Application of nucleic acid amplifi cation tests in managing COVID-19 pandemic

Background. COVID-19 pandemic highlighted the importance of sensitive and specifi c tests that would be cost-effi cient, fast and scalable. There are more than 200 COVID-19 detection tests available worldwide, with every country developing its own assays. Sample collection, preparing for a test, the test itself and interpretation of results have a strong impact on the clinical value of testing. The diversity of tests and workfl ows requires the analysis of their performance in clinics.

Проаналізувано робочі протоколи тестувань зі застосуванням полімеразної ланцюгової реакції (ПЛР) або ізотермічної кільцевої ампліфікації (LAMP). Різноманітність протоколів підготовки зразків, відмінності у виконанні тестів та протоколів оцінки результатів мають істотний вплив на чутливість та вибірковість тестів. Цю різноманітність узагальнено з акцентом на позитивні та критичні параметри робочих протоколів кожного з етапів тестування. Таке аналізування може бути корисним при виборі тестування. Праці  Introduction. The workfl ow of COVID-19 testing includes the collection of a sample, testing on-site at the point of care (POC) or transportation to a laboratory, testing with the use of advanced tools, interpreting results. By August 1, 2020, 125 COVID- 19 test systems have been approved by the Food and Drug Administration (FDA) in the USA (https://www.fda.gov/medical-devices/ coronavirus-disease-2019-covid-19-emergency-use-authorizations-medical-devices/ vitro-diagnostics-euas), and more than 100 test systems were registered in China (see references at http://ph.china-embassy.org/ eng/sgdt/P020200324570010409522.pdf) and the European Union countries (see references at https://ec.europa.eu/jrc/en/news/ coronavirus-testing-information-test-devices-and-methods-single-place). The offi cial website of the Ukrainian government reports the use of German and Chines test systems, not specifying their number and types (https://covid19.gov.ua/en). The number of reported test systems is most probably an underestimation, taking into account numerous developments in research laboratories and the legal recognition of laboratory-developed tests on the same level as in-vitro diagnostics [1][2][3]. Moreover, an analysis of the logis-tic of COVID-19 testing showed the importance of performing tests at core laboratories in the hospitals and points of care instead of outsourcing them to central laboratories [4]. Hospitals' core laboratories decreased the turn-around time from 21 to 3.7 days as compared to the outsourcing of testing [4].
COVID-19 detection tests are classifi ed into 2 types -nucleic acid detection and serological/immunological tests. Herein, we focus on nucleic acid amplifi cation-based tests. All reported workfl ows of COVID-19 testing include sample collection, preparation for a test, the test itself and interpretation of results ( Figure  1). Variations are in the origin of the sample, the time for collection, conditions of preparation and transportation of the sample, test type, logistics, e.g. POC or outsourcing, and clinical interpretation of results.
There is no test that would secure 100% sensitivity and specifi city. The effi cacy of testing depends on the probability of detecting the infection and is the main concern concerns the selection of a workfl ow. For example, viral nucleic acid detection is dependent on the viral load in a selected type of the sample over the course of infection ( Figure 2) [5,6]. The highest viral load Main steps of the workfl ow are shown. Sample collection blocklist types of samples that were successfully evaluated for COVID-19 detection. Sample preservation and transportation is the next step of the workfl ow. RNA purifi cation step is widely used but is not essential. in nasal and oropharyngeal swabs was observed upon the onset of symptoms. However, even at the points of the highest viral load, the probability of virus detection has never been 100%. It was reported that only 42% of people who died from COVID-19 tested positive for COVID-19 [7]. Failure to detect COVID-19 could be due to variability in the effi cacy at any stage of the workfl ow. Examples of test specifi city and sensitivity are reported to range from 90% to 80% [6][7][8][9]. Therefore, biomedical variables, e.g. viral load and technical suitability of test workfl ow are crucial for the interpretation of results.
To mitigate potential failure, multiple tests are ordered for a patient suspected to be infected; e.g. reportedly, the conclusion on the absence of infection may require up to 4 tests per person and at least 2 subsequent negative tests [6][7][8][9].
Therefore, accumulated clinical experience shows that there is a need for multiple testing and a clinical decision should include the interpretation of the patient's clinical condition ( Figure 3). The following sections focus on each step of the workfl ow starting with a sample collection, followed by the point of care tests and tests in centralized laboratories, and then by the interpretation of results with emphasis on lessons learned from the tests based on the amplifi cation of COVID-19 nucleic acid.

Search strategy
The PubMed database was searched with the Medical Subject Headings (MeSH) search terms "COVID-19", "detection", "test", "PCR" and "LAMP". Collected publications were screened manually for a description of COVID-19 detection methods. The same terms were used to search Google, as there are many publications deposited online but not represented on PubMed, e.g. bioRxiv.org. The third source of searches were websites of agencies involved in fi ghting COVID-19, e.g. who.int, www.fda.gov, www. ema.europa.eu and moz.gov.ua. The fourth source of information were online resources of companies producing COVID-19 detection kits.
Inclusion criteria were 1) a description of types and performance of COVID-19 detection kits, 2) description of technologies, reagents and protocols used in COVID-19 testing and/or 3) analysis and comparison of diff erent kits and protocols. The exclusion criterion was the lack of detailed information about the kit, e.g. no description of the technology, no information about reagents and a lack of detailed protocol. This search was last updated on the 15 August 2020.

Sample collection
The best option for sample collection would be self-collection at the time of the highest The window of COCID-19 detection during the disease is indicated by a red square line. The blue line illustrates the viral load in the sample collected at diff erent disease stages, e.g. initial infection, the onset of symptoms and recovery. Limit of detection of the PCR testing is defi ned as 10^3 -10^4 per ml of a sample. Reportedly, COVID-19 was detected in all tested sources, i.e. nasopharyngeal, bronchoalveolar lavage (BAL), sputum, saliva, cerebrospinal fl uid (CSF), plasma and stool [11][12][13][14][15][16]. Multiple sample sources refl ect a broad range of tissues and cells directly aff ected by the virus [11,17]. Endothelial, epithelial, myoepithelial, smooth muscle, hepatocytes, neurons and glial cells were identifi ed as targeted directly by COVID-19 [17].
The important observation is that virus detection is more dependent on the course of disease than on the sample source [3,6,7]. The stage of symptom appearance may have the highest viral load, while the load is lower at the stages of initial infection, recovery and after-recovery ( Figure 2). The detection limit of most tests ranges from 5 to 10 viral RNA molecules per reaction, which corresponds to the viral load in the range of 3x10^4 RNA copies/ml [18]. If the load of the viral RNA is below 1x10^3 copies/ml in a sample, this would require reconsideration of the source of sampling and/or a need for concentration of the RNA.
Currently, the most frequent source of samples includes nasal and oropharyngeal swabs. More than 160 designs of swabs have been reported [12]. The use of these swabs was similar, provided that its material did not interfere with the extraction of nucleic acids and PCR reaction, e.g. swabs must not contain cotton, wood or calcium alginate [12].
Collection solutions, on the other side, may have a signifi cant impact on the preservation of the viral RNA and compatibility with subsequent testing. For example, the use of Variplex system without RNA extraction in a LAMP test led to 83% false-negative rate [13]. Therefore, a sample solution must be compatible with transportation, storage conditions and the type of test to be used. Before embarking on the full-scale testing, compatibility of the sample solution with the planned sample type should be tested using control virus-contain-  There have been numerous attempts to minimize variability in sample collection. These include heating of samples and adding organic solvents and detergents. The rationale is that heating would lead to denaturation of molecules in the sample, including RNAases [19][20][21]. Organic solvents and detergents are expected to produce a similar result, i.e. inactivation of RNA degradation enzymes [19][20][21]. The additional eff ect is the dissociation of RNA-containing viral particles in the presence of detergents and subsequently, the release of RNA into a solution.
The report by Pan et al. showed that sample heating increased Ct of detection, indicating decreased sensitivity [5]. Another heating testing (at 56 0 C to 65 0 C) showed no diff erences as compared to non-treaded sample [22]. Adding ethanol to a sample solution also had an inhibitory eff ect on the microbial growth in the nasal, oropharyngeal swabs and saliva [20,22]. Strong detergents may inhibit the PCR reaction. For example, even low concentrations of sodium dodecyl sulfate, i.e. 0.1%, strongly inhibited PCR reaction. The inhibitory eff ect was also observed with detergents of Triton X-100 type at concentrations higher than 1.0% [20,21]. The use of additives, like ethanol or detergents, should be pre-tested for every workfl ow to take advantage of blocking RNA/DNAases, sterilization and solubilization of viral RNA, and to avoid any negative impact on the reactions of reverse transcriptase and nucleic acid amplifi cation.
Therefore, when selecting a sample collection protocol, consideration should be given to a) the type of sample and collection method, e.g. swab, saliva, stool and sample solution compatibility; b) expected viral load in collected samples, to ensure that LOD of the test would allow detecting the infection; and c) preservation of viral nucleic acids in the sample solution upon collection and transportation.

Sample preparation
RNA in collected samples has to be accessible for amplifi cation. The most common approach is to purify RNA and then use it in reverse tran-scriptase and amplifi cation reactions. There are many commercial kits for RNA purifi cation. The quality of kits is generally good, no serious issues have been reported. The only consideration in terms of selecting a purifi cation protocol is the cost of the kit, requirements concerning the tools, reagents and personnel. RNA purifi cation stage may also be prone to failures, especially when the viral load is low. Therefore, there were attempts to develop protocols that would omit the nucleic acid purifi cation step. Circumventing RNA purifi cation signifi cantly improves and facilitates on-site POC testing. Several reports show COVID-19 markers in nasopharyngeal swabs [15,21] and saliva [14,20]. A direct comparison with the protocol including RNA purifi cation showed similar detection accuracy and reliability. The only concerns were about the potential decrease of LOD in presence of strong denaturants in the sample solution, e.g. SDS, and interfering components in the sample itself, e.g. mucin, enzymes, etc.
Therefore, the stage of sample preparation offers options with or without RNA purifi cation. If the sample is used for further studies of COVID-19, e.g. sequencing, RNA purifi cation is There is no sharp discrimination of these two application types by technologies employed in the tests. PCR and LAMP amplifi cation can be employed in POC and laboratory-based tests. The design of devices and instruments defi nes whether the test is suitable for POC or central laboratory-based detection. Small tools even allow real-time PCR using a small benchtop instrument with minimum requirements to sample preparation. An example of such approach is reported by Wee and colleagues [23]. LAMP is usually used for POC tests, as it does not require expensive tools. LAMP detection can be performed using any device that maintains a constant temperature, e.g. a heating block or a thermostat.
Tools are becoming cheaper and more compact. On the contrary, the cost of consumables and reagents is a signifi cant part of testing expenses. In addition, the miniaturized and automated tools use dedicated consumables. It limits the use of these tools to these unique consumables and minimizes fl exibility of assays.

Loop-mediated isothermal amplifi cation (LAMP) tests
The application of LAMP to detect COVID-19 has been successful. Some publications reported and review LAMP assays to detection COVID-19 [24][25][26][27][28][29]. Herein, we focus on criteria to consider when selecting a LAMP test (Figure 4). Recognition and amplifi cation of the targeted viral sequences are dependent on the specifi city of primers, the temperature of the reaction, buff er composition, pH, and presence of interfering substances from the sample. The effi cacy of reverse transcriptase and a DNA polymerase also aff ects test performance.
There are no reported warnings for targeting specifi c COVID-19 genes and avoiding others. The consensus is that the targeted region is not crucial, as long as the sequence is unique for COVID-19 [29]. Similarity search tools, e.g. BLAST of NCBI, are a good option to fi nd primers that would be unique to COVID-19 with no overlap with other species and genes, as they detect only COVID-19.
LAMP methodology is based on the recognition of 6 sequences of the targeted gene, followed by a building and amplifi cation of a nucleic structure representing targeted sequences, and the detection of this amplifi ed structure [27][28][29]. The positioning of targeted sequences allows LAMP primers to build a structure that would be self-amplifi ed. The key to performance of a LAMP test is primers design (Figure 4). structure is formed, the amplifi cation from the viral template is not maintained any more. The amplifi cation is dominated by the DNA synthesis from the formed structure. Therefore, the applicability of the LAMP test is strongly dependent on the recognition of targeted viral sequences by primers during the initial phase of double-loop structure formation.
There is no visualization of amplifi cation products in the standard LAMP test, e.g. the size of generated DNA products cannot be controlled. The LAMP signal is dependent on the quantity of synthesized DNA and the type of DNA detection. For example, for detection using pH-sensing dyes, a buff ering capacity of the reaction should be not higher than 1 mM for a Tris buff er [30]. Frequently used pH-sensing dyes, e.g. phenol red, cresol red, neutral red, hydroxy naphthol blue, could detect the accumulation of DNA at an initial level of 3 to 30,000 copies in a reaction mix. This level of detection is comparable to real-time and classical PCR [30]. Direct comparison of the quantities of generated DNA in a LAMP and PCR assays is not relevant because the limit of detection plays a more important role, e.g. sensitivity of the detection method is crucial.
To detect the virus using DNA-interacting dyes, the capacity of dyes to inhibit the amplifi cation reaction has to be considered. Quyen and colleagues tested 23 dyes and showed that some of DNA fl uorescence dyes can inhibit LAMP reaction. The high inhibitory eff ect was reported for POPO3, DCS1, SYBR Green I, BOBO 3, Pico 488, and TOTO 3 dyes. Dyes SYTO 9, SYTO 13, SYTO 16, SYTO 64, SYTO 82, Boxto, Miami Green, Miami Yellow, and Miami Orange were found not to interfere with the amplifi cation of DNA [31]. Frequently used cresol red, neutral red and phenol red dyes have not been reported as inhibitors of the LAMP reaction. For the use of other dyes, a comparison test is recommended adding dyes before and after the reaction, followed by monitoring of generated DNA products by an agarose gel electrophoresis.
LAMP was successfully used to detect COVID-19 in versions with and without RNA purifi cation [32]. The authors targeted N-gene of the virus. Detecting a positive signal with the LAMP test was comparable with Ct below 30 cycles for a real-time PCR [32]. This indicates that the LAMP assay can be as sensitive as the real-time PCR.
Therefore, to develop an effi cient LAMP test, optimization trials have to address: a) the design of primers, e.g. computer-assisted design is required, b) sample collection conditions should be optimal and composition of the sample collection solution should not interfere with LAMP, e.g. no detergents or nucleases, c) selecting amplifi cation conditions (buff ers, enzymes, additives and the protocol should allow effi cient amplifi cation and detection), and d) the detection system should allow effi cient detection, e.g. by selecting DNA dyes/fl uorescence, pH-sensing or pyrophosphate precipitation ( Figure 4).

PCR tests: real-time reverse transcriptase and standard reverse-transcriptase tests
PCR tests are the golden standard for COVID-19 detection. PCR reaction is highly specifi c, has high fi delity, solid technology development and ensures high detection specifi city and sensitivity. A real-time reverse transcriptase (qRT-PCR) and standard reverse transcriptase (RT-PCR) use the same PCR principle, but diff erent combinations of primers and diff erent methods of signal generation and detection ( Figure 5).
Real-time PCR (qRT-PCR) is the most frequently used technique to detect COVID-19. It is explained by robust development of its theory, reagents, protocols and tools. The success of qRT-PCR is also dependent on the automation and simultaneous amplifi cation and detection of the product. The majority of approved COVID-19 detection tests are based on qRT-PCR (to see examples, see www.fda. gov/medical-devices and ec.europa.eu). They provide a good balance of high-quality PCRbased detection and a reasonable level of automation. However, some issues must be controlled to ensure high performance of tests, which are discussed in this section.
Standard RT-PCR is more laborious as compared to qRT-PCR. To assess RT-PCR result, the generated product must be visualized. Agarose gel electrophoresis is a standard technique for visualization. When the analysis quality has to be the highest, RT-PCR is the fi rst choice. The visualized product shows the size and can be sequenced for validation. Sequencing of the generated product is also used for monitoring of mutations in the viral genome. The sequencing of RT-PCR products provides data for the monitoring of viral strains and subsequent spreading of the disease. Viral mutations may aff ect treatment strategies too. Therefore, if COVID-19 testing requires the highest quality and/or is to be combined with a study of COVID-19 virus, standard RT-PCR is the method of choice ( Figure 5).
When selecting a qRT-PCR or RT-PCR test for a clinical application, the entire workfl ow must be designed. The test must be compatible with sample collection and preparation protocols. The specifi city of primers, conditions of the reaction, specifi cation of tools, and available laboratory infrastructure are other concerns.
Failure to develop a proper workfl ow design may lead to low sensitivity and specifi city. Recent reports show that qRT-PCR tests may not always detect positive cases, giving a false-negative value in 80% of cases [9]. This means that many positive cases are missed. Such a test may subsequently fail in preventing the infection spread. The analysis shows that the reason could be in a non-optimal workfl ow, and not in the performance of qRT-PCR reaction itself. Negative results may be the result of sample collection and preparation, where the viral RNA has low stability and losses of RNA during purifi cation and interference with the effi cacy of PCR reaction [9]. This calls for positive controls in samples too, not only a positive technical control of the detection system. In clinical practice, it is ensured by spiking a sample upon collection with a known quantity of COVID-19 genomic marker, e.g. adding an aliquot of the sequence probe targeted in the test DNA.
To ensure successful completion of qRT-PCR and RT-PCR tests, diff erent combinations of primers and multiplexing have been tested [33]. Primers targeting nucleocapsid (N), membrane protein (M), spike (S), envelop (E), nsp2, RNA-dependent RNA polymerase /helicase (RDRP/Hel) and orf1a regions have been reported [20,23,33,34,35,36]. The conclusion is that the location of targeted sequences in COVID-19 genome does not infl uence detection. The design of primers to ensure COVID-19 specifi city is crucial. Primers' specifi city is easy to secure with available online tools, e.g. BLAST of NCBI (blast.ncbi.nlm.nih.   [36]. The success of this optimization was due to product visualization by standard RT-PCR used for optimization, as qRT-PCR does not visualize products. 3-plexing detection limit reported by Ishige and colleagues was calculated as 25 copies of COVID-19 RNA per reaction [34]. Simultaneous detection of 4 genes (with 8 primers in one reaction) was reported by Liu et al [37]. Thus, the reported developments of COVID-19 PCR tests showed that primers can target all regions of the viral genome, and multiplexing up to 4 gene markers in one reaction is possible. The design of primers can be performed with available online tools while securing COVID-19 specifi city. Primers for qRT-PCR have to be validated using RT-PCR and amplifi ed product visualization.
The issues with PCR tests have been attributed to sample collection, RNA preparation and interference with PCR reactions. Storing viral particles and RNA upon collection, losses of RNA during purifi cation and PCR reaction inhibiting substances are the main concerns ( Figure 5).
RNA is highly sensitive to degradation. Stabilization of RNA upon collection has to be validated for sample collection solution. It is reported that Universal (UTM) and Viral (VTM) transport media are designed to preserve or lyse virus particles. If the testing workfl ow presupposes RNA purifi cation step that would remove all components of the transportation media, then there are no serious precautions to consider. RNA purifi cation for COVID-19 tests is performed with the use of commercial kits. These kits are used for an automated or semi-automated procedure. The optimization of RNA purifi cation step includes the evaluation of the lowest concentration of RNA in the Critical points of the real-time RT-PCR (qRT-PCR) and RT-PCR are illustrated in the "Concerns" block. Potential solutions to these concerns are indicated in the "Solutions" block. For description, see the text. sample that the kit can recover from the sample to ensure the acceptable limit of detection.
The prevention of RNA degradation by following the collection and transportation protocol would be the only other requirement. The technical control over the purifi cation and PCR reaction includes the detection of household human genes, e.g. RNAse P gene. If the workfl ow circumvents RNA purifi cation step, the direct detection would require lysis of the sample, release and stabilization of RNA. It was reported that detergents, e.g. Triton X-100, Tween 20, in concentrations of up to 1% in the transportation medium were tolerated in a reverse transcriptase and PCR reactions. Snap-heating of the collected sample to 70 0 C and up to 120 0 C may be considered for sample preservation [19][20][21]. Thus, optimizing the testing workfl ow may require the evaluation of the transportation media (preserving or lysing), transportation conditions (frozen or +4 0 C), and direct detection or purifi cation of RNA steps followed by the PCR reaction [19][20][21].
Test effi ciency depends on primers, enzymes and reaction buff ers. The design of primers was discussed above. Reverse transcriptase and DNA polymerases with and without exo-nuclease activity and a strand-displacement activity (e.g. Bst DNA polymerase for LAMP, Pfu and Taq DNA polymerases for PCR) are available from many suppliers. To select the enzyme, it is important to select the reaction mix, too. Enzyme suppliers off er the reaction mix to be used with their enzymes. As this master mix is already optimized with enzymes, it is recommended to evaluate proposed combinations fi rst. If the proposed enzyme-master mix combination is not performing well, an alternative combination must be considered and tested. In some cases, it is possible to develop a special master mix, but it requires signifi cant eff orts to produce in-house enzymes.
For qRT-PCR tests, positive and negative controls are standard. In addition, to optimize tests with these controls, it is recommended to include the acquisition of the melting curve. The analysis of amplifi ed products by electrophoresis is not performed, as the product is smaller and can be misinterpreted as primer dimers. qRT-PCR curves provide quantitative information, e.g. Ct values, which facilitates the interpretation of results.
Interpretation of RT-PCR results is straightforward using gel electrophoresis. The detection of amplifi cation products of the expected size, and, if required, sequencing of these products provide a secured interpretation. Standard RT-PCR is semi-quantitative. However, visualization of the amplifi ed products makes quantifi cation less important for the interpretation of results. For the clinic, the result must be "positive" or "negative", and the visualization of the product is suffi cient for such a conclusion ( Figure 5).
To optimize the testing workfl ow, it is recommended to include the detection of the endogenous human gene(s) in the testing, e.g. RNAse P gene [38]. This allows monitoring the entire workfl ow, while PCR positive and negative controls allow monitoring a PCR reaction.
The detection effi ciency is dependent on the stage of the disease (Figure 1). An example of a low consistency between COVID-19 detection and CT changes in lungs may indicate that virus detection does not correlate with specifi c clinical symptoms [39]. This is a strong indication that COVID-19 detection must be interpreted in combination with all clinical information, e.g. symptoms, history of a patient's health and travel pattern ( Figure 3).
The fi nancial drawback of PCR tests is the requirement for advanced tools and infrastructure. To take PCR-based tests to the bedside and clinics and healthcare providers on-site, portable devices (POC devices) are under development. Wee et al reported the development of the PCR tester for nucleocapsid (N) gene detection with LOD 6 copies of RNA per reaction from sputum and nasal exudate [23]. The readers are directed to the review of POC devices by Cheng et al. [8]. The performance of these devices is currently under evaluation, and if validated, it would signifi cantly ease the load on laboratories.
To sum up, PCR-based tests are and will be the main standard in the detection and study of COVID-19. Multiplexing of qRT-PCR will increase its clinical value. RT-PCR is indispensable in the development of PCR-based tests and the study of COVID-19. Both qRT-PCR and RT-PCR deliver reliable performance. However, workfl ow optimization is essential. Selecting the sample type, sample collection medium, storage and transportation conditions, RNA purifi cation step, PCR test itself and interpretation of results must be performed with patients' and control samples.

Conclusion
The success of the fi ght against COVID-19 is dependent on detection tests. Today, more than 200 tests are available in the market. Most of these tests perform well when manuals and recommended protocols are followed [40]. In addition to nucleic acid amplifi cation tests discussed herein, immunological/ serological tests, novel variants of testing by massive parallel sequencing [41] and digital droplet PCR [42] are coming into the market. However, testing does not consist in the tests only. Testing is a workfl ow that includes sample selection, collection, transportation, preparation for a test, the test itself and result interpretation ( Figure 1). Moreover, the results of testing should be interpreted together with the examination and clinical symptoms ( Figure 3). Tests detecting genetic material of COVID-19 are and will be used in the foreseen future, as they ensure the most reliable virus detection Tailoring the testing workfl ow to the specifi cs of every healthcare provider would require the optimization of sample collection, detection and interpretation processes. This review highlighted some of the concerns of such optimization.