1 Review 2 3 Neurological connections and endogenous biochemistry - potentially useful in electronic- 4 nose diagnostics for coronavirus diseases 5 6 Tiffany C. Miller1, Salvatore D. Morgera1, Stephen E. Saddow1,2, Arash Takshi1, Matthew 7 Mullarkey3, Matthew Palm4 8 9 1Department of Electrical Engineering, University of South Florida, Tampa, FL 33620, USA. 10 2Department of Medical Engineering, University of South Florida, Tampa, FL 33620, USA. 11 3Muma College of Business, University of South Florida, Tampa, FL 33620, USA. 12 4Valhall K-9 International, LLC, Hull, GA 30646, USA. 13 14 Correspondence to: Dr. Salvatore D. Morgera, Department of Electrical Engineering, 15 University of South Florida, 4202 E. Fowler Ave., Tampa, FL 33620, USA, E-mail: 16 [email protected] 17 18 How to cite this article: Miller TC, Morgera SD, Saddow SE, Takshi A, Mullarkey M, Palm M. 19 Neurological connections and endogenous biochemistry - potentially useful in electronic-nose 20 diagnostics for coronavirus diseases. Neuroimmunol Neuroinflammation 2021;8:[Accept]. 21 http://dx.doi.org/10.20517/2347-8659.2021.05 22 23 Received: 23 Feb 2021 Revised: 11 Jun 2021 Accepted: 29 Jun 2021 First online: 14 Jul 24 2021 25 26 27 ABSTRACT 28 As our understanding of infectious diseases, such as coronavirus diseases including, Coronavirus 29 Disease 2019 (COVID-19), as well as human respiratory viral and nonviral diseases, improves, 30 we expect to uncover a better understanding of the pathogenesis of the disease as it relates to 31 neuroinflammation. This may include associated biomarkers of immune response for

1 32 neuroinflammation, central nervous system injury, and/or peripheral nervous system injury 33 emitted from the breath and/or of an individual. Electronic nose gas sensing technology 34 may have the potential to substantially slow the spread of contagious diseases with rapid 35 diagnostic signal indication for detecting these emitted biomarkers of disease. The biochemistry 36 behind the disease is critical for revealing the target gasses emitted in the breath of a severe acute 37 respiratory syndrome coronavirus (SARS-CoV-2) infected individual for the development of 38 such a selective diagnostic screening tool. In this paper, we comprehensively review the 39 evidence that SARS-CoV-2 infection involves an inflammatory response mechanism involving 40 the olfactory nerve route in the brain from the nasal canal suggesting potential candidate volatile 41 organic compound target gas biomarkers, forming the breath pattern signature of COVID-19, 42 that may be detected by electronic nose technologies from a breath sample. 43 44 Key words: COVID-19, biomarkers, neuroinflammation, volatile organic compounds, CNS 45 disease, inflammatory response 46 47 48 INTRODUCTION 49 In response to problems in the field investigated in recent studies by breath specialists based in 50 Edinburgh, UK, in Dortmund, Germany[1], and in Jacksonville, Florida U.S., a weekly virtual 51 multidisciplinary meeting was established at the University of South Florida in April 2020 in 52 collaboration with canine detection specialists Valhall K-9 International, LLC to discuss and 53 begin understanding neurological manifestations of coronavirus diseases including, Coronavirus 54 Disease 2019 (COVID-19), and potential diagnostic breath sensor devices in an attempt to 55 address a long felt, yet unfulfilled need to standardize electronic nose methodologies into a 56 reproducible method for detection of low noise data signature profiles of disease[2]. In particular, 57 brain breath biochemistry and its associated biomarkers of neuroinflammation will be the main 58 focus of this paper as they show the most pertinent relevance to candidate endogenous 59 compounds of a COVID-19 breath pattern signature. 60 61 This review paper will first describe symptoms and the standard diagnostic testing method of 62 COVID-19. Next, the clinical detection of COVID-19 in a patient’s breath by various electronic

2 63 nose devices will be discussed. Third, brain breath biochemistry in neurological complications of 64 COVID-19 will be discussed including the infection mechanism of severe acute respiratory 65 syndrome coronavirus (SARS-CoV-2) with a specific focus on its angiotensin-converting 66 enzyme 2 (ACE2) interaction, viral infection routes from the olfactory epithelium at the nasal 67 cavity to the central nervous system (CNS). Fourth, neurological manifestations of COVID-19 68 will be detailed including, COVID-19 Induced Pediatric Multisystem Inflammatory Syndrome 69 (PMIS). Finally, the conclusion and future work in the development of an electronic nose for 70 detecting COVID-19 will be described in its potential to determine a specific disease from a 71 complex mixture of volatile organic compound (VOC) inflammatory biomarkers. 72 73 COVID-19 SYMPTOMS AND DIAGNOSTIC TESTING 74 Symptoms associated with COVID-19 75 On March 11, 2020 SARS-CoV-2 infection has facilitated a respiratory disease pandemic known 76 as COVID-19. Some symptoms of COVID-19 that have been observed in infected individuals 77 comprise and ageusia. Anosmia is characterized by a drastic reduction of function the 78 olfactory system, whereby, the detection of is diminished or lost. It is known that the 79 ageusia symptom features the inability to taste[3,4]. Although a loss of smell has been associated 80 with neurodegenerative diseases from other pathogens prior to the COVID-19 pandemic, both 81 anosmia and ageusia have become an increasingly known observed symptom of infection from 82 severe acute respiratory syndrome coronavirus (SARS-CoV) or SARS-CoV-2 which causes 83 COVID-19. Thus, understanding the biomechanism of anosmia may provide anosmia specific 84 biomarkers capable of being added to the known breath pattern signature of COVID-19 currently 85 consisting of increases in acetone, isoprene, heptanal, propanol, propanal, butanone, octanal with 86 a decrease on methanol, thereby, differentiating COVID-19 infection from other pathogens of 87 disease[1]. The incorporation of additional biomarkers from the anosmia symptom of COVID-19 88 contributes to the unique exhaled breath profile of COVID-19 to increase the accuracy of a 89 diagnostic electronic nose for exhaled breath applications. Although SARS-CoV-2 infection 90 causes COVID-19, studies have indicated a plurality of viruses access the CNS and/or the 91 peripheral nervous system (PNS) by traversing the length of the olfactory nerve. Once the virus 92 has been incorporated within the nervous system, the activation of T lymphocytes triggers an 93 inflammatory response from microglia and other inflammatory mediators[5,6]. This inflammatory

3 94 response generates biomarker byproducts that are released into the surrounding environment 95 through many pathways. For example, the lungs may generate disease specific metabolites that 96 are exhaled through a breath. This breath sample may be analyzed by an electronic nose for 97 abnormal chemical biomarkers that are associated with a disease[7]. 98 99 Diagnostic testing of COVID-19 100 Due to the highly contagious and life-threatening characteristics of COVID-19, the Emergency 101 Use Authorization (EUA) authority has allowed the use of real-time reverse transcription 102 polymerase chain reaction (rRT-PCR) testing to detect RNA from SARS-CoV-2 in nasal, 103 nasopharyngeal, and oropharyngeal swabs from patients exhibiting symptoms of the virus. This 104 rRT-PCR method requires target-specific fluorescent labeled oligonucleotide probes which 105 create a signal when they are bound to the amplified RNA[8]. This lengthy procedure requires 106 many steps during preparation of the sample including, but not limited to, collecting samples 107 from nasal swabs, providing quinidine isothiocyanate to disrupt the SARS-CoV-2 virions, 108 providing a magnetic nanoparticle to capture the nucleic acids, washing the sample to remove 109 inhibitors and unbound sample components, providing a nucleic acid detection kit for separating 110 the nucleic acids from the magnetic nanoparticle with a buffer solution, amplifying the nucleic 111 acids in a deep well plate, an internal control reagent is introduced to the wells and the specimen 112 is plated, amplified, and detected on a reaction plate[8]. This diagnostic process requires many 113 materials associated with an increase in cost and is subject to human error along the process 114 which may adversely affect the accuracy of the testing result. Further, between April-June 2020, 115 there have been shortages of genetic extraction kits, swabs, reagents, masks, and gowns which 116 have halted COVID-19 testing. 117 118 Many countries have been implementing reverse transcription-polymerase chain reaction PCR 119 COVID-19 testing of its population to reduce the spread of COVID-19. Studies have revealed 120 that lockdowns and mobility reductions decreased the transmission rate of COVID-19 in Europe 121 and North America[9]. However, manufacturers of the current polymerase chain reaction (PCR) 122 COVID-19 diagnostic test kits struggle to keep up with demand and experience testing kit 123 shortages and long processing times[10]. As a result, test results may be delayed which then 124 delays the intervention of quarantining the potentially infected individual who may be

4 125 transmitting viral particles within the wait period for their test results, which is typically 3 or 126 more days. Thus, there is a need for a rapid COVID-19 screening test that is capable of providing 127 results in real-time so that a confirmed COVID-19 infected individual can be alerted of their 128 infection status within seconds of taking the test. A confirmed COVID-19 positive individual 129 would be immediately advised by a healthcare professional, at point of care, to quarantine and 130 isolate to slow the spread of infection, so that businesses and schools can continue providing 131 services in a COVID-19 free environment. 132 133 Although rRT-PCR is the current standard and diagnostic testing method due to its high 134 sensitivity having an approximate range between 66%-83% for detecting the RNA of SARS- 135 CoV-2, studies indicate false-negative (FN) results from respiratory samples for SARS-CoV-2 136 and show FN rates (FNRs) having a range between 1% to 30% due to, for example, swab 137 contamination from poor sample collection, low viral load, different rates of viral shedding over 138 time in the disease process, and decreased analytic sensitivity[11,12]. Thus, the substantially 139 increased range of FNRs with rRT-PCR test results serves as a motivation for a diagnostic 140 electronic nose breath analysis testing method. In particular, it depending on the type of 141 electronic chemical gas sensors used and the amount of these sensors that comprise a rapid 142 electronic nose device, the high selectivity of these sensors to the breath pattern signature of 143 coronavirus disease would greatly reduce the number of false positives and false negatives that 144 have been demonstrated by existing rRT-PCR diagnostic testing methods. 145 146 Further, it would be more desirable to have a method of COVID-19 testing that eliminates the 147 need for excessive equipment, supplies, and multiple steps. Since there have been shortages of 148 testing materials such as swabs and associated reagents to perform rRT-PCR, some non- 149 electronic nose investigations have sought out COVID-19 patient breath samples for processing 150 and COVID-19 VOC signature detection with a gas chromatograph ion mobility spectrometer 151 (GC-IMS) device. Evidence suggests a signature of inflammatory and oxidative stress 152 metabolites detected from COVID-19 positive patients using (GC) and Ion 153 Mobility Spectrometry (IMS) include, acetone, alcohol, butanone, methanol, isoprene, heptanal, 154 propanol, propanal and octanal[1]. Other studies have indicated VOC concentrations of 155 approximately 10 to 250 parts-per-billion (ppb) of methylpent-2-enal, 2,4-octadiene, 1-

5 156 chloroheptane, and nonanal in breath samples obtained from exhaled breath samples of COVID- 157 19 infected patients[13,14]. 158 159 ELECTRONIC NOSE DETECTION OF BIOMARKERS OF RESPIRATORY DISEASE 160 Electronic noses have been shown to successfully recognize complex VOC mixtures from a 161 sample breath indicative of a metabolic process in respiratory diseases. Many electronic nose 162 devices comprise an electronic configured for detection of a target gas, 163 such as VOCs, and implements to differentiate the target gas from a diversity 164 of gasses[15]. Current existing electronic nose technologies have been developed and investigated 165 for their diagnostic performance in respiratory diseases including, but not limited to, 166 inflammatory lung disease such as asthma[16,17]. For example, the Cyranose 320 is an electronic 167 nose having a carbon black-polymer sensor array[17]. Studies have investigated the performance 168 of the Cyranose 320, configured for 96% accuracy in distinguishing patients with asthma from 169 patients with Chronic obstructive pulmonary disease (COPD)[16]. Another study indicated the 170 Cyranose 320 could differentiate patients having mild asthma from patients having severe 171 asthma at a cross-validation value (CVV) of 65%[17,18]. In yet another study, the Cyranose 320 172 was shown to discriminate between healthy controls, controlled asthma, and partially-controlled 173 and uncontrolled asthma with an area under the receiver operator characteristic curve of 0.85, a 174 sensitivity of 0.79, and a specificity of 0.84[17,19]. The Aeonose is an electronic nose device 175 having micro hotplate metal-oxide sensors and demonstrated a lower accuracy and sensitivity 176 compared to the Cyranose 320 when discriminating asthma from healthy controls, whereby, the 177 Aeonose performance results indicate an area under the receiver operator characteristic curve 178 curve of 0.79, a sensitivity of 0.74, and a specificity of 0.91[17]. 179 180 Although this section presents and discusses performance results of electronic nose devices 181 configured for use in diagnostic breath analysis applications that have shown moderate to good 182 accuracy in providing a rapid detection of the breath pattern of biomarkers of respiratory disease, 183 electronic noses may also be usefully extended to diagnostic applications of other inflammatory 184 and infectious diseases[15]. Due to their portability, ability to rapidly provide point-of-care 185 diagnostic test results without a laboratory technician and their low cost elevate the electronic 186 nose to an attractive diagnostic tool in breath analysis. In regards to low cost, for example, the

6 187 electronic nose electrical components of, an electronic chemical gas sensor, a microcontroller, 188 and wiring may cost less than approximately $15.00 in parts. 189 190 ELECTRONIC NOSE SIGNAL PROCESSING CORRECTION 191 As a result of metal oxide semiconductor (MOS) based gas sensor commercial availability, low 192 cost, and favorable sensor properties, including a high selectivity to target gases and a short 193 recovery time, the MQ-2 and MQ-135 gas sensors are commonly used gas sensor component for 194 electronic nose technologies for biomedical applications. However, their suitability has yet been 195 determined for use in electronic nose systems directed at real-time monitoring for breath analysis 196 applications[20,21]. The problem with the use of MOS-based sensors in electronic nose technology 197 is the system experiencing instability due to the output response of the sensor not being under 198 stable control as a result of gradual drifts in the output response of the MQ-2 and MQ-135 199 sensors, resulting in the inability to successfully reproduce output responses using these sensors. 200 For example, a COVID-19 breath simulation study utilized regression analysis to determine the 201 linear model data between Time (s) and the Output Response (V) of an MQ-2 and MQ-135 gas 202 sensor for an electronic nose prototype[22]. 203 204 The study results determined the correlation values of the output response peaks and the 205 concentration gradient of the sample solution have a fitting formula with a measured relativity 206 (R2) > 0.9073 for the MQ-2 gas sensor and a R2 > 0.3963 for the MQ-135 gas sensor[22]. It would 207 be more desirable to have a gas sensor having an increased reproducibility of more controlled 208 output responses so that acquired gas sensor data obtained from a laboratory setting may be 209 validated and more efficiently applied toward breath analysis applications[23]. This baseline 210 instability directly pertains to the baseline output response of the MQ-2 and MQ-135 sensors not 211 being accurately obtained due to factors such as drifts, environmental influences, and sampling 212 means[24]. In particular, as it relates to drifts, as a slight variation is observed in MOS-based gas 213 sensor output responses of electronic nose devices having the sensors exposed to an unchanged 214 target gas sample with an unchanged concentration value having the surrounding environment at 215 an ambient status. 216

7 217 Drift in a sensor is a natural phenomenon in which a low frequency change occurs. Drift changes 218 occurring between sample intervals has been investigated to be attributed to many factors 219 including, the degradation of the heating element comprising a nickel-chromium coil surrounded 220 by a layer of aluminum oxide-based ceramic of the sensor over time or damage to the sensing 221 element, comprising a layer of tin dioxide overlaying the heating element, from chemicals that 222 are capable of compromising the integrity of the sensor components[25,26]. Further, unstable drift 223 changes of MOS-based gas sensors may result from changes in the environment of the sample 224 chamber such as, temperature, humidity, flow rate of the sample, and concentration of target 225 gas[27]. In an example, drift variation is magnified in breath analysis applications when the 226 sensors of an electronic nose device are exposed to a diversity of target gasses within an exhaled 227 breath sample, at varying concentrations, and a changing flow rate corresponding to the 228 inconsistent force of the exhaled breath emitted from an individual’s lungs. 229 230 In addition, undesirable drift variation exists in MOS-based gas sensors due in part to the 231 sampling mechanism in which the heated tin dioxide layer of the gas sensor is configured for 232 oxygen adsorption on its surface. This oxygen adsorption decreases the flow of electric current 233 flowing through the coil of the MOS gas sensor until the surfaced density of adsorbed oxygen 234 lowers during sensor exposure with a target gas. As a result of target gas exposure to a MOS gas 235 sensor, a resistance change having a value measured using a voltage divider network occurs and 236 is associated with the detection of a specific gas. Thus, in the presence of reducing gases, more 237 electrons are capable of flowing across the coil. The sampling mechanism comprising the 238 electron flow across a coil of a MOS gas sensors requires lengthy cycling time when a saturated 239 sensor refreshes back to baseline drift. This response and recovery time may hinder the accuracy 240 of results and not be suitable for breath analysis applications unless corrected using signal 241 processing[24]. Although an efficient cleaning process is a solution to minimizing drifts after each 242 gas exposure phase, the cleaning process requires a length of time that is not suitable for an 243 electronic nose in clinical applications for breath analysis. Thus, there is a need for a signal 244 processing method to minimize the effects of drift on MOS-based gas sensor output responses so 245 that the sensor response drifts are reduced to provide quicker and continuous measurements. As a 246 result, fast measurements as well as maintaining the accuracy of the results are expected from 247 applying the signal processing cleaning step[28].

8 248 249 Current signal correction methods, such as univariate and multivariate are configured for 250 increasing the quality of the sensor output response. In an example, univariate methods are 251 capable of applying the correction to one variable, such as the gas sensor during baseline 252 manipulation[29]. In particular, the output response of the sensor is altered by the initial 253 measurement of the output response. This alteration is commonly referred to as a transformation 254 and is configured for the correction, such as differential, relative, or fractional, of the sensor’s 255 signal baseline. Some studies have indicated the use of baseline manipulation as a means for 256 preliminary processing of the output response of the gas sensor[30]. Other studies have utilized 257 various filtering techniques such as Moving Median Filter (MMF), Fourier Bandpass Filter 258 (FBF), and Discrete Wavelet Transform (DWT) in an attempt to eliminate adverse drift effects 259 from acquired gas data[30,31]. Due to the capabilities of the discrete wavelet transform, a signal 260 may be examined at differing variables such as frequency band levels and resolutions. 261 262 Studies have indicated a correlation between output responses of gas sensors and drift effects that 263 have influenced the direction of electronic nose technologies towards utilizing multivariate 264 signal correction methods to apply the correction to many variables, such as a plurality of gas 265 sensors, resulting in an increased amount of data being detected for modeling of nonlinear drift 266 effects[32]. In an example, Self-Organizing Maps (SOMs) are a type of adaptive neural network 267 that have been incorporated within the signal processing of some electronic nose sensors[33]. 268 Although the use of this adaptive neural network is capable of accomplishing accurate results for 269 the application of gas analysis, the quantification of acquired gas sensor output responses have 270 been identified as a major limitation[34,35]. Taking into consideration of the literature review, it 271 can be discerned that drift is associated with the output response of a gas sensor even during 272 exposure of to a target gas having consistent variables such as, temperature, humidity, and time. 273 It would be more desirable to minimize the variance of drift. Further, some studies have 274 investigated and favored the use of OSC algorithms to eliminate the variance that is absent any 275 correlation to the variable for approximate calculation[36]. 276 277 In another example of multivariate linear correction being applied in other studies, the use of 278 Partial Least Square (PLS) and Principal Component Analysis (PCA) has contributed to the

9 279 decrease in variance of the gas sensor drift as it flows in a single pathway[27,35,37]. Although a 280 high accuracy is achieved with PLS and PLS with orthogonal signal correction (OSC), when 281 measuring the error of the model in predicting quantitative data, it is more desirable in some 282 applications to have a lower Root Mean Square Error (RMSE) value because the average 283 distance from the predicted values of the model and the dataset values are in closer proximity. A 284 comparison of the stability and accuracy of the regression model of acquired data from an 285 electronic nose prototype with the application of PLS and PLS with OSC resulted in an increased 286 accuracy with both methods, however, the PLS with OSC method required less components than 287 with PLS alone to achieve the same performance[38]. In particular, the signal correction method 288 comprising both PLS with OSC for corrected data achieved a RMSE value of approximately 289 0.12% using 1 component compared to PLS alone achieving a RMSE value of approximately 290 0.15% using 5 components. In a study, OSC was used for the correction of near-infrared spectra 291 (NIR) and many publications described machine learning algorithms to improve the processing 292 of the OSC method to eliminate the variance not correlated to the variable to estimate[38,39]. 293 294 The signal processing approach to quantification of acquired gas sensor output response data 295 includes applying baseline manipulation with OSC in an attempt to eliminate and/or minimize 296 the drift effects of the gas sensors responding to varying concentrations of an exhaled breath 297 sample. Next, a regression model to determine the approximate concentration of gas detected in 298 ppm is realized after the corrected datasheet undergoes PLS regression. As the gas sensors are 299 exposed to an exhaled breath sample, PLS regression is a good candidate to be applied to model 300 the output responses of the gas sensors. Validation of acquired gas data results are critical for 301 increasing the reproducibility of the sensor output responses for enhancing the regression model 302 stability while maintaining an accuracy suitable for breath analysis applications. It is envisioned 303 that this drift compensation method presented will enhance electronic nose design capabilities. 304 305 REPORTING DIAGNOSTIC ACCURACY 306 The reproducibility, completeness, and transparent reporting of research may be facilitated by 307 the recommendations set forth by the Standards for Reporting Diagnostic Accuracy (STARD) 308 statement[40]. Some of the STARD recommendations are directed to the study design, 309 participants, test methods, analysis, test results. For example, a study design method

10 310 recommendation includes, but is not limited to, identification as a study of diagnostic accuracy 311 such as sensitivity, specificity, predictive values, or AUC[40]. As it pertains to the diagnostic 312 testing of individuals for SARS-CoV-2, an electronic nose device having gas sensors with a high 313 selectivity to target gasses would be more desirable. Further, the Transparent Reporting of a 314 multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD) statement sets 315 forth reporting guidelines for the development and validation of prediction models in diagnostic 316 applications[41]. Some of the TRIPOD recommendations are directed to the title, abstract, 317 background and objective, source of data, participants, outcome, predictors, sample size, missing 318 data, statistical analysis methods, risk groups, and development vs. validation. For example, a 319 statistical analysis method recommendation includes, but is not limited to, specify the type of 320 model, all model-building procedures, and methods for internal validation[41]. As it pertains to 321 the diagnostic testing of individuals for SARS-CoV-2, an electronic nose device incorporated 322 with OSC, PLS, and PCA to correct drift variations is most desirable. 323 324 DISCUSSION OF NEUROLOGICAL MANIFESTATIONS OF COVID-19 325 Brain breath biochemistry in neurological complications of COVID-19 326 Currently, not many investigations have actively explored the infection mechanism of SARS- 327 CoV-2 and its ability to infiltrate the nervous system. Understanding the biological 328 communication network of the olfactory nerve is essential in expanding the knowledge in 329 regards to SARS-CoV-2 transmission between individuals and metabolic byproducts of infection. 330 It would be particularly useful in the field of diagnostic electronic nose technologies to identify 331 quantitative biomarkers of immune response to include in the breath pattern signature of 332 COVID-19. As a result, electronic nose devices may be fitted with a gas sensor array configured 333 for the detection of disease specific target gasses, SARS-CoV-2 and/or viral components such as 334 the amino acid receptor binding domains (RBDs) of Spike 1 (S1) glycoprotein of SARS-CoV-2, 335 and/or a plurality of inflammatory biomarkers emitted from the breath of an infected individual. 336 Infection Mechanism of SARS-CoV-2. 337 338 Studies have indicated that SARS-CoV-2 has four main structural proteins including, spike (S) 339 glycoproteins, a membrane (M), an envelope (E), and a nucleocapsid (N)[42-45]. These S 340 glycoproteins are embedded on the surface of the outer E portion of SARS-CoV-2 and have a

11 341 1273 amino-acid-long structure configured to connect with the human cell membrane protein 342 ACE2 during infection[44,46,47]. Studies have shown ACE2 to be located at various host sites such 343 as at the nasal epithelial cells, the heart, the esophagus, the kidneys, the bladder, and the ileum[48]. 344 The S glycoprotein is divided into subunit 1 (S1) and subunit 2 (S2)[49]. S1 is primarily in 345 communication with ACE2 during infection and has beta strands and comprises an N-terminal 346 domain (NTD), a first C-terminal domain (CTD1), a second C-terminal domain (CTD2), and a 347 third C-terminal domain (CTD3)[50]. SARS-CoV-2 binds to the host during infection, studies 348 have shown evidence that some viral proteins contribute to the suppression of cytokines, 349 interferon (IFN) - α and β of macrophages during immune response[51]. This immune response 350 delay corresponds to the peak of viral load detected in the upper respiratory tract of SARS-CoV- 351 2 infected patients at day 10 of infection[48,51]. This suppression of IFN - α and β may contribute 352 to this infection load peak. 353 354 Although other pathogens of disease may share a similar infection mechanism of SARS-CoV-2, 355 it is an important aspect of this research for the combination of more broadly known neurological 356 inflammatory biomarkers, such as NO, from a plurality of pathogens of disease to be 357 incorporated with the known breath pattern signature of COVID-19. The VOCs of COVID-19 358 currently consist of specific increases in acetone, isoprene, heptanal, propanol, propanal, 359 butanone, methylpent-2-enal, 2,4-octadiene, 1-chloroheptane, nonanal, octanal, and a decrease in 360 methanol[1,13,14]. In regards to diagnostic electronic nose detection of SARS-CoV-2 infection and 361 COVID-19 disease biomarkers, a breath pattern signature having a wider scope including 362 neurological inflammatory biomarkers is a strategy described in this paper to enhance the 363 accuracy of differentiating COVID-19 infection from other pathogens of disease. 364 365 Olfactory nerve 366 The cribriform plate separates the nasal cavity from the brain[52]. The goblet cells, secretory cells, 367 and ciliated cells of the nasal epithelium lines the nasal cavity and paranasal sinuses[53]. Olfactory 368 receptor neurons of the olfactory neural cells have a dendritic end portion extending through the 369 olfactory epithelium of the nasal cavity located opposite an end traversing through approximately 370 15-20 foramina or openings of the cribriform plate and are connected to the olfactory bulb[7,52,54]. 371 The olfactory bulb is positioned within the frontal lobe of the brain and the olfactory tract

12 372 traverses through the medial temporal lobe of the brain[55]. Next, the olfactory tracts divide at the 373 olfactory trigone into three channels. One of the channels is referred to as the lateral olfactory 374 stria which continues laterally, traversing the horizontal Sylvian cistern and ending at the medial 375 temporal lobe approximately at the uncus[56]. It has been shown that an infection route of SARS- 376 CoV-2 is through the nasal cavity in communication with the medial temporal lobe of the brain 377 by the olfactory bulb[57]. 378 379 Studies have confirmed that viral particles up to approximately 100 nm can enter the CNS 380 through the olfactory epithelium[7]. Interestingly, clinical investigations of coronavirus-specific 381 viral particles were observed to range in size from 60-140 nm in size[58]. Thus, SARS-CoV-2 is 382 more than capable of infiltrating the CNS through olfactory receptor neurons and traversing the 383 cribriform plate. Studies have indicated infection of the olfactory bulb epithelium as being a 384 contributor in neurological symptoms in that SARS-CoV-2 damages ACE2 and then infiltrates 385 into the blood-brain barrier (BBB)[59]. The unique communication of SARS-CoV-2 with the 386 olfactory receptor neurons of the olfactory epithelium may serve the basis of diagnostic sensing 387 devices and/or canine sensing methods configured for pathogen detection and/or pathogen 388 classification of the nervous system. In younger children, the cribriform plate is undergoing 389 development and may have substantially larger foramina which could result in an increased 390 susceptibility to viral infiltration especially since SARS-CoV-2 is attracted to the ACE2 391 receptors within the nasal cavity. Studies have indicated that viral transmission along the 392 olfactory nerve is associated with histological lesions[60]. These histological lesions in CNS 393 disease have been known to induce an inflammatory response including, but not limited to, NO 394 mediated demyelination and could trigger COVID-19 induced pediatric inflammatory 395 multisystem syndrome[60]. Therefore, NO is a potential biomarker candidates for detecting 396 COVID-19 using an electronic nose device. 397 398 Nitric oxide expression in COVID-19 induced PMIS 399 A recently observed symptom of COVID-19 is anosmia or loss of smell. Studies have shown an 400 increased ACE2 expression in SARS-CoV-2 infection at the location of nasal epithelial cells[41,53]. 401 Because the S1 proteins of SARS-CoV-2 bind to ACE2 for infection to occur, a strong 402 correlation exists between peripheral nerve injury including virus induced olfactory loss and/or

13 403 injury stemming from viral communication at the olfactory epithelium, resulting in the anosmia 404 symptom of COVID-19. Studies have shown that demyelinating diseases have an increased 405 inflammatory response compared to non-neurological diseases. Hence, this atrophy of the 406 olfactory bulb and olfactory tract might be correlated with demyelination, which could trigger an 407 inflammatory response with potentially fatal outcomes in pediatric patients. This inflammatory 408 response induced from COVID-19 can be detected through VOC brain breath analysis with an 409 electronic nose device. 410 411 Although, the majority of anosmia observations reported in COVID-19 diagnosed cases appear 412 to be caused from mucosal congestion, peripheral nerve injury could explain anosmia in 413 asymptomatic COVID-19 patients not having mucosal congestion[61]. In particular, as a result of 414 the inflammatory response of SARS-CoV-2 connecting with ACE-2 at the olfactory epithelium, 415 macrophages and microglia generate cellular waste such as, inflammatory cytokines and NO. 416 These byproducts contribute to demyelination and axonal loss[62]. NO has been shown to damage 417 mitochondria of axons[63], resulting in a reduction of the enzyme Na+/K+ ATPase, a decrease in 418 ATP supply, and an increase in the amount of cellular Na ATP production is necessary for 419 nervous tissue to maintain stable resting potentials for propagation of electrical signals along the 420 length of nerve fibers. When ATP undergoes hydrolysis and is converted into adenosine 421 diphosphate (ADP) and phosphate, the Na+/K+ pumps become activated again to, thereby, 422 reversing the influx of Na+ within the cell[64]. There may be a correlation between anosmia as a 423 symptom of COVID-19 with the NO mediated injury to the olfactory receptor neurons. 424 425 Although the epithelium of the paranasal sinuses produces NO even without viral infection, 426 macrophages generate excessive NO as a result of proinflammatory cytokine activation during 427 viral infection defense[65,66]. When NO and oxygen free radicals, such as the hydroxyl radical

428 (OH) and superoxide (O2-) are in communication with each other from some viral lung infections,

429 toxic byproducts such as peroxynitrite (ONOO-) and NOx (NO2 and N2O3) are seen to contribute 430 to the mechanism of viral pneumonia resulting in tissue injuries to the lungs[66,67]. Further, 431 scientific evidence has indicated measurable amounts of NO in nasally exhaled air[68]. Thus, 432 excessive levels of NO may be a biomarker of inflammation especially when a collects 433 breath samples from nasally exhaled air as opposed to orally exhaled air.

14 434 435 CONCLUSION 436 Electronic nose technology configured to detect exhaled VOC biomarkers from a breath sample 437 have been observed as a successful approach to noninvasive diagnosis demonstrated in clinical 438 trials for breast cancer, colorectal cancer, lung cancer, liver disease, infectious disease such as 439 influenza and inflammatory lung disease[69]. Further, electronic nose technologies are currently 440 used in monitoring air quality, spoiled meats, and ripened fruit; however, its impact on emerging 441 pandemics, such as the respiratory illness pandemic in the form of COVID-19, is still under 442 investigation. Within the coronavirus family, seven viruses are currently known to infect humans, 443 including, NL63 and 229E from the alpha genus and OC43, HKU1, SARS-CoV, MERS-CoV, 444 and SARS-CoV-2 from the beta genus[70]. Of these seven viruses, studies have indicated infected 445 SARS-CoV-2 individuals have significant exhaled breath concentrations of ethanal, octanal, 446 acetone, butanone, methylpent-2-enal, 2,4-octadiene, 1-chloroheptane, nonanal, and a decrease in 447 methanol are capable of distinguishing between non-SARS-CoV-2 infected individuals and 448 SARS-CoV-2 infected individuals[1,13,20]. 449 450 The research results provide for the advancement of the configuration of electronic nose 451 technologies for the application of diagnostic breath analysis of not only SARS-CoV-2, but the 452 remaining six viruses of the coronavirus family, human respiratory viral , and nonviral diseases. 453 For example, studies have probed the role of conducting polymer type electronic nose devices 454 having an array comprising of 32 gas sensors. Although the low power consumption of 455 conducting polymer electronic noses have shown a good sensitivity and reproducibility at room 456 temperatures, their sensitivity to moisture, shortened sensor life, and susceptibility to damage 457 from polar analytes makes MOS-based electronic nose device a more attractive alternative for 458 effective detection of disease specific chemical biomarkers for diagnostic applications[71]. In 459 summary, breath analysis studies investigating the exhaled breath VOCs of individuals infected 460 with other pathogens of disease may provide clues to specific metabolic byproducts of infection 461 within the coronavirus family. Although it has been shown that SARS-CoV-2 may enter the 462 nasal cavity of a host and traverse the olfactory nerve to generate biomarkers of inflammation 463 within the brain, studies have determined that the circulatory system absorbs these biomarkers 464 which are then ultimately emitted from the lungs through an exhaled breath[71]. As our

15 465 understanding of infectious diseases, such as COVID-19 improves, we expect diagnostic 466 electronic nose devices to include a diversity of MOS-based gas sensors and electrochemical 467 sensors that can more accurately detect the exhaled breath biomarkers of disease as well as 468 changes associated with the pathogenesis of the disease. For diagnostic applications, electronic 469 nose devices are configured for detecting the unique breath pattern signature of a specific disease 470 from a complex mixture of VOCs with the use of a plurality of gas sensors having high 471 sensitivity to targeted biomarkers, thus, improving the disease specificity of the electronic nose 472 device[72]. 473 474 The highly transmissible nature of SARS-CoV-2 infection of COVID-19 disease has resulted in 475 the need for a rapid COVID-19 breath test, that may serve to diagnose both children and adults 476 around the world. Our research group has investigated the SARS-CoV-2 infection mechanisms 477 and its inflammatory response by the central nervous system by analyzing the symptoms of 478 COVID-19 and similar diseases while partnering with breath specialists that have collected, 479 measured, and validated VOCs from breath samples of COVID-19 positive individuals. 480 Substantial evidence suggests VOCs associated with the inflammatory response of a COVID-19 481 infection may include, but are not limited to, NO, alcohol, and acetone. These VOCs are capable 482 of being detected when they are exhaled through the breath of an individual. A rapid COVID-19 483 breath test with an electronic nose gas sensor array configured for noninvasive and rapid 484 detection of the breath pattern signature of COVID-19 may contribute to a substantial reduction 485 of the transmission rate of the global health threat of infection and result in a reduction of 486 COVID-19 related deaths. 487 488 DECLARATIONS 489 Acknowledgment 490 Since the beginning of the COVID-19 outbreak, it has caused an incredible amount of damage to 491 humans as well as economic repercussions. We thank The University of South Florida COVID- 492 19 Rapid Response Research Grants program and the efforts made by the University of South 493 Florida researchers for their contributions to high impact research and public service during this 494 global pandemic. 495

16 496 Authors’ contributions 497 Wrote the manuscript: Miller TC, Morgera SD, Saddow SE, Takshi A, Mullarkey M, Palm M. 498 Provided critical revision and final approval of the article: Miller TC, Morgera SD. 499 500 Availability of data and materials 501 Not applicable. 502 503 Financial support and sponsorship 504 This work is supported by a University of South Florida Pandemic Response Research Network 505 funding for COVID-19 to Dr. Salvatore D. Morgera. 506 507 Conflicts of interest 508 All authors declared that there are no conflicts of interest. 509 510 Ethical approval and consent to participate 511 Not applicable. 512 513 Consent for publication 514 Not applicable. 515 516 Copyright 517 © The Author(s) 2021. 518 519 REFERENCES 520 1. Ruszkiewicz DM, Sanders D, O'Brien R, et al. Diagnosis of COVID-19 by analysis of breath 521 with gas chromatography-ion mobility spectrometry - a feasibility study. EClinicalMedicine 522 2020;29:100609. doi: 10.1016/j.eclinm.2020.100609. PMID: 33134902; PMCID: PMC7585499. 523 2. Henderson B, Ruszkiewicz DM, Wilkinson M, et al. A benchmarking protocol for breath 524 analysis: the peppermint experiment. J Breath Res 2020;14:046008. doi: 10.1088/1752- 525 7163/aba130. PMID: 32604084.

17 526 3. Boesveldt S, Postma EM, Boak D, et al. Anosmia-a clinical review. Chem Senses 527 2017;42:513-23. doi: 10.1093/chemse/bjx025. PMID: 28531300; PMCID: PMC5863566. 528 4. Marinosci A, Landis BN, Calmy A. Possible link between anosmia and COVID-19: 529 out the truth. Eur Arch Otorhinolaryngol 2020;277:2149-50. doi: 10.1007/s00405-020-05966-0. 530 PMID: 32303881; PMCID: PMC7164410. 531 5. Palao M, Fernández-Díaz E, Gracia-Gil J, Romero-Sánchez CM, Díaz-Maroto I, Segura T. 532 Multiple sclerosis following SARS-CoV-2 infection. Mult Scler Relat Disord 2020;45:102377. 533 doi: 10.1016/j.msard.2020.102377. PMID: 32698095; PMCID: PMC7340057. 534 6. Wang L, Shen Y, Li M, et al. Clinical manifestations and evidence of neurological 535 involvement in 2019 novel coronavirus SARS-CoV-2: a systematic review and meta-analysis. J 536 Neurol 2020;267:2777-89. doi: 10.1007/s00415-020-09974-2. PMID: 32529575; PMCID: 537 PMC7288253. 538 7. van Riel D, Verdijk R, Kuiken T. The olfactory nerve: a shortcut for influenza and other viral 539 diseases into the central nervous system. J Pathol 2015;235:277-87. doi: 10.1002/path.4461. 540 PMID: 25294743. 541 8. Wang P. Combination of serological total antibody and RT-PCR test for detection of SARS- 542 COV-2 infections. J Virol Methods 2020;283:113919. doi: 10.1016/j.jviromet.2020.113919. 543 PMID: 32554043; PMCID: PMC7295513. 544 9. Rannan-Eliya RP, Wijemunige N, Gunawardana JRNA, et al. Increased intensity of PCR 545 testing reduced COVID-19 transmission within countries during the first pandemic wave. Health 546 Aff (Millwood) 2021;40:70-81. doi: 10.1377/hlthaff.2020.01409. PMID: 33264048. 547 10. Garg A, Ghoshal U, Patel SS, et al. Evaluation of seven commercial RT-PCR kits for 548 COVID-19 testing in pooled clinical specimens. J Med Virol 2021;93:2281-6. doi: 549 10.1002/jmv.26691. PMID: 33230819; PMCID: PMC7753435. 550 11. Kanji JN, Zelyas N, MacDonald C, et al. False negative rate of COVID-19 PCR testing: a 551 discordant testing analysis. Virol J 2021;18:13. doi: 10.1186/s12985-021-01489-0. PMID: 552 33422083; PMCID: PMC7794619. 553 12. Grover A, Lall B. A novel method for removing baseline drifts in multivariate chemical 554 sensor. IEEE Trans Instrum Meas 2020;69:7306-16. doi: 10.1109/TIM.2020.2976224. 555 13. Grassin-Delyle S, Roquencourt C, Moine P, et al; Garches COVID-19 Collaborative Group 556 RECORDS Collaborators and Exhalomics® Collaborators. Metabolomics of exhaled breath in

18 557 critically ill COVID-19 patients: a pilot study. EBioMedicine 2021;63:103154. doi: 558 10.1016/j.ebiom.2020.103154. PMID: 33279860; PMCID: PMC7714658. 559 14. Giovannini G, Haick H, Garoli D. Detecting COVID-19 from Breath: a game changer for a 560 big challenge. ACS Sens 2021;6:1408-17. doi: 10.1021/acssensors.1c00312. PMID: 33825440; 561 PMCID: PMC8043202. 562 15. Farraia MV, Cavaleiro Rufo J, Paciência I, Mendes F, Delgado L, Moreira A. The electronic 563 nose technology in clinical diagnosis: a systematic review. Porto Biomed J 2019;4:e42. doi: 564 10.1097/j.pbj.0000000000000042. PMID: 31930178; PMCID: PMC6924976. 565 16. Dragonieri S, Pennazza G, Carratu P, Resta O. Electronic nose technology in respiratory 566 diseases. Lung 2017;195:157-65. doi: 10.1007/s00408-017-9987-3. PMID: 28238110. 567 17. Licht JC, Grasemann H. Potential of the electronic nose for the detection of respiratory 568 diseases with and without Infection. Int J Mol Sci 2020;21:9416. doi: 10.3390/ijms21249416. 569 PMID: 33321951; PMCID: PMC7763696. 570 18. Dragonieri S, Schot R, Mertens BJ, et al. An electronic nose in the discrimination of patients 571 with asthma and controls. J Allergy Clin Immunol 2007;120:856-62. doi: 572 10.1016/j.jaci.2007.05.043. PMID: 17658592. 573 19. Tenero L, Sandri M, Piazza M, Paiola G, Zaffanello M, Piacentini G. Electronic nose in 574 discrimination of children with uncontrolled asthma. J Breath Res 2020;14:046003. doi: 575 10.1088/1752-7163/ab9ab0. PMID: 32512553. 576 20. Wilson AD, Baietto M. Advances in electronic-nose technologies developed for biomedical 577 applications. Sensors (Basel) 2011;11:1105-76. doi: 10.3390/s110101105. PMID: 22346620; 578 PMCID: PMC3274093. 579 21. Kalidoss R, Umapathy S. An overview on the exponential growth of non-invasive diagnosis 580 of diabetes mellitus from exhaled breath by nanostructured metal oxide Chemi-resistive gas 581 sensors and μ-preconcentrator. Biomed Microdevices 2019;22:2. doi: 10.1007/s10544-019-0448- 582 z. PMID: 31797133. 583 22. Miller TC, Morgera SD, Saddow SE, Takshi A, Palm M. Electronic nose with detection 584 method for alcohol, acetone, and carbon monoxide in coronavirus disease 2019 breath simulation 585 model. IEEE Sensors J 2021;1-9. doi: 10.1109/JSEN.2021.3076102. 586 23. Marco S, Gutierrez-galvez A. Signal and data processing for and chemical 587 sensing: a review. IEEE Sensors J 2012;12:3189-214. doi:10.1109/JSEN.2012.2192920.

19 588 24. Monroy JG, González-Jiménez J, Blanco JL. Overcoming the slow recovery of MOX gas 589 sensors through a system modeling approach. Sensors (Basel) 2012;12:13664-80. doi: 590 10.3390/s121013664. PMID: 23202015; PMCID: PMC3545586. 591 25. Korotcenkov G, Cho B. Instability of metal oxide-based conductometric gas sensors and 592 approaches to stability improvement (short survey). Sens Actuators B Chem 2011;156:527-38. 593 doi:10.1016/j.snb.2011.02.024. 594 26. Sharma RK, Chan PC, Tang Z, Yan G, Hsing I, Sin JK. Investigation of stability and 595 reliability of tin oxide thin-film for integrated micro-machined gas sensor devices. Sens 596 Actuators B Chem 2001;81:9-16. doi: 10.1016/S0925-4005(01)00920-0. 597 27. Di Natale C, Martinelli E, D’amico A. Counteraction of environmental disturbances of 598 electronic nose data by independent component analysis. Sens Actuators B Chem 2002;82:158- 599 65. DOI: 10.1016/S0925-4005(01)01001-2. 600 28. Wold S, Antti H, Lindgren F, Öhman J. Orthogonal signal correction of near-infrared spectra. 601 Chemom. Intell. Lab. Syst 1998;44:175-85. doi: 10.1016/S0169-7439(98)00109-9. 602 29. Hines E, Llobet E, Gardner J. Electronic noses: a review of signal processing techniques. IEE 603 Proc, Circuits Devices Syst 1999;146:297. doi: 10.1049/ip-cds:19990670. 604 30. Gardner JW, Bartlett PN. Electronic noses. Principles and applications. Meas Sci Technol 605 2000;11:1087. doi: 10.1088/0957-0233/11/7/702. 606 31. Llobet E, Brezmes J, Ionescu R, et al. Wavelet transform and fuzzy ARTMAP-based pattern 607 recognition for fast gas identification using a micro-hotplate gas sensor. Sens Actuators B Chem 608 2002;83:238-44. doi:10.1016/S0925-4005(01)01047-4. 609 32. Zuppa M, Distante C, Persaud KC, Siciliano P. Recovery of drifting sensor responses by 610 means of DWT analysis. Sens Actuators B Chem 2007;120:411-6. doi:10.1016/j.snb.2006.02.049. 611 33. Ziyatdinov A, Marco S, Chaudry A, Persaud K, Caminal P, Perera A. Drift compensation of 612 gas sensor array data by common principal component analysis. Sens Actuators B Chem 613 2010;146:460-5. doi:10.1016/j.snb.2009.11.034. 614 34. Marco S, Ortega A, Pardo A, Samitier J. Gas identification with tin oxide sensor array and 615 self-organizing maps: adaptive correction of sensor drifts. IEEE Trans Instrum Meas 1998;47: 616 316-21. doi: 10.1109/19.728841. 617 35. Fenske JD, Paulson SE. Human breath emissions of VOCs. J Air Waste Manag Assoc 618 1999;49:594-8. doi: 10.1080/10473289.1999.10463831. PMID: 10352577.

20 619 36. Artursson T, Eklöv T, Lundström I, Mårtensson P, Sjöström M, Holmberg M. Drift 620 correction for gas sensors using multivariate methods. J Chemom 2000;14:711-23. doi: 621 10.1002/1099-128X(200009/12)14:5/6<711::AID-CEM607>3.0.CO;2-4. 622 37. Ding H, Liu JH, Shen ZR. Drift reduction of gas sensor by wavelet and principal component 623 analysis. Sens Actuators B Chem 2003;96:354-63. doi: 10.1016/S0925-4005(03)00569-0. 624 38. Laref R, Ahmadou D, Losson E, Siadat M. Orthogonal signal correction to improve stability 625 regression model in gas sensor systems. J Sens 2017;2017:1-8. doi:10.1155/2017/9851406. 626 39. Ahmadou D, Laref R, Losson E, Siadat M. Reduction of drift impact in gas sensor response 627 to improve quantitative odor analysis. In: Proceedings of the IEEE International Conference on 628 Industrial Technology; 2017 Mar 22-25; Toronto, ON, Canada. 2017. p. 928-33. 629 40. Bossuyt PM, Reitsma JB, Bruns DE, et al; STARD Group. STARD 2015: an updated list of 630 essential items for reporting diagnostic accuracy studies. BMJ 2015;351:h5527. doi: 631 10.1136/bmj.h5527. PMID: 26511519; PMCID: PMC4623764. 632 41. Moons KG, Altman DG, Reitsma JB, et al. Transparent reporting of a multivariable 633 prediction model for individual prognosis or diagnosis (TRIPOD): explanation and elaboration. 634 Ann Intern Med 2015;162:W1-73. doi: 10.7326/M14-0698. PMID: 25560730. 635 42. Seo G, Lee G, Kim MJ, et al. Rapid detection of COVID-19 causative virus (SARS-CoV-2) 636 in human nasopharyngeal swab specimens using field-effect transistor-based biosensor. ACS 637 Nano 2020;14:5135-42. doi: 10.1021/acsnano.0c02823. PMID: 32293168; PMCID: 638 PMC7172500. 639 43. Gao L, Ding YS, Dai H, Huang ZD, Shao SH. A novel fingerprint map of SARS-CoV with 640 visualization analysis. Third International Conference on Image and Graphics (ICIG'04), 2004, 641 pp. 226-9, doi: 10.1109/ICIG.2004.18. 642 44. Wang Q, Zhang Y, Wu L, et al. Structural and functional basis of SARS-CoV-2 entry by 643 using human ACE2. Cell 2020;181:894-904.e9. doi: 10.1016/j.cell.2020.03.045. PMID: 644 32275855; PMCID: PMC7144619. 645 45. Hoffmann M, Hofmann-winkler H, Pöhlmann S. Priming time: how cellular proteases arm 646 coronavirus spike proteins. In: Böttcher-friebertshäuser E, Garten W, Klenk HD, editors. 647 Activation of Viruses by Host Proteases. Cham: Springer International Publishing; 2018. pp. 71- 648 98. doi:10.1007/978-3-319-75474-1_4.

21 649 46. Ortega JT, Serrano ML, Pujol FH, Rangel HR. Role of changes in SARS-CoV-2 spike 650 protein in the interaction with the human ACE2 receptor: an in silico analysis. EXCLI J 651 2020;19:410-7. doi: 10.17179/excli2020-1167. PMID: 32210742; PMCID: PMC7081066. 652 47. Walls AC, Park YJ, Tortorici MA, Wall A, McGuire AT, Veesler D. Structure, function, and 653 antigenicity of the SARS-CoV-2 spike glycoprotein. Cell 2020;181:281-92.e6. doi: 654 10.1016/j.cell.2020.02.058. PMID: 32155444; PMCID: PMC7102599. 655 48. Dong D, Tang Z, Wang S, et al. The Role of Imaging in the Detection and Management of 656 COVID-19: a review. IEEE Rev Biomed Eng 2021;14:16-29. doi: 10.1109/RBME.2020.2990959. 657 49. Cascella M, Rajnik M, Cuomo A, Dulebohn SC, Di Napoli R. Features, evaluation and 658 treatment coronavirus (COVID-19). StatPearls [Internet] StatPearls Publishing. 2020. 659 50. Gui M, Song W, Zhou H, et al. Cryo-electron microscopy structures of the SARS-CoV spike 660 glycoprotein reveal a prerequisite conformational state for receptor binding. Cell Res 661 2017;27:119-29. doi: 10.1038/cr.2016.152. PMID: 28008928; PMCID: PMC5223232. 662 51. Cheung CY, Poon LL, Ng IH, et al. Cytokine responses in severe acute respiratory syndrome 663 coronavirus-infected macrophages in vitro: possible relevance to pathogenesis. J Virol 664 2005;79:7819-26. doi: 10.1128/JVI.79.12.7819-7826.2005. PMID: 15919935; PMCID: 665 PMC1143636. 666 52. Doty RL, Kamath V. The influences of age on olfaction: a review. Front Psychol 2014;5:20. 667 doi: 10.3389/fpsyg.2014.00020. PMID: 24570664; PMCID: PMC3916729. 668 53. Sungnak W, Huang N, Bécavin C, et al; HCA Lung Biological Network. SARS-CoV-2 entry 669 factors are highly expressed in nasal epithelial cells together with innate immune genes. Nat Med 670 2020;26:681-7. doi: 10.1038/s41591-020-0868-6. PMID: 32327758. 671 54. Wenig BM. Undifferentiated malignant neoplasms of the sinonasal tract. Arch Pathol Lab 672 Med 2009;133:699-712. doi: 10.5858/133.5.699. PMID: 19415944. 673 55. Olfactory Pathways and Limbic System. pp. 903-4. 674 http://www.neuroanatomy.wisc.edu/coursebook/neuro3(2).pdf 675 56. Borden NM, Forseen SE, Stefan C. Imaging anatomy of the human brain a comprehensive 676 atlas including adjacent structures. New York: Demos Medical Publishing. 2016. 677 57. Montalvan V, Lee J, Bueso T, De Toledo J, Rivas K. Neurological manifestations of 678 COVID-19 and other coronavirus infections: a systematic review. Clin Neurol Neurosurg

22 679 2020;194:105921. doi: 10.1016/j.clineuro.2020.105921. PMID: 32422545; PMCID: 680 PMC7227498. 681 58. Kim JM, Chung YS, Jo HJ, et al. Identification of coronavirus isolated from a patient in 682 Korea with COVID-19. Osong Public Health Res Perspect 2020;11:3-7. doi: 683 10.24171/j.phrp.2020.11.1.02. PMID: 32149036; PMCID: PMC7045880. 684 59. Hernández Ramos FJ, Palomino García A, Jiménez Hernández MD. Neurology during the 685 pandemic. Is COVID-19 changing the organisation of neurology departments? Neurologia (Engl 686 Ed) 2020;35:269-71. doi: 10.1016/j.nrl.2020.04.009. PMID: 32364118; PMCID: PMC7183930. 687 60. Ryan MA, Zhou H, Buehler MG, et al. Monitoring space shuttle air quality using the Jet 688 Propulsion Laboratory electronic nose. IEEE Sens J 2004;4:337-47. doi: 689 10.1109/jsen.2004.827275. PMID: 15875356. 690 61. Galougahi MK, Ghorbani J, Bakhshayeshkaram M, Naeini AS, Haseli S. Olfactory bulb 691 magnetic resonance imaging in SARS-CoV-2-induced anosmia: the first report. Acad Radiol 692 2020;27:892-3. doi: 10.1016/j.acra.2020.04.002. PMID: 32295727; PMCID: PMC7151240. 693 62. Caruso D, Pesaresi M, Maschi O, Giatti S, Garcia-Segura LM, Melcangi RC. Effect of short- 694 and long-term gonadectomy on neuroactive steroid levels in the central and peripheral nervous 695 system of male and female rats. J Neuroendocrinol 2010;22:1137-47. doi: 10.1111/j.1365- 696 2826.2010.02064.x. PMID: 20819120. 697 63. Waxman SG. Axonal conduction and injury in multiple sclerosis: the role of sodium channels. 698 Nat Rev Neurosci 2006;7:932-41. doi: 10.1038/nrn2023. PMID: 17115075. 699 64. Miller TC. Brain-computer interface based on magnetic particle imaging for diagnostic and 700 neurological rehabilitation in multiple sclerosis. In: Proceedings of the 8th International Winter 701 Conference on Brain-Computer Interface (BCI), Brain-Computer Interface (BCI); 2020 Feb 26- 702 28; Gangwon, Korea (South). 2020. p. 1-6. 703 65. Stefano GB, Esch T, Kream RM. Potential immunoregulatory and antiviral/SARS-CoV-2 704 activities of nitric oxide. Med Sci Monit 2020;26:e925679. doi: 10.12659/MSM.925679. PMID: 705 32454510; PMCID: PMC7271680. 706 66. Babizhayev MA, Deyev AI. Management of the virulent influenza virus infection by oral 707 formulation of nonhydrolized carnosine and isopeptide of carnosine attenuating proinflammatory 708 cytokine-induced nitric oxide production. Am J Ther 2012;19:e25-47. doi: 709 10.1097/MJT.0b013e3181dcf589. PMID: 20841992.

23 710 67. Akaike T, Suga M, Maeda H. Free radicals in viral pathogenesis: molecular mechanisms 711 involving superoxide and NO. Proc Soc Exp Biol Med 1998;217:64-73. doi: 10.3181/00379727- 712 217-44206. PMID: 9421208. 713 68. Lundberg JO. Airborne nitric oxide: inflammatory marker and aerocrine messenger in man. 714 Acta Physiol Scand Suppl 1996;633:1-27. PMID: 8876755. 715 69. Lourenço C, Turner C. Breath analysis in disease diagnosis: methodological considerations 716 and applications. Metabolites 2014;4:465-98. doi: 10.3390/metabo4020465. PMID: 24957037; 717 PMCID: PMC4101517. 718 70. Abdelrahman Z, Li M, Wang X. Comparative review of SARS-CoV-2, SARS-CoV, MERS- 719 CoV, and influenza a respiratory viruses. Front Immunol 2020;11:552909. doi: 720 10.3389/fimmu.2020.552909. PMID: 33013925; PMCID: PMC7516028. 721 71. Wilson AD. Advances in electronic-nose technologies for the detection of volatile biomarker 722 metabolites in the human breath. Metabolites 2015;5:140-63. doi: 10.3390/metabo5010140. 723 PMID: 25738426; PMCID: PMC4381294. 724 72. Wilson AD. Biomarker metabolite signatures pave the way for electronic-nose applications 725 in early clinical disease diagnoses. Curr Metabolom 2017;5:90-101. doi: 726 10.2174/2213235X04666160728161251.

24