Detection of Coronavirus in Simultaneously Collected Tear and Throat Swab Samples from the Patients with 2019- new Coronavirus Infection Disease: A Single Center Cross-sectional Study

Objecive: This study aimed to evaluate whether the coronavirus can be detected in the tears in the eyes of 2019- novel coronavirus (2019-nCoV) infected patients and compa re the detection consistency of two simultaneously collected samples, from the tears and throat swabs, of these patients. Methods: A total of 35 diagnosis conrmed patients with 2019-nCoV infection disease,from January 26 to February 19, 2020, at central theater Wuhan general hospital of PLA, were included in this cross-sectional case series study. The diagnosis was conrmed by both clinical and laboratory ways. Throat samples from all enrolled patients were collected with sampling swab, and simultaneously ,tear samples were collected w ith sampling swab from 9 patients (No.1-9) and with Schirmer's strip from the remaining patients (No.10-35) (bilateral eyes for all patients). Sample collecting and testing were performed in three separate time points: rst from patients No.1-9,second from p atients No.10-29,and third from patients No.30-35. Reverse-transcriptase-polymerase-chai n-reaction (RT-PCR) assay was performed by the same technician(Researcher Zhang), who was blind to the patients prole, within 4 hours after each collection. Results: Among enrolled patients, 29 (No.1-29) had mild or moderate clinical symptoms and 6 (No.30-35) had severe symptoms. The mean time interval from the sample collection day to diagnosis conrmation day was 9.71±6.50 days (ranged from 3 to 29 days). None of the patients had conjunctivitis. Nineteen out of 35 (54.3%) throat samples presented positive Rt-PCR results. Three (no.13,21,31) out of 35 (8.6%) tear samples presented positive RT-PCR results. Two(no.21, 31) of these three patients were throat swab positive and one (No. 13) was negative. The consistency analysis indicated that tears and throat samples showed poor consistency ( Kappa=-0.009, P=0.9 ).The cycle threshold value (Ct-value) of tears samples collected by sampling swab was signicantly higher than that by Schirmer's strip(t=2.288, P=0.03). Conclusion: In spite of the low 2019-nCoV positive detection rate of tear samples from 2019-nCoV infected patients, we cannot fully rule out the transmission by ocular surface. Whether tear testing can be used as an aid in judging of 2019-nCoV infection need further investigation. Authors Ying Yan and Bo Zeng contributed equally to this study The baseline characteristics stratified by positive and negative throat swab Rt-PCR results of the mild and moderate


Introduction
A novel coronavirus-infected disease was rst reported in Wuhan, a city in central China in December 2019 [1] . Up to March 10, 2020, a total of 80928 2019-nCoV infected cases in China have been con rmed [2].The outbreak of the disease has become a global health threat, and almost 24 countries and 5 continents have been a icted with 2019-nCoV infected disease [3]. On January 7, 2020, the 2019 novel coronavirus (2019-nCoV) was isolated from the clinical specimens of bronchoalveolar lavage uid and was con rmed as the cause of the disease [4],and a rapid human to human transmission has been con rmed [5]. Although respiratory droplets or contact has been identi ed as the route of human-to-human transmission [6], other potential routes including ocular surface secretions transmission have not been investigated [7].
It is reported that,the tears, being one of the body uids, may potentially harbor the coronavirus. In the past, researchers detected many in uenza viruses, including Ebola virus [8], severe acute respiratory syndrome coronavirus [9],and human immunode ciency virus [10], in tears using polymerase chain reaction (PCR) technique. Therefore, we conducted this study to investigate whether 2019-nCoV can be detected in the tears of patients with 2019-nCoV infection.

Methods
This is a cross-sectional case series study.Thirty-ve patients with con rmed 2019-nCoV infection by both the clinical way (includesymptoms, blood routine test and chest computed tomography) and the laboratory way (the revere-transcriptase-polymeras e-chain-reaction (RT-PCR) from throat swab sample), were recruited at the time of ad mission in the Central Theater General Hospital, Wuhan, China, from January 26 to F ebruary 19, 2020. Approval was obtained from the Ethics Committee of the Central T heater General Hospital(No. [2020]009-1). The patients' signed consents were obtained.
Throat samples and tears of all patients were collected simultaneously. Throat samples were obtained by cotton swab scraping ministry of epiglottis,and tear samples from both eyes were collected with sampling swab (No.1-9) and with Schirmer's strip(No.10-35), respectively. For sampling swab, the samples were taken by putting sterile cotton wool sticks into the deep lower fornix of patients' eyes. For Schirmer's strip collection, a Schirmer's strip ( Schirmer Tear Test Strips; Liaoning Meizilin Pharmaceutical Co.,Ltd) was placed at the temporal canthus of each eye for 5 min without anesthetic and then removed from the eye.
The amount of wetting in millimeters was recorded by observing the location of the leading edge of moisture on the printed millimeter marks. All samples were placed into disposable virus sample tubes.
Gloves were changed when collecting specimens in between patients to minimal contamination. . Basically, 300uL lysis buffer was added into the tube with the test paper, then mixed by vortex for 3 minutes, and the nucleic acid was puri ed following the instruction of the kit. The 2019-nCoV was tested by real-time RT-PCR using 2019-nCoV Nucleic acid Diagnostic Kit (Sansure Biotech, Changsha, Hunan Province, China), and two targets (open reading frame 1ab, nucleocapsid protein) were tested.
If either of the two targets (open reading frame 1a or 1b, nucleocapsid protein) was detected by speci c real-time RT-PCR, the case would be considered as laboratory-con rmed. A cycle threshold value (Ctvalue) less than 39 was de ned as a positive test, and a Ct-value of 39 or more was de ned as a negative test.
The mean age was 58.9±15.5 years ranged from 20 to 89 years and 17 (48.6%) were female. The mean time interval from the sample collection day to diagnosis con rmation day was 9.71±6.50days (ranged from 3 to 29 days). Details of the characteristics of the patients were provided in Table 1.
All patients have been free of conjunctivitis since the onset and have not used any eye drops.  Table 2. The average age of positive patients was larger than that of negative patients (63.5 vs. 50.4 years,p = 0.017). The average time interval between the sample collection and diagnosis con rmation day of positive patients was shorter than that of negative patients (6.15 vs. 8.44 days, P = 0.034). There was no difference in sex body temperature, cough, expectoration symptoms, degree of blood oxygen saturation, and the use of antiviral drugs.
Three (no.13,21,31) out of 35 tear samples (8.6%) presented positive RT-PCR results, 2 of which were throat swab positive and one (No. 13) was negative. Two of the these 3 were mild and moderate patients, both were 51 years old. And, the sample collection interval from diagnosis con rmed was 8 days and 7 days, respectively. (Table 3).
The consistency analysis indicated that tears and throat samples showed poor consistency. ( Kappa = -0.009, P = 0.9).
Mean cycle threshold value (Ct-value) of tear samples collected by conventional sampling swab was signi cantly higher than that collected by Schirmer's strip ( 32.68±1.30 Vs 31.39±1.45 P = 0.03).
The tear volumes of Schirmer's strip were negatively associated with the Ct-value for the 6 severe patients ( coe cient value -0.928, P = 0.008) ( Table 4.).

Discussion
The transmission of 2019-novel coronavirus through eyes has recently attracted the researchers' attention. There were sporadic cases of conjunctivitis reported in infected patients whose conjunctival swabs were positive for nucleic acid of coronavirus [11][12][13]. However, for all infected patients, the positive rate of conjunctival swabs was very low, lower than that of throat swabs [11][12][13], which was consistent with the results of our study. It seemed that the possibility of virus existence in the ocular surface is low.
However, more detailed analysis of the eye sample collection process and the laboratory nucleic acid test results can help to obtain more information.
In previous studies, conjunctival swabs were used to collect eye secretions including the tears, which were usually insu cient for test [14]. It was reported that the amount of tears collected by Schirmer's strip at one time was greater than that collected by other methods such as capillary tubes and surgical sponges [15].In the rst nine patients (no.1-9) of this study, conjunctival swabs were used to collect samples and all the test results were negative. In the subsequent patients (no.10-29), Schirmer's strips were used to collect sample, and 2 samples were detected positive. This result suggested that the positive rate of the virus detection may be higher when Schirmer's strip was used. In addition, through the analysis of the volume of tear and the value of nucleic acid detection CT in 6 severe patients, it was found that there was indeed a quantity effect relationship between the two. This, to some extent, indicated that,the larger the amount of tear samples collected, the greater the probability of positive detection.
Another potential factor that affected the test results was the viral load in the sample. In theory, the more viral load, the higher the detection rate. The low positive rate of this study also indicated that the viral load in the ocular surface was relatively limited.
Due to the in uence of multiple factors, the interpretation of the detection results should be more cautious-even throat swabs may have a high negative rate. In this study, throat swabs showed a positive detection rate of 54.3%, which was far lower than that in recently published online study. Different from ours, the sputum (non throat swab) samples of patients were taken for nucleic acid detection, so a high positive detection rate (91.7%) was obtained [13]. In practice, we found that most patients did not have sputum available. In addition to the way of sampling from bronchial secretion, the patients pro le may also affect the nucleic acid detection results. For the patients with mild and moderate disease, we compared the baseline characteristics strati ed by positive and negative throat swab Rt-PCR results (Table 2). Since the average age of the positive patients was greater than that of the negative patients, and the average interval between the sampled and diagnosis con rmed day for positive patients was shorter than that for negative patients,we suggested that the older the patients were, the earlier the detections were performed, the easier it was to detect nucleic acid fragments from throat swabs. However, for severe patients, because of their 100% positive rate, the possibility of the test results affected by the factors mentioned above should be small.
The fact that the consistency between the test results of throat swab and Schirmer's strip was low and that only two of three patients with tear Rt-PCR positive result had throat swab positive result, suggested that the tear test results were not related to throat swab test results. Moreover, the result that one patient (No. 13) showed positive result of tear but negative of throat swab reminded us that Schirmer's strip detection may be used as an auxiliary means to identify patients with false negative throat swabs test result. Although the positive rate of tear test was very low, when facing this highly infectious disease, any additional screening for the control of disease transmission is of practical signi cance.
Limitation of this study was the relatively small sample size, especially the few severe patients included. In practice, in the face of the more infectious severe patients, sample collection performed in a single center had certain challenges. Another limitation also came from the small number of samples: no conjunctivitis patients could be included. Therefore, it was impossible to judge whether it was more likely to detect virus from the tears of conjunctivitis patients. Furthermore, we did not repeatedly collect and test samples,neither tears nor throat swab. Hence, we could not reduce the random errors occurred during the sample collection and testing process.
However, one of the advantages of this study was the simultaneous collection of throat swabs and tears from patients, which was helpful for us to compare the detection results of throat swab and tear. Another advantage came from the use of Schirmer's strip, which could collect larger volume of tear samples.
From this perspective, Schirmer's strip outweighs the conjunctival swab. Most importantly, this study con rmed the practical value of Schirmer's strip by demonstrating the positive tear test result from patient no.13 who had negative throat test result.
In the future, we should pay more attention to the virus detection from ocular surface in different stages of the disease and different degrees of infection. Isolating the virus and determining the virus load, if exists, from the ocular surface, and investigating the existence of eye-body infection pathway will help us to clarify the human-human transmission route via eyes.

Conclusions
We cannot fully rule out the transmission by ocular surface. Whether tear testing can be used as an aid in judging of 2019-nCoV infection need further investigation. Moreover, we propose the use of Schirmer's strips that allow tear collection more e ciently.    The tears volume of Schirmer's strip were negatively associated with the Ct-value of tears sample of 6 severe patients. (Spearman correlation coefficient , coefficient value = -0.928, P =0.008).