged by the type of container used (V2 cartridge or tank) and e-liquid flavor. Additionally, sucralose was  About twice as much e-liquid was consumed from the tank over the course of 20 puffs compared to the cartridge system (two-tailed t-test, both p < 0.0001) The amount of condensate found inside the mouthpiece of the tank-system was also significantly greater than for the cartridge-system (two-tailed t-test, both p < 0.0001). The latter difference might be explained by the difference in size and geometry of the two mouthpieces. For the cartridge, the channel that delivers the vapor from the vaporizer to the mouth is only~4mm long with a diameter of ca. 2.5mm, while for the tank the channel is conically shaped with an overall length~24.4mm and a diameter that increases from~3.4mm to~5.2mm at the tip. Fig 4 shows the average sucralose concentrations found for the cartridge and tank across all 4 flavors. Concentrations were significantly higher in the mouthpiece condensate than in the vapor trap for both the tank and the cartridge (two-tailed t-test, both p< 0.0001), indicating that sucralose tended to accumulate in the mouthpieces. Although the absolute amounts of sucralose in the vapor trap were not significantly different between the cartridge and tank (two-tailed t-test, p = 0.8694; see Table 1), the cartridge yielded higher sucralose concentrations than the tank in both the condensate of the mouthpiece (two-tailed t-test, p< 0.05) and the vapor trap (two-tailed t-test, p< 0.002).

Discussion
As was pointed out in the Introduction, it is well known that some odors have sweet qualities that can be confused with sweet taste, and that the sweetness of odors sensed retronasally contributes to the perceived sweetness of foods and beverages [18][19][20]. The present results show that by themselves, sweet odors can produce the characterizing sweetness of e-cigarette sweet and fruity flavors. Given the absence to date of published evidence of artificial sweeteners in Table 1. Mass change of the e-liquid after 20 puffs as well as amount of condensate found inside of the mouthpiece, the mass, and resulting concentration (μg/mg) of sucralose found in the trapped vapor and the mouthpiece condensate.

Vapor (found in trap)
Condensate (found in mouthpiece) commercial e-liquid flavors, it is highly likely that the sweetening agents in most e-cigarette flavors are sweet-smelling odors. In addition, our results show that while adding sucralose to commercial sweet e-liquid flavors can add to their sweetness, the increases in sweetness were proportionally small and were insufficient to significantly increase flavor liking. Although higher concentrations of sucralose could potentially increase sweetness more effectively, the 1% concentration used here is more than 100 times higher than the concentration needed to produce a clearly perceptible sweet taste in aqueous solutions (% 0.008%) [35], and informal testing with 5% sucralose did not seem to produce a noticeable increase in e-liquid sweetness. The finding that sucralose produced significantly more sweetness with the V2 cartridge than with the V2 EX tank system indicates that the effectiveness of adding an artificial sweetener can depend on the e-cigarette delivery system. This result was somewhat surprising because both systems use the same 4.2V battery and 3O heating coil and so should produce similar atomizing temperatures. Therefore other design features (see Fig 1) must have affected delivery of sucralose into the vapor. For example, e-liquid delivery to the heating coils is different between the V2 cartridge and the V2 EX tank. Although in both systems the e-liquid is brought to the heating coil via a cotton-fiber blend wicking system, in the cartridge the e-liquid is contained within a sheet of rolled cotton-fiber that surrounds the heated air channel, while in the tank the e-liquid is drawn from the tank into the heating channel through a cotton-fiber strip. These different wicking designs may differentially affect diffusion of sucralose into the heating channel. In addition, as noted above, the mouth piece of the tank system has a much larger surface area on which the aerosol could condense during vaping. During analytical testing we noticed that tiny droplets of condensate often appeared on the outer edge of the cartridge mouth piece but not the tank mouth piece, raising the possibility that sucralose in the condensate could have been tasted if participants touched their tongue tip to the mouth piece. However, because the droplets were observed during analytical testing after bouts of 20 simulated puffs, it is unclear whether they would form after the 3 puffs that were used during psychophysical testing.
As shown in Table 1, the greater total mass of sucralose measured in the mouthpiece and trap for the tank compared to the cartridge system seems in conflict with the higher ratings of overall flavor and sweetness when sucralose was added in the cartridge system. However, it is well known that the potency of taste stimulation depends more upon stimulus concentration than upon stimulus mass, as is evident from the use of molarity or molar concentration (defined as the number of moles of a substance per liter of solute) as the standard measure of stimulus strength in taste research [36]. The significantly higher sucralose concentration in the vapor of the cartridge compared to the vapor from the tank is therefore a likely reason for the higher sweetness reported with the cartridge (Table 1 and Fig 4). The source of this difference in sucralose concentration is unclear, as it could depend upon one or more of the differences between the delivery systems discussed above. The finding nevertheless demonstrates that the effectiveness of adding non-volatile sweeteners to e-liquids can depend in part on e-cigarette design, and so may vary even more between delivery systems that also have different atomizer voltages and temperatures.
More generally, the evidence that the e-cigarette delivery system can alter the effectiveness of sucralose as a sweetening agent has implications for the use of high-potency sweeteners in other types of tobacco products. While non-volatile sweeteners like sucralose have the potential to strongly affect the flavor of products in which sweeteners come in direct contact with the mouth and tongue (e.g., chewing tobacco or cigars in which sweeteners are added to the tobacco leaf wrapping), the present findings suggest their effectiveness in products that rely on delivery of flavoring agents solely via inhalation (e.g., hookas) is likely to be relatively small compared to volatile flavor molecules and dependent upon the specific design of the devices.

Conclusions
The present results demonstrate that the perceived sweetness of flavors derives primarily from volatile components of e-liquids that are sensed via retronasal olfaction rather than by taste. Although it was found that the artificial sweetener sucralose can add to the sweetness of flavors, the amount of sweetness added was small and had little effect on flavor liking. However, the added sweetness depended on the e-cigarette delivery system. Together these findings suggest that any future regulation of flavor sweetness in e-cigarettes should focus primarily on the volatile (olfactory) constituents of e-liquids, with the caveat that use of more powerful e-cigarette systems, or systems that are specifically designed to deliver artificial sweeteners into the inhaled aerosol, may enable non-volatile sweeteners to contribute more strongly to the appeal of e-cigarettes.