【病毒外文文獻】2003 Prediction of amino acid pairs sensitive to mutations in the spike protein from SARS related coronavirus

Peptides 24 2003 1837 1845 Prediction of amino acid pairs sensitive to mutations in the spike protein from SARS related coronavirus Guang Wu Shaomin Yan DreamSciTech Consulting Co Ltd 301 Building 12 Nanyou A zone Jainnan Road CN 518054 Shenzhen PR China Received 4 October 2003 received in revised form 27 October 2003 accepted 28 October 2003 Abstract In this study we analyzed the amino acid pairs affected by mutations in two spike proteins from human coronavirus strains 229E and OC43 by means of random analysis in order to gain some insight into the possible mutations in the spike protein from SARS CoV The results demonstrate that the randomly unpredictable amino acid pairs are more sensitive to the mutations The larger is the difference between actual and predicted frequencies the higher is the chance of mutation occurring The effect induced by mutations is to reduce the difference between actual and predicted frequencies The amino acid pairs whose actual frequencies are larger than their predicted frequencies are more likely to be targeted by mutations whereas the amino acid pairs whose actual frequencies are smaller than their predicted frequencies are more likely to be formed after mutations These findings are identical to our several recent studies i e the mutations represent a process of degeneration inducing human diseases 2003 Elsevier Inc All rights reserved Keywords Amino acid pairs Coronavirus Mutations SARS 1 Introduction Although the severe acute respiratory syndrome SARS has gone after hitting the world for several months every one is intuitively expecting the possible return of SARS in the near future and human logic seems to have such an assumption i e if the SARS would return it would be another mutated form following its first battle with hu mans This is possible because the accumulating evidence shows that there are several mutations in SARS related coronavirus SARS CoV So far 15 point mutations have been documented in SARS CoV proteins two in the 221 amino acid long membrane glycoprotein 2 15 17 19 24 three in the 1255 amino acid long spike glycoprotein 2 15 17 24 and ten in the 7073 amino acid long replicase polyprotein 2 8 11 15 17 24 Naturally we would expect these three SARS CoV proteins to have other forms of mu tations rather than those documented and the new mutations would lead to the difficulties in diagnosis and treatment of SARS An intriguing question is whether or not we can predict the new mutations of SARS CoV If so it would be greatly Corresponding author Tel 86 755 2202 9353 fax 86 755 2520 8256 E mail address hongguanglishibahao G Wu helpful for identification of SARS CoV and a great advance in understanding of the evolutionary process in SARS CoV Also it would be useful for studying the mutant patterns in other human coronavirus which would give us some insight into the mutations in coronavirus Among encoded structural replicase spike envelope membrane nucleocapsid proteins from human SARS CoV the spike protein is incorporated into the viral envelope The spike proteins of coronaviruses are large type I membrane glycoproteins that are responsible for both binding to recep tors on host cells and for membrane fusion The spike pro teins of some coronaviruses are cleaved into S1 and S2 sub units S proteins also contain important virus neutralizing epitopes and amino acid changes in the spike proteins can dramatically affect the virulence and in vitro host cell tropism of the virus 6 7 22 Still at present it is only the spike glycoprotein in which a considerable amount of mu tations has been documented Using the Blastp program to align three spike glycoproteins from humans we find little cue on the likelihood of which amino acid would mutate in SARS spike glycoprotein In the past three years we have developed two models to analyze the primary structure in proteins 25 conducted a series of studies on mutations in different proteins 26 34 Our studies show there is a clearly probabilistic pattern in the amino acids which are subject to mutations In this 0196 9781 see front matter 2003 Elsevier Inc All rights reserved doi 10 1016 j peptides 2003 10 008 1838 G Wu S Yan Peptides 24 2003 1837 1845 study we use our model to analyze two spike glycoproteins from human coronavirus in order to gain the insight on the prediction of amino acid pairs being sensitive to mutations in human SARS CoV 2 Materials and methods The amino acid sequences of the spike glycoproteins were obtained from the Swiss Protein data bank 1 The ac cess number is P59594 for human SARS CoV with 3 point mutations 2 15 17 24 P15423 for human coronavirus strain 229E with 38 point mutations 3 5 10 12 20 23 and P36334 for human coronavirus strain OC43 with 80 point mutations 10 13 14 18 In order to determine the amino acid pairs probabilistically sensitive to mutations we conduct the following calculations 25 which is briefly described follows with the SARS CoV spike protein as the example 2 1 Amino acid pairs in spike proteins The spike protein from human SARS CoV consists of 1255 amino acids The first and second amino acids are con sidered as an amino acid pair the second and third as another pair the third and fourth until the 1254th and 1255th thus there are 1254 pairs Because there are 20 types of amino acids an amino acid pair can be composed from any of 20 types of amino acids so there are 400 types of theoretically possible amino acid pairs Again there are 1254 pairs in the spike protein which are more than 400 types of theoreti cally possible amino acid pairs clearly some of 400 types should appear more than once Meanwhile we may expect that some of 400 types are absent from the spike protein Similarly there are 1172 and 1352 amino acid pairs in the spike proteins from strain 229E and OC43 respectively 2 2 Actual frequency and randomly predicted frequency in amino acid pairs The randomly predicted frequency is governed by the simple permutation principle 9 For example there are 39 arginines R and 96 serines S in SARS CoV spike pro tein the random frequency of amino acid pair RS would be 3 39 1255 96 1254 1254 2 983 Actually we can find three RS s in the spike protein so the actual frequency of RS is 3 Hence we have three relationships between actual and predicted frequencies i e the actual frequency is smaller than equal to and larger than the predicted fre quency respectively 2 3 Randomly predictable present amino acid pairs in SARS CoV spike protein As described in the last section the predicted frequency of randomly present pair RS would be 3 and RS does appear 3 times in the spike protein so the presence of RS is randomly predictable 2 4 Randomly unpredictable present amino acid pairs in SARS CoV spike protein There are 84 alanines A in SARS CoV spike protein the frequency of random presence of AA would be 6 84 1255 83 1254 1254 5 555 i e there would be 6 AA s in the spike protein In fact the AA appears 10 times in the spike protein so the presence of AA is ran domly unpredictable This illustrates the case that the actual frequency of AA is larger than its predicted frequency Another case is that the actual frequency is smaller than the predicted one For example there are 91 valines V in the spike protein and the predicted frequency of AV is 6 84 1255 91 1254 1254 6 091 whereas the actual frequency is only three 2 5 Randomly predictable absent amino acid pairs in SARS CoV spike protein There are 11 tryptophans W in SARS CoV spike pro tein the frequency of random presence of RW would be 0 39 1255 11 1254 1254 0 342 i e the RW would not appear in the spike protein which is true in the real sit uation Thus the absence of RW is randomly predictable 2 6 Randomly unpredictable absent amino acid pairs in SARS CoV spike protein There are 99 threonines T in SARS CoV spike pro tein the frequency of random presence of RT would be 3 39 1255 99 1254 1254 3 076 i e there would be three RT s in the spike protein However no RT is found in this protein therefore the absence of RT from the spike protein is randomly unpredictable 2 7 Mutations in randomly predictable and unpredictable amino acid pairs A point missense mutation results in two amino acid pairs being substituted by another two As each pair has its actual and predicted frequencies the difference between them rep resents a probabilistic measure for the comparison in substi tuted and substituting amino acid pairs before and after mu tation After calculating the predicted frequency and com paring with the actual frequency we can classify the sub stituted amino acid pairs into the predictable unpredictable amino acid pairs 3 Results Table 1 details the appearance of theoretically possible types of amino acids in three spike proteins for example G Wu S Yan Peptides 24 2003 1837 1845 1839 Table 1 Number of theoretical types of amino acid pairs in the spike proteins from different human coronaviruses Appearance SARS CoV Strain 229E Strain OC43 Number Percentage Number Percentage Number Percentage 0 59 14 75 86 21 5 66 16 5 1 76 19 78 19 5 61 15 25 2 61 15 25 60 15 56 14 3 51 12 75 45 11 25 55 13 75 4 46 11 5 43 10 75 41 10 25 5 38 9 5 18 4 5 35 8 75 6 22 5 5 19 4 75 30 7 5 7 15 3 75 14 3 5 17 4 25 8 15 3 75 12 3 11 2 75 9 7 1 75 8 2 11 2 75 10 4 1 7 1 75 6 1 5 11 4 1 5 1 25 6 1 5 12 1 0 25 3 0 75 4 1 13 1 0 25 0 0 0 0 14 00 00 00 15 0 0 0 0 1 0 25 16 0 0 1 0 25 0 0 17 00 00 00 8 9 20 0 0 1 0 25 0 0 the third row shows how many types do not appear From the viewpoint of amino acid pairs no matter the length of a protein is the number of its theoretically possible types can not be more than 400 and therefore the difference between proteins is either how many types of theoretically possible amino acid pairs appear or how many times a theoretically possible type of amino acid pair repeats or both Table 1 shows 59 86 and 66 types are absent from the spike protein of SARS CoV strain 229E and strain OC43 third row in the Table respectively Still Table 1 shows that 76 78 and 61 types appear once in the spike protein of SARS CoV strain 229E and strain OC43 fourth row in the Table respectively and so on The absent types include 17 randomly predictable and 42 randomly unpredictable with regard to SARS CoV spike protein 37 randomly predictable and 49 randomly un predictable with regard to strain 229E spike protein and 12 randomly predictable and 54 randomly unpredictable with regard to strain OC43 spike protein Table 2 Occurrence of mutations with respect to randomly predictable and unpredictable amino acid pairs in the spike proteins from different human coronaviruses Spike protein Amino acid pairs Types Pairs Mutations Ratio Number Percentage Number Percentage Number Percentage Mutations types Mutations pairs SARS CoV Predictable 86 25 22 226 18 02 0 0 0 86 0 0 226 0 Unpredictable 255 74 78 1028 81 98 3 100 3 255 0 012 3 1028 0 003 Strain 229E Predictable 81 25 8 206 17 58 1 2 63 1 81 0 012 1 206 0 005 Unpredictable 233 74 2 966 82 42 37 97 37 37 233 0 159 37 966 0 038 Strain OC43 Predictable 97 29 04 286 21 15 4 5 4 97 0 041 4 286 0 014 Unpredictable 237 70 96 1066 78 85 76 95 76 237 0 321 76 1066 0 071 Still we can classify the present amino acid pairs as ran domly predictable and unpredictable with respect to theoret ically possible types and pairs because some theoretically possible types appear many times from row 5 to row 23 in Table 1 The columns 3 4 5 and 6 in Table 2 show how many predictable and unpredictable types and pairs in human spike proteins When corresponding the position of each mutation to predictable pairs and unpredictable pairs we find that a vast majority of mutations occurs at the un predictable pairs columns 7 8 9 and 10 in Table 2 Fig 1 shows the ratios of frequency difference AF PF versus mutation number per each type of substituted amino acid pairs in spike proteins It can be seen that there is a gen eral tendency in the ratios i e the larger the difference the higher the chance of mutation occurring Therefore the dif ference between actual and predicted frequencies indicates the potential chance of mutation occurring in amino acid pairs 1840 G Wu S Yan Peptides 24 2003 1837 1845 Fig 1 Ratios of difference between actual and predicted frequencies versus mutations per theoretically possible type of amino acid pair in the spike proteins from different human coronaviruses Fig 2 Sum of differences between actual and predicted frequencies in substituted and substituting amino acid pairs in the spike proteins from different human coronaviruses the data are presented as mean S E G Wu S Yan Peptides 24 2003 1837 1845 1841 Table 3 Sum of difference between actual and predicted frequencies with respect to amino acid pairs being substituted and substituting by mutations SARS CoV Sigma1 AF PF Strain 229E Sigma1 AF PF Strain OC43 Sigma1 AF PF mutation position Substituted pairs Substituting pairs Mutation position Substituted pairs Substituting pairs Mutation position Substituted pairs Substituting pairs 77 3 498642930 244 2 5 120 2 22 1 577 7 2 176 3 140 2 7 210 3 0 62 5 2 223 3 663 20 230 2 3 115 3 2 230 2 0 115 3 0 248 16 1 116 3 1 270 6 2 152 22 295 3 3 161 11 300 1 1 167 0 0 307 5 5 173 3 2 336 4 2 173 3 2 401 2 2 190 1 1 414 4 3 222 2 3 424 2 5 248 1 3 430 4 1 252 2 1 441 1 9 272 5 4 444 2 5 283 3 0 462 4 0 288 2 3 481 1 6 291 3 2 488 4 2 303 1 3 530 9 2 308 4 2 577 3 7 329 3 1 578 3 1 334 2 1 590 11 1 451 4 4 642 3 5 454 32 681 1 1 467 0 2 700 3 1 488 1 3 711 4 1 496 4 1 714 3 1 544 1 1 765 5 2 557 1 7 775 3 1 566 5 5 846 0 4 570 11 871 1 3 579 1 0 937 6 5 587 6 0 971 1 2 603 1 1 1005 1 3 612 0 5 630 3 4 641 12 665 5 3 694 15 700 0 2 728 3 2 758 4 1 783 0 0 802 1 0 817 3 0 824 2 2 833 3 2 884 4 9 896 3 4 912 1 2 915 3 2 933 2 2 944 22 955 1 8 955 1 1 969 3 1 975 0 1 1842 G Wu S Yan Peptides 24 2003 1837 1845 Table 3 Continued SARS CoV Sigma1 AF PF Strain 229E Sigma1 AF PF Strain OC43 Sigma1 AF PF mutation position Substituted pairs Substituting pairs Mutation position Substituted pairs Substituting pairs Mutation position Substituted pairs Substituting pairs 993 4 4 1012 2 6 1016 1 0 1039 4 1 1058 1 2 1059 2 4 1074 4 1 1089 4 0 1160 0 2 1189 1 2 1193 6 2 1197 11 1202 33 1211 3 0 1220 2 2 1231 0 2 1246 5 6 1265 0 3 1331 2 0 1342 3 2 AF actual frequency PF predicted frequency As the point missense mutations substitute one type of amino acid to another one we can gain some insight into the mutation tendency after comparing the difference be tween actual and predicted frequencies in substituted and substituting amino acid pairs For the numerical analysis we calculate the difference between actual frequency AF and predicted frequency PF in amino acid pairs before and after mutation i e Sigma1 AF PF For instance a muta tion at position 244 substitutes I to T which results in two amino acid pairs DI and IW changing to DT and TW because the amino acid is D at position 243 and W at position 245 The actual frequency and predicted frequency are 7 and 5 for DI 1 and 1 for IW 2 and 6 for DT and 0 and 1 for TW respectively Thus the dif ference between actual frequency and predicted frequency is 2 with regard to the substituted amino acid pairs i e 7 5 1 1 2 and 5 with regard to the substituting amino acid pairs i e 2 6 0 1 5 In this way we can compare the frequency difference in the amino acid Table 4 Classification of substituted amino acid pairs with respect to mutations in the spike proteins from different human coronaviruses Spike protein Amino acid pairs Mutations in SARS CoV Mutations in strain 229E Mutations in strain OC43 I II Number Percentage Number Percentage Number Percentage Predictable AF PF AF PF 0 0 1 2 63 4 5 Unpredictable AF PF AF PF 1 33 33 12 31 58 19 23 75 AF PF AF PF 2 66 67 10 26 32 20 25 AF PF AF PF 0 0 10 26 32 23 28 75 AF PF AF PF 0 0 3 7 89 9 11 25 AF PF AF PF 0 0 2 5 26 5 6 25 AF actual frequency PF predicted frequency pairs affected by mutations Fig 2 shows the difference be tween actual and predicted frequencies in both substituted and substituting amino acid pairs in spike proteins It can be seen that the substituting pairs distribute more centrally and symmetrically than the substituted pairs do The sum of differences between actual and predicted frequencies is statistically smaller in substituting amino acid pairs than in substituted ones in Table 3 the Student s t test P 0 AF 0 PF 0 0 a 00 a 02 a 2 5 AF 0 PF 0 AF PF 00 a a 01 a 1 25 AF 0 PF 0 AF PF 0 0 a a 00 a 0 AF 0 PF 0 AF 0 AF PF 0 a 01 a 2 63 7 a 8 75 AF PF 0AF PF 00 0 0 0 0 0 AF PF PF 00 00 00 AF PF 0AF PF 1 0 1 2 63 0 0 AF PF AF negationslash 0AF PF AF negationslash 01 a 33 33 9 a 23 68 6 a 7 5 AF 0 0 a 02 a 5 26 11 a 13 75 AF PF 0 a 011 a 28 95 26 a 32 5 AF PF 0 AF PF 0 0 0 0 0 3 3 75 AF PF AF PF 0 0 8 21 05 9 11 25 AF PF 0 AF PF 1 33 33 6 15 79 11 13 75 a It indicates one or both substituting amino acid pairs with their actual frequency smaller than predicted frequency These amino acid pairs are 66 67 60 52 and 69 of total amino acid pairs in SARS CoV strain 229E and strain OC43 respectively dicted frequency in one or both substituted pairs the first three rows in unpredictable pairs Comparing the first three rows with the last two rows in unpredictable pairs we can see that the mutations are more likely to target the pairs whose actual frequencies are larger than predicted frequen cies Therefore Table 4 suggests which type of amino acid pairs are more likely to be substituted i e the different sen sitivities of amino acid pairs to mutations in spike proteins Table 5 shows in which types of amino acid pairs the mu tations are likely or unlikely to form in spike proteins We can find that more than 60 of mutations result in one or both substituting pairs whose actual frequencies are smaller than their predicted frequencies Taking the results in both Tables 3 and 4 into account the mutations are likely to at tack the pairs whose actual frequencies are larger than their predicted ones and the consequences of mutations are likely to form the pairs whose actual frequencies are smaller than their predicted ones In such a manner the mutations re duce the difference between actual and predicted frequen cies Fig 2 4 Discussion In this study we have analyzed the amino acid pairs af fected by mutations in three spike proteins in order to gain some insight into the possible mutations from SARS CoV Firstly the present results demonstrate that the randomly unpredictable amino acid pairs are more sensitive to the mu tations Table 2 although these 3 spike proteins are con structed by different types of amino acid pairs which repeat different times Table 1 Furthermore the larger the differ ence between actual and predicted frequencies is the higher the chance of mutation occurring is Fig 1 The effect in duced by mutations is to reduce the difference between ac tual and predicted frequencies Fig 2 Finally the amino acid pairs whose actual frequencies are larger than their pre dicted frequencies are more likely to be targeted by muta tions Table 4 whereas the amino acid pairs whose actual frequencies are smaller than their predicted frequencies are more likely to be formed after mutations Table 5 These findings are identical to our recently publications 26 34 Combining the results with our previous studies our model suggests that the mutations go along a pathway which is probabilistically more likely to occur As such a pathway is less energy and time consuming in fact the mutations represent a process of degeneration inducing hu man diseases Although this study shows that the mutations in the spike proteins from strains 229E and OC43 go along the direction of degeneration i e the mutations go along a probabilistically easy pathway the documented evidence in literature still cannot suggest whether or not the mutations belong to degeneration in the spike protein from human SARS CoV If the potential mutations in the spike protein from SARS CoV would go along a probabilistically easy path way according to t