Recent advances in RNA sample preparation techniques for the detection of SARS-CoV-2 in saliva and gargle

Molecular detection of SARS-CoV-2 in gargle and saliva complements the standard analysis of nasopharyngeal swabs (NPS) specimens. Although gargle and saliva specimens can be readily obtained non-invasively, appropriate collection and processing of gargle and saliva specimens are critical to the accuracy and sensitivity of the overall analytical method. This review highlights challenges and recent advances in the treatment of gargle and saliva samples for subsequent analysis using reverse transcription polymerase chain reaction (RT-PCR) and isothermal amplification techniques. Important considerations include appropriate collection of gargle and saliva samples, on-site inactivation of viruses in the sample, preservation of viral RNA, extraction and concentration of viral RNA, removal of substances that inhibit nucleic acid amplification reactions, and the compatibility of sample treatment protocols with the subsequent nucleic acid amplification and detection techniques. The principles and approaches discussed in this review are applicable to molecular detection of other microbial pathogens.


Introduction
Analytical techniques have played an important role in clinical diagnosis, community surveillance, and control and mitigation of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) infection as the SARS-CoV-2 virus continues to evolve and spread [1]. Emerging new variants can escape the defense of neutralizing antibodies produced by vaccines and infections [2,3]. The currently circulating Omicron is the dominant variant worldwide, and it contains many sub-variants. Omicron sub-variants have continued to evolve and accumulate mutations, resulting in multiple descendants with increased ability of immune escape [1]. Despite the development of new vaccines such as the bivalent mRNA vaccines, it is still possible that new variants capable of evading immunity from new vaccines will emerge [4]. The COVID-19 pandemic highlights the need for better preparedness, including rapid and accessible methods for the detection of viral infection.
Molecular detection of SARS-CoV-2 RNA has primarily relied on reverse transcription polymerase chain reaction (RT-PCR) and isothermal amplification techniques, such as loop-mediated amplification (LAMP) and recombinase polymerase amplification (RPA), and the incorporation of CRISPR technology with these techniques. Extensive research has advanced nucleic acid amplification techniques, improved analytical sensitivity and specificity, and explored the point-of-care potential of new amplification and detection techniques. However, research focusing on sampling and sample preparation before the nucleic acid amplification and detection steps is very limited. This review aims to fill this important gap, discuss current sample treatment techniques, and highlight important considerations of specimen collection and nucleic acid extraction. Ultimately, the accuracy and reliability of analytical results require the validity of the entire analytical process, from specimen collection and treatment to signal amplification and detection.
The most commonly collected specimens for the diagnosis of SARS-CoV-2 infection are nasopharyngeal swabs (NPS) [5]. Appropriate collection of NPS specimens often requires trained healthcare professionals. Although NPS specimens can be selfcollected, the quality of NPS specimens is highly variable and depends on sampling skills. NPS sampling could cause discomfort, coughing and even bleeding in patients with bleeding disorders [6]. Inadequate sampling of viral components on NPS could cause false negative results [7].
Gargle and saliva samples are suitable alternatives to NPS and have been recently used for the detection of SARS-CoV-2 (Tables S1eS4). SARS-CoV-2 can enter saliva through the oral epithelial mucosa, salivary gland secretions, the gingival crevicular fluid, blood circulation, and sputum and respiratory secretions from the lower and upper respiratory tracts [8e11]. In addition, detection of SARS-CoV-2 in the early stages of infection is more sensitive from the analysis of saliva than of NPS specimens [12]. SARS-CoV-2 RNA can be stable in saliva for up to 25 days at room temperature [13]. Gargle specimens are typically obtained by gargling with saline or water [7], which is convenient for those who have difficulty generating enough saliva. Several studies have analyzed the overall agreement as well as clinical sensitivity of using saliva relative to using NPS. The overall concordance of saliva with NPS ranges from 89.7% to 99% and the overall clinical sensitivity ranges from 83% to 88% [8,14e22].
Gargle and saliva sampling has several advantages: (1) the method of collection is non-invasive; (2) convenient self-collection at home avoids crowding testing spaces and frees up time of health professionals; (3) the competing demand for swabs, reagents (e.g., viral transport media), and personal protective equipment required by healthcare workers for NPS sample collection is alleviated; (4) repeated collection of multiple gargle and saliva specimens is acceptable and useful for assessing changes of viral concentration over time in infected individuals.
The use of gargle and saliva samples for SARS-CoV-2 detection has encountered several challenges. First, infectious viruses collected in gargle and saliva pose a transmission risk to those who deliver, handle, and analyze gargle and saliva samples. On-site inactivation of virus-containing gargle and saliva samples during self-collection is necessary to avoid potential transmission risks. Second, the released viral RNA is prone to digestion by enzymes present in gargle and saliva samples. Techniques must maintain the stability of released viral RNA during the inactivation and treatment of the gargle and saliva samples. Third, gargle and saliva samples contain viscous materials, which vary with individuals and sampling protocols. The viscous sample matrix can hinder downstream processing, affect pipetting, and lower the efficiency of RNA extraction [7,23]. Commercial RNA extraction kits widely used for NPS samples may not be suitable for gargle and saliva samples. Fourth, collection and processing methods for saliva and gargle are not standardized. Conflicting results have been reported regarding comparisons of viral loads between NPS and saliva. Chau et al. [24] and Jamal et al. [25] reported similar amounts of viral RNA detected in NPS and saliva from symptomatic patients. Some studies reported that viral load in saliva was higher than in NPS [5,26e28], whereas other studies showed that viral load in saliva was lower than in NPS and that the detection of SARS-CoV-2 in saliva gave higher false negative rates [6,29e31]. These conflicting results could be due to both biological differences and analytical discrepancy. Differences in sampling procedures, RNA extraction and concentration approaches, and RNA detection methods could lead to analytical discrepancy and should be critically evaluated. In this review, we focus on discussing the collection of gargle and saliva samples, inactivation of viruses, release and preservation of viral RNA, extraction and/or concentration of viral RNA, and the subsequent RT-PCR or isothermal amplification techniques.

Collection of saliva and gargle samples
Procedures for collecting saliva and gargle samples differ between studies. Different studies employed varying abstention times (15e60 min) of patients from eating, drinking, chewing gum, or smoking prior to saliva and gargle collection [23,32]. A longer abstention time before sample collection allows individuals to accumulate more viruses in the mouth. Thus, in several studies, individuals were requested to produce a saliva sample in the morning right after waking up [27,33,34], which can lead to saliva samples with higher viral RNA concentrations.
Five self-collection methods have been commonly used for saliva collection in many studies ( Fig. 1, Tables S1eS4): (1) repeatedly spitting (200 mL ̶ 5 mL) into a container [35], avoiding mucus and sputum; (2) swirling saliva in the mouth for 30 s to 2 min and then repeatedly spitting into a container (200 mL ̶ 4 mL) [23,36,37]; (3) coughing up deep throat saliva (and potentially sputum) (~1 mL) and then spitting into a container [23,38,39]; (4) drooling saliva (0.5 ̶ 2 mL) into a container [23,40e42]; (5) chewing a cotton swab for about 1 min to stimulate salivation or swabbing the back of tongue and buccal mucosa, to coat the swab with saliva, and then keeping the swab in a sterile tube with medium or buffer [43,44]. Saliva collected by spitting or coughing followed by spitting is generally thicker and more difficult to process [23]. Additional steps are required to reduce viscosity; this increases the risk of cross-contamination and reduces the viral concentration of the sample [23]. Clear saliva samples collected by passive drooling or chewing on parafilm exhibited high clinical sensitivities for the detection of SARS-CoV-2 ranging from 87.3% to 95% relative to NPS [44,45]. Passively collected saliva samples without any forceful coughing had good clinical sensitivity, ranging from 85.7% to 98.6% in asymptomatic and mildly symptomatic patients [32]. The supernatant of saliva was found to contain higher SARS-CoV-2 RNA than saliva sediment and resuspended saliva [46]. In general, clear saliva is the best saliva sample. Saliva collected with swabs or cotton pads contained less viral RNA than saliva collected by either . This may be attributed to two reasons: (1) using swabs or chewing cotton pads may not collect all of the saliva present in the mouth; (2) salivary swabs or pads containing saliva samples are usually suspended in buffer [44] or centrifuged [43] to release saliva and viruses, which can lower the viral concentration because viruses may not be fully released. In addition, the buffer used for suspending swabs or pads can dilute the viral concentration in samples. Therefore, passive drooling is the best option. For gargle collection, different gargle media, volumes of gargling solution, and durations of gargling were used in different studies (Tables S1eS4). Patients swished and gargled varying volumes (3e10 mL) of saline [7,48e52], tap water [47,53], or natural spring water [54,55]) in their throat and mouth for varying durations (5e30 s), and the solutions were spat into sterile containers (Fig. 1). Regarding the gargle medium, studies using saline gargle demonstrated a slightly higher overall sensitivity (97%) than those using water gargle (86%) [56]. Regarding the gargle solution volume, studies using 5 mL of gargling solution reported a higher overall clinical sensitivity (92%) than those using > 5 mL of gargling solution (87%). Regarding the duration of gargling, studies using a gargling time of > 10 s reported higher overall clinical sensitivity (95% vs 86%) than those using 10 s of gargling [56]. Overall, saline, a gargling solution of 5 mL, and a gargling time of > 10 s are recommended for gargle collection. Gargle samples were found to have poorer clinical and analytical sensitivity when compared to saliva [34,37,47]. This is most likely due to the fact that gargle samples are diluted with gargling solution. A lower volume of gargle solution (1e2 mL) might be used to limit dilution of viral concentration in the gargle.
In most published studies, self-collected saliva and gargle were not inactivated prior to delivery, which could pose a transmission risk to those who deliver and analyze the samples. Thus, on-site viral inactivation is critical. Our research group formulated a viral inactivation and RNA preservation (VIP) buffer, enabling simultaneous viral inactivation and RNA preservation during on-site selfcollection of saliva and gargle samples [7]. The VIP buffer contained Triton X-100 (2.5%) and RLT lysis buffer (Qiagen) (main component: guanidinium isothiocyanate (GITC)). Triton X-100 (2.5%e10%) was able to destroy the envelopes of virions. The combination of Triton X-100 with guanidinium solution enhanced the inactivation of SARS-CoV-2 viruses [7]. This VIP buffer also prevented degradation of the released RNA. SARS-CoV-2 RNA (65 copies/200 mL sample) was stable in the VIP buffer for at least 3 weeks at room temperature [7].

Sample treatment and extraction of SARS-CoV-2 RNA
RNA extraction usually entails multiple steps, including (a) RNA release by breaking the host cell membrane as well as the envelope and capsid of virions, (b) RNA concentration and purification by removing non-RNA material such as proteins, DNA, and salts, and (c) recovery of RNA via precipitation or elution. Specimens can be treated using physical (e.g., heat, sonication, bead beating, or homogenization), enzymatic (e.g., proteinase K and lysozymes), or chemical procedures alone or in combination to release viral RNA from host cells and virions [57e59]. Chemical lysis involves the use of small-molecule reagents (e.g., chaotropic agents, reducing reagents, detergents, or organic solvents). The commonly used chaotropic agents are guanidinium hydrochloride (GuHCl) or GITC; reducing reagents are 2-mercaptoethanol, Tris (2-carboxyethyl) phosphine) (TCEP), and dithiothreitol (DTT); surfactants are sodium dodecyl sulfate (SDS), Triton X-100, and Tween 20; organic solvents are alcohols, ether, and chloroform [57,58]. Chaotropic agents are used to denature viral capsids and cell membrane proteins, as well as to inactivate RNases. Reducing reagents are utilized to break disulfide bonds of RNases and other proteins. Surfactants (detergents) are used to disrupt the lipid membrane and lipid viral envelope. Organic solvents can also cause cell disruption by permeating the cell walls and membranes. In addition, enzymes are used to assist chemicals to break/hydrolyze the cell wall and membrane, as well as virions, and to partially inactivate RNases. Reducing reagents, chaotropic agents, and proteinase K can reduce the viscosity of saliva samples through denaturing proteins [60,61]. In some studies, RNA-extraction-free methods, heat-based specimen lysis methods (Tables S1 and S2) and lysis approaches without the need of heating (Fig. 3) were developed and used to release SARS-CoV-2 RNA, and the resultant crude SARS-CoV-2 RNA was directly analyzed using RT-PCR or isothermal techniques [62,63]. These direct analysis methods aimed at reducing the time and cost of SARS-CoV-2 detection. Nevertheless, these methods can result in limited analytical sensitivity due to the remaining sample matrices in crude RNA and the lack of RNA preconcentration. In other studies, SARS-CoV-2 RNA was extracted using commercial reagents or kits. In this review, RNA extraction using commercial extraction kits will be referred to as "standard RNA extraction". The RNA preservation during the specimen lysis and RNA extraction is critical for subsequent RNA detection. The following section will go through tactics for preserving RNA integrity. Afterwards, RNAextraction-free techniques as well as RNA extraction methods will be discussed. 3.1. RNA stability during treatment of specimens RNA integrity is a key factor affecting the detection sensitivity of SARS-CoV-2 RNA by RT-PCR, reverse transcription loop-mediated isothermal amplification (RT-LAMP), or reverse transcription recombinase polymerase amplification (RT-RPA). Thus, maintaining RNA integrity during heat lysis or RNA extraction is crucial. Achieving this goal requires understanding of the factors that can affect RNA stability, as well as their remedies. RNA can be degraded under the following conditions. First, alkaline conditions hydrolyze the bases of RNA [64] [61,71]. TCEP is more widely used than DTT because it is more stable than DTT [61]. Strong denaturants such as GuHCl and GITC are used to further denature reduced RNases [69]. Additionally, proteinase K can partially inactivate RNases [72]. RNase contamination from the environment should be avoided. Work environments must be kept clean and decontaminated with products such as RNase AWAY (Thermo Scientific). Furthermore, 3.2. RNA-extraction-free approach: heat treatment of specimens, followed by direct analysis Saliva is known to have constituents that inhibit RT-PCR [73]. However, regular RNA extraction is time-consuming, expensive, and subject to supply chain bottlenecks, which limits the scalability of diagnostic tests for SARS-CoV-2. To overcome these limitations, researchers have developed several heat-based specimen lysis methods that do not require the extraction of RNA (Figs. 2 and 3, Tables S1 and S2). Heat lysis was used to achieve three objectives: (1) inactivate SARS-CoV-2 viruses, (2) release RNA and preserve it from degradation, and (3) disrupt inhibitors in samples. Abraham et al. [74] reported that heating a sample for 3 min at > 75 C is sufficient to inactivate any SARS-CoV-2 virions. Heat-based lysis methods with > 75 C temperature can offer safe sample processing and analysis. Heating can also break down viral particles and host cells, releasing RNA and degrading certain inhibitors [57]. To maximize RNA release, protect RNA from degradation, and degrade inhibitors, many studies utilized enzymatic and chemical treatments to complement heat lysis [45,75]. Because RT-PCR and isothermal amplification techniques have different tolerance to chemicals such as reducing reagents and chelating reagents [76e78], different heat lysis methods are compatible with the subsequent RT-PCR (Fig. 2, Table S1) and isothermal amplification techniques (Fig. 3, Table S2).

Heat treatment of specimens and direct RT-PCR analysis
3.2.1.1. Heating only. Ham et al. [79] heated saliva samples for 30 min at 95 C and directly analyzed the crude RNA solution using RT-PCR. The limit of detection (LOD) using this whole process was 5000 RNA copies/mL of simulated positive saliva sample [79]. White et al. [80] treated saliva samples using the same approach, except that the heat-processed samples were diluted with an equal volume of TBE-Tween 20 buffer prior to RT-PCR analysis. A low clinical sensitivity of 68% was observed [80]. The investigators confirmed that heating saliva samples at 95 C for 30 min did not completely denature all of the RNases and inhibitors present in the samples [80], which probably caused the low analytical sensitivity and clinical sensitivity reported in these two studies. Furthermore, in the presence of divalent cations, heating over an extended period could result in RNA degradation [61,64,65]. Therefore, the low analytical and clinical sensitivities of these two studies could be attributed to insufficient RNA release and disruption of RNases and inhibitors, as well as RNA degradation owing to heating.

Heating and homogenization.
To further eliminate the inhibitors while maximally releasing RNA with heating treatment, Sahajpal et al. [81] developed a SalivaSTAT protocol by adding an additional homogenization step after heating. In this protocol, saliva was heated to 95 C for 30 min and then homogenized at 4.5 m/s for 30 s using the Omni bead mill homogenizer. This protocol, combined with the subsequent RT-PCR analysis, had a LOD of 0.06e0.18 RNA copies/mL, which is much lower than that obtained by Ham et al. [79] through heating only (5000 copies/mL). The increased analytical sensitivity is attributed to homogenization, and homogenized saliva samples were detected as positive with significantly lower Ct values compared to the samples subjected to vortexing [81]. Homogenization in the SalivaSTAT protocol made the viscosity of saliva samples similar to that of water and reduced the inhibitory effect so that 20 mL of processed saliva was well tolerated in 30 mL of RT-PCR reaction. The SalivaSTAT protocol exhibited a clinical sensitivity of 95.0% for saliva samples when compared to a standard RNA extraction protocol. However, the Ct values were significantly higher with SalivaSTAT compared to those with a standard RNA extraction method. The higher Ct values may be due to two reasons: (1) lower amounts of viral RNA was input for RT-PCR because RNA was not concentrated during the SalivaSTAT process; (2) a small portion of RNA may be degraded during extended heating.
3.2.1.3. Proteinase K treatment followed by heating. Proteinase K can be helpful for release of SARS-CoV-2 RNA, protection of RNA, and reduction of inhibitors because it can inactivate some RNases and degrade proteins. Vogels et al. [75] combined proteinase K with heating to effectively treat saliva. Coined the SalivaDirect protocol, Vogels et al. [75] first treated saliva with proteinase K (2.5 mg/mL) for 1 min and then heated the sample (containing proteinase K) at 95 C for 5 min. This protocol received Emergency Use Authorization (EUA) from the U.S. Food and Drug Administration (FDA) [75], and was used in a hospital cohort study for SARS-CoV-2 testing. SalivaDirect provided a LOD of 6 virus copies/mL of the saliva sample, which is significantly lower than that of the heating-only approach (5000 copies/mL) [79], indicating teh advantage of the combined treatment using proteinase K and heating. This technique offered a clinical sensitivity of 97.1% for saliva samples and a clinical sensitivity of 94% for NPS samples. However, saliva processed with SalivaDirect exhibited higher Ct values than saliva processed using a standard RNA extraction. The higher Ct values (i.e., lower analytical sensitivity) are mostly caused by the inhibitors remaining in the processed saliva following SalivaDirect treatment [82]. Due to a lack of RNA preconcentration in the Sali-vaDirect protocol, the limited viral RNA input into RT-PCR also contributes to the lower analytical sensitivity. Two studies reported that saliva samples processed using Sali-vaDirect had clinical sensitivities of 88.2% and 91.3% compared to NPS samples processed with a standard RNA extraction [83,84]. These sensitivities are comparable to the 94% clinical sensitivity reported by Vogels et al. [75].
RNA sequencing results from saliva samples treated with Sali-vaDirect have significantly lower genome completeness (7.8% ± 33%) than those obtained using standard RNA extraction (89.3% ± 29%) [85], indicating that heat inactivation breaks RNA down into shorter fragments. This finding is in line with that of a previous study [86]. Lowering proteinase K concentration from 2.5 to 1.2 mg/mL in saliva treatment resulted in a slight decrease in analytical sensitivity (LOD: 9.4 RNA copies/mL) and clinical sensitivity (78%) [87].
Morais et al. [88] further evaluated the SalivaDirect protocol and found that extending the 95 C heating time from 5 min to 15 min considerably reduced Ct values, whereas extending heating beyond 15 min did not have a significant impact on Ct values [88]. Athough RT-PCR inhibition by sample matrix is reduced after heating saliva at 95 C for 15 min, inhibitors are not completely removed. Substances that are not thermally labile might be present and inhibit RT-PCR, albeit at a lower level [88].
3.2.1.4. Proteinase K and other chemical treatment followed by heating. Enzymatic and chemical treatment combined with heat lysis can facilitate the release of RNA from saliva while preventing its degradation and reducing inhibitors [45,88]. In a study, 1.25 mg/ mL of proteinase K, 0.5% of N-acetylcysteine (NAC), and 0.5% of Triton X-100 were added to saliva samples and incubated at 37 C for 10 min, followed by heating at 95 C for 5 min. NAC is a mucolytic agent that is used to reduce saliva viscosity, while 0.5% of Triton X-100 is used to rupture the envelope of SARS-CoV-2 [89] and is compatible with RT-PCR [86]. This lysis process and the subsequent RT-PCR allowed the detection of 2.5 RNA copies/mL of sample, with a clinical sensitivity of 95.7% compared to standard RNA extraction processes [89]. This study demonstrates that enzymatic and chemical treatment in combination with heat lysis can enhance detection sensitivity by facilitating RNA release and providing more effective RNA protection.
In summary, all of the heat-based specimen treatment techniques (Fig. 2) for saliva samples discussed can provide clinical sensitivity that is comparable to that of standard extraction protocols with NPS specimens, making large-scale testing more accessible and affordable. Based on published LOD values (Table 1,  Table S1), the effectiveness of the heat treatment methods is ranked as follows: heating plus homogenization > heating plus proteinase K, NAC, and Triton X-100 > heating plus proteinase K > heating alone. However, due to the absence of an RNA purification step to completely remove PCR inhibitors and a step to concentrate RNA, heat lysis procedures lead to a considerable decrease in analytical sensitivities and an increase in Ct values when compared to standard extraction approaches. Additionally, the inhibition level of processed saliva is highly affected by different heating conditions (heating time, heating with or without the addition of enzymes and chemicals). Finally, all of the heat lysis methods discussed are followed by direct RT-PCR detection and the saliva samples were collected in sterile tubes without preservatives, because RT-PCR is vulnerable to inhibitors such as preservatives [75,81].

Heat treatment of specimens and isothermal amplification and detection
Compared to RT-PCR, isothermal amplification techniques do not require a thermal cycler. Moreover, isothermal methods are more tolerant to some common inhibitory components in a crude sample. For example, LAMP is more tolerant than PCR to typical sample matrix inhibitors [90], and RPA is more tolerant than PCR to some known inhibitors and complicated sample matrices [91e93]. Researchers have developed several heat-based specimen lysis protocols that are compatible with downstream RT-LAMP and RT-RPA detection (Fig. 3, Table S2). The combination of heat lysis with isothermal detection enables rapid, sensitive, and repeatable testing directly from saliva without requiring the purification of RNA, which greatly simplifies SARS-CoV-2 detection and is more applicable to point-of-care diagnosis.
3.2.2.1. Heating only. Taki et al. [94] simply heated saliva samples at 95 C for 10 min and used 5 mL of treated saliva in 20 mL of RT-LAMP reaction volume. The clinical sensitivity of this assay was 47% relative to standard RNA extraction followed by RT-PCR in saliva samples. In comparison, the combination of standard RNA extraction methods with RT-LAMP showed 94% clinical sensitivity compared to standard RNA extraction followed by RT-PCR for saliva samples. This finding suggests that heat lysis, rather than RT-LAMP, is primarily responsible for the low clinical sensitivity (47%) produced by heat lysis in combination with RT-LAMP. Heating at 95 C for 10 min is insufficient to fully liberate RNA from particles, deactivate RNases, and eliminate the matrix inhibitory effects of saliva. This is consistent with a previous study reporting that heating saliva samples at 95 C for 30 min did not completely denature all of the RNases and inhibitors [80]. Wei et al. [95] developed the high-performance loop-mediated isothermal amplification (HP-LAMP) assay. They heated saliva samples at a 95 C for 5 min and then added 5 mL of the processed sample to the one-step HP-LAMP reaction mixture for a 500 mL total reaction volume. With saliva samples, HP-LAMP had a clinical sensitivity of >96% compared to a standard RNA extraction-RT-PCR protocol with NPS specimens [95]. The high clinical sensitivity in this study may be attributed to the huge reaction volume of the HP-LAMP application (500 mL), which dilutes the inhibitors present in the input sample lysed by heating alone.
3.2.2.2. Chemical treatment followed by heating. Chemical treatment has been combined with heat-based specimen lysis in order to improve the inactivation and disruption of virions while inactivating RNases. The treatment of TCEP and EDTA has been used together with heat-based salivary sample lysis [46,61]. Rabe et al. [61] treated saliva with 1 Â TCEP/EDTA (2.5 mM TCEP, 1 mM EDTA), which was then heated at 95 C for 7 min. The resulting crude sample was directly used in the RT-LAMP assay. Five mL of processed saliva could be tolerated in 20 mL of the RT-LAMP reaction. In this lysis protocol, TCEP was used to reduce saliva viscosity by denaturing proteins and to inactivate RNases. EDTA was used to chelate divalent metal cations to prevent RNA degradation during heating. The investigators found that the treatment of 1 Â TCEP/EDTA combined with heating is sufficient to inactivate RNase activity [61]. The LOD of this assay was 50 RNA copies/mL of saliva. The 1 Â TCEP/EDTA has been applied in many heat-based protocols with slight modifications. In one study [46], saliva samples were treated with 1 Â TCEP/EDTA and 0.2% Pluronic F-68, followed by heating at 95 C for 5 min. Then, the processed saliva was directly analyzed using RT-LAMP. The LOD of this assay was 39 virus copies/ mL of saliva. The inclusion of Pluronic F-68 in this protocol enabled more sensitive detection of SARS-CoV-2. Pluronic F-68 is a detergent that can rupture the envelope of the SARS-CoV-2 virus. Kundrod et al. [96] treated samples with 5 Â TCEP/EDTA and 200 mM GuHCl and heated samples at 95 C for 6 min, before RT-LAMP analysis. In this assay, the treatment of TCEP/EDTA and GuHCl prior to heating improved the efficiency of heat lysis. These chemicals are compatible with the RT-LAMP reaction. This assay displayed a LOD of~20 virus copies/mL of saliva. The analytical sensitivity is slightly higher than those obtained by using 1 Â TCEP/ EDTA alone or 1 Â TCEP/EDTA plus Pluronic F-68, suggesting that combining 5 Â TCEP/EDTA with GuHCl can slightly improve RNA release and reduce inhibitory effect. Overall, the treatment of TCEP/ EDTA combined with detergent or GuHCl prior to heat lysis can facilitate RNA release and inhibitor reduction.

3.2.2.3.
Treatment with proteinase K and other chemicals, followed by heating. Similar to RT-PCR, RT-LAMP is also compatible with proteinase K treatment followed by heating (Table S2). In a highthroughput CRISPR-Cas13 assay for SARS-CoV-2 [97], saliva was heated to 65 C for 6 min in the presence of proteinase K (2 mg/mL) and was further incubated at 98 C for 3 min. The assay of RT-LAMP amplification followed by CRISPR-mediated detection demonstrated a LOD of 5 virus copies/mL of saliva, which is comparable to that obtained by SalivaDirect plus RT-PCR [75]. The saliva samples tested by this assay showed 88% clinical sensitivity relative to NPS specimens analyzed by standard RNA extraction with RT-PCR [97]. Huang et al. [98] used RNAsecure (main component: DTT) and proteinase K (25 units/mL) to treat saliva at room temperature for 10 min, and then at 65 C for 10 min. The resulting mixture was heated at 100 C for 12 min. This protocol in conjunction with multiplex RT-LAMP could detect as few as 1.5 virus copies/mL of SARS-CoV-2 in saliva, indicating that the treatment of DTT and proteinase K combined with heating is effective. Cook et al. [99] utilized 1 Â TCEP/EDTA with 0.1 mg/mL proteinase K to treat saliva samples by vortexing, and then heated samples for 10 min at 95 C. Using this heat-based lysis protocol in conjunction with an RT-LAMP, they correctly detected two positive samples out of 1649 samples collected over the course of a 9-month voluntary SARS-CoV-2 testing program. These positive results were comfirmed by RT-PCR tests with a standard RNA extraction [99]. According to the results of these studies, the treatment of proteinase K alone or in combination with chemicals prior to heat lysis can improve the overall RNA release while also inactivating RNases and lowering inhibitors. It is worth noting that the heating after proteinase K treatment helped proteinase K deactivation. On the basis of the reported LOD values ( Table 1, Table S2), the effectiveness is in the following decreasing order: heating plus proteinase K and other chemicals (e.g., reducing agents) > heating plus proteinase K > heating plus chemicals (reducing and/or chelating agents). The effectiveness of heating alone cannot be ranked among other methods in this case, despite the fact that the HP-LAMP assay, in which heating alone was employed to treat specimens, produced a LOD of 1.4 virus copies/L. However, the LAMP reaction volume was 500 mL, which is cost-ineffective and not suitable for widespread use.

Integrated devices for heating specimens and isothermal detection.
Portable devices with integrated heat lysis and isothermal detection allow convenient, fast, and sensitive detection of SARS-CoV-2 in saliva specimens, and are promising for point-ofcare testing. Two integrated devices for the SARS-CoV-2 test in saliva have been reported. The first [100] is a minimally instrumented SHERLOCK (miSHERLOCK) device in which a saliva sample was treated with 10 mM DTT and 5 mM EGTA in a column with an integrated 4-mm polyethersulfone (PES) membrane (Millipore) at the bottom, followed by heating at 95 C in the chamber for 3 min. The processed saliva flows through the membrane by gravity and by capillary action of an absorbent cellulose underneath the membrane. The resulting PES membrane was directly put to the RT-RPA reaction, followed by CRISPR-mediated detection. The PES membrane contains 0.22-mm pores and is functionalized with a hydrophilic surface that serves as a porous matrix to capture and concentrate SARS-CoV-2 RNA. This miSHERLOCK device enabled notable RNA concentration, resulting in a sensitivity improvement. A LOD of 1.24 viral RNA copies/mL was achieved, which is one of the lowest among all techniques using heat treatment of saliva ( Table 1). The miSHERLOCK technology achieved 96% clinical sensitivity for the detection of SARS-CoV-2 in saliva samples, and it was comparable to that obtained concurrently using the standard RNA extraction and RT-PCR technique [100]. The other integrated device is SLIDE (saliva-based SARS-CoV-2 self-testing with RT-LAMP in a mobile device) [101]. In this device, the saliva was treated by heating at 95 C for 5 min, and 10 mL of treated saliva was used in 40 mL of RT-LAMP reaction. The heattreated saliva flows into the reaction chamber by capillary force, which could diminish the amount of debris entering the reaction chamber. The LOD was 5 viral RNA copies/mL of the saliva sample, which is comparable to the LOD (6 viral RNA copies/mL) of Saliva-Direct plus RT-PCR [75]. The high analytical sensitivity of SLIDE may in part be due to limiting the debris in the RT-LAMP reaction.
3.3. RNA-extraction-free approach without heating, followed by colorimetric assay using nanoparticles To further simplify the processing of specimens, two studies did not use heat treatment but used chemical lysis buffers to treat saliva samples. In the first study [62], the saliva sample, GITC lysis buffer (from Purelink RNA mini kit, Invitrogen), and RNase-free water were mixed together at a volume ratio of 2:1:2. Then 2 mL of the resulting solution was used in 10 mL of RT-LAMP reaction.
Plasmonic gold nanoparticles capped with antisense oligonucleotides (ASOs) were used as reporters for colorimetric detection [62]. The LOD of this assay was 10 viral RNA copies/mL of the artificial saliva sample, which is comparable to those of heat-based lysis methods, such as SalivaDirect plus RT-PCR and the integrated SLIDE method [73,99]. In the second study, saliva was mixed with lysis buffer (TE buffer supplemented with 0.11% SDS, 0.11% Triton X-100, and 0.58 mg/mL proteinase K), and then incubated for 20 min at room temperature. The incubation allowed the released RNA to bind to nanoparticles for generating colormetric readout. This method had a LOD of 2.1 Â 10 2 RNA copies/mL by visual inspection or 8 Â 10 1 RNA copies/mL by spectrophotometric measurement within 20 min [63]. Although the sensitivity of this assay is much lower than that of assays involving RNA amplification, it is better than those of antigen tests (LOD: 5 Â 10 3 copies/mL) [63].

Extraction of SARS-CoV-2 RNA from saliva and gargle samples
RNA-extraction-free approaches often provide low analytical sensitivity because they are unable to completely eliminate sample matrix inhibitors and concentrate RNA. Appropriate RNA extraction and preconcentration can enhance the overall analytical sensitivity. RNA extraction is commonly performed by using commercially available reagents or kits.

RNA release
Commercial RNA extraction kits usually contain various reagents, such as chaotropic agents, reducing agents, surfactants, and enzymes. Current commercial kits are primarily designed for extraction of RNA from nasopharyngeal swabs, but they are not intended for saliva or gargle samples. Although several studies extracted SARS-CoV-2 RNA from saliva and gargle samples using commercial kits, the RNA extraction efficiency was low. Saliva and gargle samples contain viscous matrices. The viscosity should be reduced first to efficiently extract RNA from these samples. In several studies, saliva or gargle was diluted with PBS or other types of buffers before RNA extraction. However, dilution of the sample also lowers the concentration of the target RNA [6,22,30,36]; Lopes et al. [6] observed that when saliva was 1:1 diluted with viral transport media, clinical sensitivity was decreased (66% vs. 73.3%). To reduce sample viscosity, several studies treated saliva samples with DTT [102,103], proteinase K [104], or homogenization [105].
Our research group formulated a viral inactivation and RNA preservation (VIP) buffer that is suitable for the extraction of RNA from both saliva and gargle samples. The VIP buffer is composed of 2-ME (1%), Triton X-100 (2.5%), proteinase K (170 ng/mL), glycogen (17 ng/mL), and RLT lysis buffer (main component: GITC, Qiagen). In the VIP buffer, 2-ME is used to break disulfide linkages in RNases and other proteins; GITC denatures proteins; and proteinase K digests RNases and viscous proteins in saliva and gargle samples. Triton X-100 breaks host cell membranes and viral envelopes. Glycogen serves as an RNA carrier to enhance the recovery of low amounts of RNA from samples. The VIP buffer has sufficient capability to inactivate viruses, reduce the viscosity of saliva and gargle samples, release RNA, protect RNA from degradation, and facilitate the extraction of low-abundance RNA [7].

Purification and concentration of RNA
After RNA is released, the next step is to separate RNA from any non-RNA materials, such as proteins, salts, DNA, and other interfering substances in saliva and gargle. The process of separation also concentrates RNA. Two basic methods for RNA extraction are phenol-chloroform and solid-phase extraction. Solid-phase extraction techniques, such as with silica columns and silicacoated magnetic beads, are now widely used in commercial RNA extraction kits. Beads coated with carboxyl groups have also been used to purify SARS-CoV-2 RNA from saliva samples. RNA extraction procedures that include solid-phase RNA purification steps have high analytical sensitivities, with LOD ranging from 0.02 to 0.13 RNA copies/mL of saliva samples (Tables S3 and S4). These reported LOD values are comparable to those achieved using Sali-vaSTART (LOD~0.06e0.18 RNA copies/mL of saliva samples) and are orders of magnitude better than those obtained with all other RNAextraction-free methods.

Phenolechloroform method of purification and concentration.
Phenol-chloroform method (phenol/chloroform in 1:1 vol ratio) is a conventional RNA separation process that removes proteins and DNA while leaving RNA in the aqueous phase. The RNA in the aqueous phase is precipitated from the supernatant by adding ethanol or isopropanol along with a high concentration of salt. The resulting RNA pellet is washed, dried, and finally dissolved with RNase-free water. During purification, phenol can denature proteins rapidly, while chloroform efficiently inhibits RNase activity. Thus, this phenol-chloroform mixture safeguards RNA during purification [106,107]. The phenol-chloroform extraction method is incorporated into TRIzol (Invitrogen), a commonly used commercial reagent for RNA extraction. This reagent comprises GITC, phenol, and other proprietary components to aid in the isolation of RNA. This reagent can be mixed with a saliva sample, followed by the addition of chloroform for separating RNA from DNA, proteins, and lipids. This method is time-consuming and labor-intensive. In addition, residual organic solvents, if remaining in the extracted RNA, can impact the downstream analysis. Nonetheless, this method is well established and cheaper than commercial kits. Abasiyanik et al. [108] reported that the TRIzol method had a relative efficacy of 17% compared with the QIAmp Viral RNA Mini Kit when extracting SARS-CoV-2 RNA from saliva. In contrast, using the TRIzol method, Paz et al. [109] observed a recovery of viral RNA ranging from 50% to 90% when extracting RNA from saliva, NPS, and oropharyngeal swabs (OPS). These findings imply that the extraction efficiency of the TRIzol method is greatly dependent on operator competence and sample type. TRIzol was also used in one study to extract RNA from paired saliva and NPS samples simultaneously. The TRIzol saliva testing was as sensitive as the TRIzol NPS testing [110]. In another study involving the TRIzol method, the investigators compared viral RNA concentration in saliva and gargle samples to NPS specimens. They found that the clinical sensitivity for saliva and gargle was 87.8% and 80.5%, respectively, when compared to NPS tested using the same method [111]. These results are in line with those reported in several other studies [34,37,47].

Solid-phase extraction of RNA
3.4.2.2.1. Purification and concentration of RNA using silica columns. Silica column-based RNA purification has been incorporated into many commercial RNA extraction kits. This method relies on the fact that chaotropic salts (such as GITC and GuHCl) can denature biomolecules by disrupting the hydration shell around them, allowing positively charged ions to form a salt bridge between the negatively charged silica and negatively charged RNA backbone at high salt concentrations [112,113]. A binding solution (lower pH and high salt concentration) usually contains GuHCl and/or GITC, Triton X-100, isopropanol, and a pH indicator. Column-based extraction typically consists of three steps: (1) The sample is introduced to the column, and the RNA is captured on the column; (2) the column is washed with a suitable wash buffer to remove unbound components; and (3) the captured RNA is eluted off from the column using a buffer or RNase-free water (Fig. 4A).
Many studies used commercial kits with silica-columns to extract SARS-CoV-2 RNA from saliva and gargle samples (Table S3). One study reported a LOD of~0.02 RNA copies/mL in saliva samples (Table 1) [114], much better than those obtained with RNAextraction-free approaches. The clinical sensitivities of saliva sample testing in reported studies (Table S3) range from 74% to 100% when compared to paired NPS tested utilizing the same RNA extraction and detection procedures (Table S3). These results demonstrate that saliva is as effective as NPS for detecting SARS-CoV-2. The slight differences in sensitivity between studies could be attributed to different sampling procedures, RNA extraction and detection methods, and study populations. The differences in sensitivity are also influenced by the various pretreatment procedures used prior to extraction (Table S3). One study isolated RNA from matched saline gargles and NPS from inpatients and compared the viral load in these two types of samples. They showed 100% clinical sensitivity in comparing gargle and the corresponding NPS samples [52]. The Ct values for gargle samples were slightly higher because of the dilution of RNA in the sample by saline.
3.4.2.2.2. Purification and concentration of RNA using silicacoated magnetic beads. Magnetic beads with silica surfaces are also commonly used to extract SARS-CoV-2 RNA from saliva and gargle samples. RNA is captured on the magnetic beads; the magnetic beads are collected using a simple magnet; and the sample matrix is discarded. After the magnetic beads are washed, the RNA is eluted with an elution buffer (Fig. 4B). Both the magnetic bead and silica column methods have achieved comparable analytical sensitivity (LOD~0.02e0.06 RNA copies/mL) with saliva and gargle samples ( Table 1, Table S4). Compared to using silica columns, the use of magnetic beads eliminates the need for repeated centrifugation or vacuum filtration. The procedure is simple, convenient, fast, and it may be used in batch processes. It is simple to implement and amenable for automation and high-throughput applications. The magnetic bead method is also appropriate for use in resource-limited settings.
Many researchers used silica-bead-mediated commercial RNA extraction kits in conjunction with RT-PCR or RT-LAMP to compare the viral load in paired saliva and NPS samples (Table S4). In most of these studies, clinical sensitivities vary from 78.9% to 95.1%, indicating the viral RNA levels in saliva samples are comparable to those in NPS [22,28,29,36,82,115,116]. One study reported a clinical sensitivity of 68% due to a portion of the samples collected from convalescent patients [100]. However, Rao et al. [27] found that the detection rate for SARS-CoV-2 in saliva was higher than in NPS samples (93.1% vs. 52.5%) when testing 217 convalescent patients, suggesting that the viral RNA levels in saliva and NPS are different in convalescent patients. Furthermore, clinical sensitivity of 73.3% was reported when 2/3 samples were collected from inpatients with hematological disease [6], implying that this disease may influence SARS-CoV-2 distribution in patients.
The viral RNA concentrations in paired gargle and NPS samples were also compared using commercial RNA extraction kits with incorporated silica beads plus RT-PCR or RT-LAMP (Table S4). The clinical sensitivities range from 80% to 98%, indicating the gargle specimens are also comparable to NPS for SARS-CoV-2 testing. In these studies, when NPS and gargle specimens were both positive, the Ct values for gargles were significantly higher, probably because gargle is more dilute than saliva. The solution type, volume, and gargling time used for gargle collection in each study varied, which contributes to the observed variation.
3.4.2.2.3. Purification and concentration of RNA using carboxylated magnetic beads. Magnetic beads with carboxyl functional groups on the surface have also been used to capture SARS-CoV-2 RNA in the lysed saliva samples (Table S4). The principle behind the binding between RNA and carboxylated magnetic beads is that in the presence of saturated PEG [H-(OeCH2eCH2)n-OH] and salt, PEG can grab H 2 O molecules and destroy the protective layer surrounding RNA, leaving RNA naked to each other. The Na þ ions (high concentration of salt) shield the negative phosphate backbones, causing RNA to stick together and to the magnetic beads [117]. Liu et al. [7] used carboxylated magnetic beads (SPRIselect beads, Beckman) to capture RNA from saliva and gargle samples that were treated with a viral inactivation and RNA preservation (VIP) buffer.
The SPRIselect beads (10e20 mL) had no inhibitory effect on the downstream RT-PCR analysis. Thus, the beads were directly used in the RT-PCR reaction without the need for elution of the captured RNA. Because of the higher amount of sample input (e.g., 200 mL) and no dilution from elution, the analytical sensitivity of the assay was substantially increased [7]. This assay yielded a LOD of 25 RNA copies per 200 mL of gargle or saliva sample, which is equivalent to 0.125 viral RNA copies/mL. This analytical sensitivity is higher than that obtained using a conventional viral RNA extraction kit (QIAamp Viral RNA Mini Kit, Qiagen) when the two methods were compared in parallel. The VIP-magnetic-beads method is 10 times cheaper than the QIAamp Viral RNA Mini Kit (~$0.55 vs.~$5.6 per sample) and is suitable for point-of-care testing when combined with isothermal amplification techniques [7]. Kellner et al. [118] developed a Bead-LAMP assay that used carboxylated magnetic beads (RNAClean Xp, Beckman). Gargle solution or sputum samples were treated with QuickExtract DNA extraction buffer followed by heating at 95 C for 5 min. The resultant crude solution was mixed with a bead solution (containing beads, PEG, and buffer) to concentrate RNA. Finally, the captured RNA was eluted using LAMP reaction mixture and directly analyzed using RT-LAMP. This Bead-LAMP assay has analytical sensitivity comparable to RT-PCR for testing SARS-CoV-2 RNA in QuickExtract lysate. These two studies demonstrate that RNA captured on carboxylated magnetic beads can be entirely put into RT-PCR or RT-LAMP for amplification and detection. Thus, concentrating RNA with carboxylated magnetic beads allows enhanced analytical sensitivity.

Overall comparison of RNA-extraction-free methods and RNA extraction methods for saliva and gargle samples
Each of the RNA-extraction-free methods and RNA extraction methods has advantages and disadvantages. To help readers choose suitable methods based on their specific experimental needs, we have compared and summarized the advantages and disadvantages of each method in Table 1. In general, RNA-extraction-free methods are cheaper and faster than RNA extraction methods, but their analytical sensitivity is lower. Among RNA-extraction-free methods, heating plus homogenization in combination with RT-PCR has the highest analytical sensitivity. The method of heating plus homogenization has the potential to be used with isothermal amplification techniques. However, a special homogenizer may not be available or suitable for point-of-care settings. Replacing the homogenization equipment with chemical treatment (proteinase K and other reagents) makes the sample processing suitable for point-of-care testing, although the analytical sensitivity is slightly lower. Two devices with integrated heating lysis and isothermal Fig. 4. Schematic of RNA extraction using silica columns (A) or silica-coated magnetic beads and carboxylated magnetic beads (B). (A) RNA is captured on the column. After washing away non-RNA materials, the captured RNA is eluted with RNase-free water or buffer. (B) RNA is captured on magnetic beads. The beads are collected using a magnet, and the supernatant is discarded. The beads are washed several times to remove non-RNA materials. Carboxylated magnetic beads with captured RNA can be directly used as input for RT-PCR or isothermal amplification and detection. Captured RNA can also be eluted off from the beads for subsequent analysis. Graphics created with BioRender.com. detection are also good options for point-of-care testing [98,99]. In remote areas where heating at 95 C is not available, a lysis solution containing chemical reagents is a suitable choice for releasing RNA. Isothermal amplification techniques, such as RT-LAMP, and nanoparticle-based colorimetric detection provide point-of-care testing potential. RNA extraction methods involving silica columns, and silica-coated, carboxylated magnetic beads outperform RNA-extracton-free methods in terms of analytical sensitivity. The method of simultaneous viral inactivation and RNA preservation, in combination with capture of RNA on carboxylated magnetic beads, is the best option because it is ten times less expensive than commercial kits while simultaneously providing high analytical sensitivity.

Conclusions and perspectives
Saliva and gargle samples are excellent alternatives to NPS samples. The collection of saliva and gargle samples is noninvasive. Self-collection of saliva and gargle samples avoids crowding public testing spaces and minimizes the risk of crossinfection. Frequent and repeated collection of multiple gargle and saliva specimens is feasible and could be used for studying temporal variations in viral concentration in infected individuals.
A critical assessment of the published research enabled us to make the following four recommendations with regard to the collection of saliva and gargle samples: First, the duration of abstention from eating, drinking, chewing gum or tobacco, or smoking should be consistent prior to sample collection. It is best to collect saliva and gargle samples immediately upon waking up in the morning. Alternatively, 1 h of abstention prior to sample collection is advised. Second, drooling is recommended in order to collect clear saliva samples that can be easily processed for RNA extraction. Third, 1e2 mL of saline solution and gargling for >10 s are advised for gargle sample collection. Finally, the saliva and gargle samples should be inactivated during collection to minimize the risk of viral transmission to people who subsequently handle, deliver, and analyze the samples.
Processing of saliva and gargle specimens includes RNA extraction and RNA-extraction-free approaches. RNA extraction methods involving silica columns, silica-coated magnetic beads, and carboxylated beads provide superior analytical sensitivity than RNA-extraction-free methods. RNA extraction methods are more expensive than RNA-extraction-free methods. When RNA extraction reagents and kits were in shortage due to high demand for SARS-CoV-2 tests worldwide, RNA-extraction-free techniques provided much needed alternatives. Heating combined with homogenization is the best option when a homogenizer is available. Two RNA-extraction-free techniques are promising for point-ofcare testing. One is the miSHERLOCK device with integrated heating and RT-RPA and the other combines heating with proteinase K and RNAsecure (a major component being DTT) treatment followed by RT-LAMP. These techniques offer good analytical sensitivity, with LOD of 1.2e1.5 viral RNA copies/mL. Sample processing techniques must be adjusted to further boost analytical sensitivity and decrease analytical discrepancy between investigations.
For heating-mediated RNA-extraction-free methods, certain enzymes, surfactants, and reducing agents have unique capabilities for denaturing proteins, breaking viral particles, and inactivating RNases. Methods combining heating with the use of these reagents should be optimized to efficiently release RNA, protect RNA from degradation, and eliminate inhibitors that could interfere with the subsequent analysis. These methods should be separately optimized for RT-PCR and isothermal amplification techniques, such as RT-LAMP and RT-RPA, because they have different tolerances to inhibitors. For example, proteinase K and Triton-X100 can be used in the sample treatment protocols for the subsequent RT-PCR analysis while proteinase K, surfactants, and TCEP/EDTA are appropriate reagents used in the sample treatment protocol for the downstream RT-LAMP and RT-RPA analyses.
Compared to NPS samples, saliva and gargle contain a more complicated matrix. The commonly used commercially available kits for RNA extraction from NPS are not suitable for RNA extraction from saliva or gargle. Treatment of saliva and gargle samples prior to RNA extraction should be optimized and then applied. New materials, such as magnetic beads with carboxyl functional groups on the surface, are useful for capturing RNA from saliva and gargle samples. Direct analysis of the captured RNA on these magnetic beads without the need for elution of RNA is advantageous. Concentrating RNA from a larger volume (e.g., 1e2 mL) of saliva and gargle samples onto magnetic beads increases the overall analytical sensitivity. Continuing development of methods for self-collection of saliva and gargle samples, treatment of samples, preservation of nucleic acid targets, and amplification techniques will further advance molecular detection of SARS-CoV-2 and other emerging pathogens.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
Data will be made available on request.