- Email: [email protected]
Diagnostic performance of an optimized transcriptomic signature of risk of tuberculosis in cryopreserved peripheral blood mononuclear cells
Accepted Manuscript Diagnostic performance of an optimized transcriptomic signature of risk of tuberculosis in cryopreserved peripheral blood mononuclear cells Fatoumatta Darboe, Stanley Kimbung Mbandi, Ethan G. Thompson, Michelle Fisher, Miguel Rodo, Michele van Rooyen, Elizabeth Filander, Nicole Bilek, Simbarashe Mabwe, Mark Hatherill, Daniel E. Zak, Adam Penn-Nicholson, Thomas J. Scriba, Sindile Matiwane, Lungisa Jaxa, Noncedo Xoyana, Constance Schreuder, Janelle Botes, Hadn Africa, Lebohang Makhethe, Marcia Steyn PII:
S1472-9792(17)30321-9
DOI:
10.1016/j.tube.2017.11.001
Reference:
YTUBE 1640
To appear in:
Tuberculosis
Received Date: 28 July 2017 Revised Date:
30 October 2017
Accepted Date: 2 November 2017
Please cite this article as: Darboe F, Mbandi SK, Thompson EG, Fisher M, Rodo M, van Rooyen M, Filander E, Bilek N, Mabwe S, Hatherill M, Zak DE, Penn-Nicholson A, Scriba TJ, The SATVI Clinical Immunology Team, Matiwane S, Jaxa L, Xoyana N, Schreuder C, Botes J, Africa H, Makhethe L, Steyn M, Diagnostic performance of an optimized transcriptomic signature of risk of tuberculosis in cryopreserved peripheral blood mononuclear cells, Tuberculosis (2017), doi: 10.1016/ j.tube.2017.11.001. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT 1
Diagnostic performance of an optimized transcriptomic signature of risk of
2
tuberculosis in cryopreserved peripheral blood mononuclear cells
3 Fatoumatta Darboea*, Stanley Kimbung Mbandi a*, Ethan G. Thompsonb, Michelle
5
Fishera, Miguel Rodoa, c, Michele van Rooyena, Elizabeth Filandera, Nicole Bileka,
6
Simbarashe Mabwea, Mark Hatherilla, Daniel E. Zakb, Adam Penn-Nicholsona$,
7
Thomas J. Scribaa$ and the SATVI Clinical Immunology Team#
RI PT
4
8 #
The SATVI Clinical Immunology Team: Sindile Matiwane1, Lungisa Jaxa1 (nee
SC
9
Nkantsu), Noncedo Xoyana1, Constance Schreuder1, Janelle Botes1, Hadn Africa1,
11
Lebohang Makhethe1, Marcia Steyn1.
M AN U
10
12 13
a
14
Molecular Medicine and Division of Immunology, Department of Pathology, University
15
of Cape Town, 7925, Cape Town, South Africa.
TE D
South African Tuberculosis Vaccine Initiative, Institute of Infectious Disease and
16 b
The Center for Infectious Disease Research, Seattle, 98109, WA, USA.
18 19
c
20
South Africa
EP
17
21
AC C
Department of Statistical Sciences, University of Cape Town, Cape Town, 7925,
22
*These authors contributed equally
23
$
Shared senior authors
24 25
Send correspondence to Thomas J. Scriba: [email protected], Tel: +27 021
26
406 6427
27
Keywords: biomarker, correlate of risk, diagnostic, whole blood, PBMC, mRNA
ACCEPTED MANUSCRIPT A staggering 23% of the global population is estimated to be infected with
29
Mycobacterium tuberculosis (M.tb) [1], representing an enormous reservoir of
30
individuals at risk of progressing to active tuberculosis (TB) disease. Identification of
31
those at risk of TB disease would allow targeted preventive therapy, potentially
32
curtailing transmission [2]. We recently identified and validated a 16-gene whole
33
blood transcriptomic correlate of risk (CoR) signature that can predict active TB
34
disease in M.tb-infected individuals more than a year before disease manifestations
35
and diagnosis [3]. The transcriptomic signature is measured by microfluidic
36
quantitative reverse transcription polymerase chain reaction (qRT-PCR), comprising
37
a set of 57 primer-probes (47 primer-probes detecting 47 transcripts representing 16
38
interferon (IFN) response genes, and 10 primer-probes used as a reference for
39
standardization). To improve high-throughput processing we sought to trim 9 primer-
40
probes from the signature such that the 48 primer-probes can be conveniently
41
accommodated in duplicate in a typical 96.96 Fluidigm gene expression platform.
M AN U
SC
RI PT
28
TE D
42
We selected the transcripts for removal based on 3 criteria: (1) those that showed the
44
least reproducibility in qRT-PCR assays, (2) those that yielded redundant signal to
45
the overall signature, and (3) when transcripts were ranked by the number of pairings
46
they form with other transcripts in the signature only those that form very few pairs
47
were removed (thus contributing little to the signature; data not shown). Based on
48
these criteria, we removed nine transcripts detected by 9 primer-probes, which
49
represented five IFN response genes.
50
The reduced 48-primer-probe signature comprised 38 primer-probes representing 11
51
IFN response genes, and the 10 reference primer-probes. The removed primer-
52
probes represented the genes, SEPT4, ANKRD22, APOL1, FCGR1A and GBP4. We
53
compared the prognostic performance of the original, 57 primer-probe signature to
AC C
EP
43
ACCEPTED MANUSCRIPT the reduced, 48 primer-probe signature (Adolescent Cohort Study (ACS) 11-gene
55
signature), measured by microfluidic qRT-PCR on samples collected from adolescent
56
progressors and controls that comprised the original training and test sets from the
57
ACS, in which the signature was discovered [3]. Ribonucleic acid (RNA) was isolated
58
with the PAXgene™ blood RNA kit (QIAGEN); complimentary DNA (cDNA) was
59
synthesised by reverse transcription and pre-amplified using PCR, as described [3].
60
Signature transcripts were quantified using TaqMan primer-probes on 96.96 Gene
61
Expression Chips (Fluidigm) on a Biomark HD multiplex microfluidic qRT-PCR
62
instrument. Signature scores were calculated as previously described [3]. Risk
63
scores of the original signature and reduced ACS 11-gene signature were essentially
64
identical (Spearman r=0.99, p=0.0 (Figure 1A)). Predictive performance of TB risk,
65
expressed as receiver operating characteristic (ROC) area under the curve (AUC)
66
also did not differ (Figure 1B). Thus, reduction of the CoR signature to 48 primer-
67
probes resulted in equivalent performance.
M AN U
SC
RI PT
54
TE D
68
Next, we sought to determine the performance of the 48 primer-probe whole blood
70
ACS 11-gene signature when measured in peripheral blood mononuclear cell
71
(PBMC) samples. Reliable measurement of the signature using RNA from PBMC
72
would allow testing and/or further validation using biobanked samples from historical
73
clinical studies. Because the signature score increases most proximal to the onset of
74
TB disease, we evaluated diagnostic performance using RNA from PAXgene whole
75
blood and cryopreserved PBMC samples (RNEasy mini kit, QIAGEN) collected in
76
parallel from 30 HIV-uninfected adults with newly diagnosed active TB disease
77
(sputum Xpert MTB/RIF-positive) and 30 healthy, M.tb-infected (QuantiFERON Gold
78
In-tube-positive, QIAGEN) adult controls.
79
We first compared messenger RNA (mRNA) expression differences for each IFN
80
response transcript in the reduced signature between TB cases and controls in whole
81
blood and PBMC. The difference in median mRNA expression between TB cases
AC C
EP
69
ACCEPTED MANUSCRIPT and controls (delta cycle threshold (Ct)) was estimated and 95% confidence intervals
83
(CI) of the median were obtained using the rank inversion method [4], with code from
84
the R package quantreg. All 38 IFN response transcripts were expressed at higher
85
levels in TB cases than in controls, whether measured in whole blood or PBMC
86
(Figure 1C). However, for transcripts including SCARF1, GBP2, GBP5, BATF2 and
87
TRAFD1, the relative differences in transcript expression between TB cases and
88
controls were greater in whole blood than in PBMC. This suggests a substantial
89
contribution of cell subsets present in whole blood but absent in PBMC, such as
90
granulocytes, to the disease-associated IFN response, as previously described [5].
91
Nevertheless, the ACS 11-gene signature classified samples from TB cases and
92
healthy controls with equivalent accuracy (Fig 1D). ROC-AUCs for whole blood and
93
PBMC were indistinguishable at 0.97 (95% CI 0.91-1.00) and 0.98 (0.95-1.00),
94
respectively. Risk signature scores were markedly lower in PBMC than in whole
95
blood regardless of disease status, again reflecting the comparatively weaker IFN
96
response transcriptomic signal in PBMC (Figure 1E). This necessitates the use of
97
independent, sample type-specific thresholds for classification of signature-positive
98
or negative samples.
99
The comparable performance of the CoR signature in discriminating active TB cases
100
from healthy controls in PBMC and whole blood demonstrates feasibility for applying
101
this signature to ascertain risk of TB in historical studies that have prospectively
102
collected cryopreserved PBMC.
SC
M AN U
TE D
EP
AC C
103
RI PT
82
104
In conclusion, the ACS 11-gene, 48-primer-probe transcriptomic signature has
105
diagnostic utility for TB with identical accuracy as the original 16-gene, 57- primer-
106
probe signature, and can be measured in cryopreserved PBMC.
107 108
Funding
ACCEPTED MANUSCRIPT This work was supported by the Strategic Health Innovation Partnerships (SHIP) Unit
110
of the South African Medical Research Council with funds received from the South
111
African Department of Science and Technology.
112
F.D. is supported by the Margaret McNamara educational grant for women in
113
developing countries.
RI PT
109
114 Conflicts of Interest
116
EF, MH, NB, MR, MF, SM, FD and MvR declare no conflicts of interest.
117
APN, DEZ, EGT and TJS report pending patent of the gene signature.
118
DEZ and TJS report receiving grants from the South African Medical Research
119
Council, National Institutes of Health and/or Bill and Melinda Gates Foundation
120
related to the gene signature during course of study.
M AN U
SC
115
121 122
References
123
[1]
TE D
Houben RMGJ, Dodd PJ. The Global Burden of Latent Tuberculosis Infection:
124
A Re-estimation Using Mathematical Modelling. PLoS Med 2016;13:1–13.
125
doi:10.1371/journal.pmed.1002152. [2]
Penn-Nicholson A, Scriba TJ, Hatherill M, White RG, Sumner T. A novel blood
EP
126
test for tuberculosis prevention and treatment. South African Med J
128
2016;107:4. doi:10.7196/SAMJ.2017.v107i1.12230.
129 130 131 132
AC C
127
[3]
al. A blood RNA signature for tuberculosis disease risk: a prospective cohort study. Lancet 2016;6736:1–11. doi:10.1016/S0140-6736(15)01316-1.
[4]
133 134
Zak DE, Penn-Nicholson A, Scriba TJ, Thompson E, Suliman S, Amon LM, et
Koenker R. Confidence intervals for Quantile Regression. Asymptomatic Stat. Proc. 5th Prague Symp., 1994.
[5]
Berry MPR, Graham CM, McNab FW, Xu Z, Bloch S a a, Oni T, et al. An
135
interferon-inducible neutrophil-driven blood transcriptional signature in human
136
tuberculosis. Nature 2010;466:973–7. doi:10.1038/nature09247.
ACCEPTED MANUSCRIPT 137 138 139 140
RI PT
141 Figure 1. Performance of the reduced ACS 11-gene signature of TB risk as a
143
classifier of TB disease from M.tb-infection in PBMC. (A) Correlation of the
144
signature scores generated from the original 57 primer-probe (16 genes) and the
145
reduced, 48 primer-probe (11-genes) qRT-PCR transcriptomic signatures of risk of
146
TB disease in progressors and controls from the Adolescent Cohort Study (ACS).
147
The Spearman correlation coefficient is shown. (B) Performance of the original and
148
reduced CoR signatures in classifying progressor and control samples from the
149
Adolescent Cohort Study. (C) Differences in transcript expression between TB cases
150
(Xpert MTB/RIF-positive) and healthy controls (QuantiFERON-positive) when
151
measured in whole blood or PBMC samples. Dots represent medians and the bars
152
95% CI for each transcript, identified by its TaqMan primer-probe set. (D) CoR
153
signature scores measured in whole blood or PBMC from active TB cases or healthy
154
QuantiFERON-positive persons. Horizontal lines represent medians, the box
155
represents the IQR and whiskers the range. (E) Receiver operating characteristic
156
curves illustrating diagnostic performance of the 48-primer-probe ACS 11-gene
157
signature in distinguishing individuals with active TB disease from healthy
158
QuantiFERON-positive controls when measured in RNA from whole blood or PBMC
159
samples.
160
AC C
EP
TE D
M AN U
SC
142
40 Primer Probes
D
100
TE D
C
RI PT
20
Sample Type Whole Blood PBMC
80
TB cases
SC
Sensitivity
B
100
80
40
0.2
Spearman r = 0.99 p = 0.0
0 20 40 60 80 57 primers: signature score (%)
20
0
QFT+ controls 100 0.0
E
Whole blood PBMC
0.4
0.2
0.0
0.2
0.2 ETV7.Hs00903228_m1 TAP1.Hs00897093_g1 TAP1.Hs00388675_m1 GBP5−j4 GBP5.Hs00369472_m1 BATF2.Hs00912736_m1 FCGR1C.Hs00417598_m1 TRAFD1.Hs00938765_m1
M AN U
48 primers: signature score (%) 60
Sensitivity
AC C
60 STAT1.Hs01013995_g1 STAT1.Hs01014002_m1 STAT1.Hs01013998_m1 STAT1.Hs01013993_m1 STAT1.Hs01013997_m1 STAT1.Hs01013996_m1 STAT1.Hs01013994_m1 STAT1.Hs01013992_g1 STAT1.Hs01013991_m1 STAT1.Hs01013989_m1 STAT1.Hs01014000_m1 GBP2.Hs00894837_m1 GBP2.Hs00894846_g1 GBP2−j1 GBP2.Hs00894842_g1 GBP2.Hs00894840_mH SERPING1.Hs00934330_m1 SERPING1.Hs00934329_m1 SERPING1.Hs00935959_m1 SERPING1.Hs00934328_g1 SERPING1.Hs00163781_m1 SCARF1.Hs01092483_m1 SCARF1.Hs01092485_g1 SCARF1.Hs01092482_g1 SCARF1.Hs00186503_m1 GBP1−j1 GBP1.Hs00266717_m1 GBP1.Hs00977005_m1 ETV7−j2 ETV7.Hs00903230_g1
Relative diference TB-LTBI (dCT) −4 −2 0 0
EP
Signature score (%)
A 1.0
0.8
ACCEPTED MANUSCRIPT
0.6
0.4 AUC (95% CI)
0.4 0.6 1 - Specificity
0.4 0.6 1 - Specificity
p-value
57 primers 0.83 (0.77-0.83) 1.1 x 10-14
48 primers 0.84 (0.77-0.83) 6.0 x 10-15
0.0 0.8
0.8
1.0
1.0
0.8
0.6
p-value AUC (95% CI) Whole blood 0.97 (0.91-1.00) 7.5 x 10-10 PBMC 0.98 (0.95-1.00) 2.1 x 10-10
0.0
1.0
S1472-9792(17)30321-9
DOI:
10.1016/j.tube.2017.11.001
Reference:
YTUBE 1640
To appear in:
Tuberculosis
Received Date: 28 July 2017 Revised Date:
30 October 2017
Accepted Date: 2 November 2017
Please cite this article as: Darboe F, Mbandi SK, Thompson EG, Fisher M, Rodo M, van Rooyen M, Filander E, Bilek N, Mabwe S, Hatherill M, Zak DE, Penn-Nicholson A, Scriba TJ, The SATVI Clinical Immunology Team, Matiwane S, Jaxa L, Xoyana N, Schreuder C, Botes J, Africa H, Makhethe L, Steyn M, Diagnostic performance of an optimized transcriptomic signature of risk of tuberculosis in cryopreserved peripheral blood mononuclear cells, Tuberculosis (2017), doi: 10.1016/ j.tube.2017.11.001. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT 1
Diagnostic performance of an optimized transcriptomic signature of risk of
2
tuberculosis in cryopreserved peripheral blood mononuclear cells
3 Fatoumatta Darboea*, Stanley Kimbung Mbandi a*, Ethan G. Thompsonb, Michelle
5
Fishera, Miguel Rodoa, c, Michele van Rooyena, Elizabeth Filandera, Nicole Bileka,
6
Simbarashe Mabwea, Mark Hatherilla, Daniel E. Zakb, Adam Penn-Nicholsona$,
7
Thomas J. Scribaa$ and the SATVI Clinical Immunology Team#
RI PT
4
8 #
The SATVI Clinical Immunology Team: Sindile Matiwane1, Lungisa Jaxa1 (nee
SC
9
Nkantsu), Noncedo Xoyana1, Constance Schreuder1, Janelle Botes1, Hadn Africa1,
11
Lebohang Makhethe1, Marcia Steyn1.
M AN U
10
12 13
a
14
Molecular Medicine and Division of Immunology, Department of Pathology, University
15
of Cape Town, 7925, Cape Town, South Africa.
TE D
South African Tuberculosis Vaccine Initiative, Institute of Infectious Disease and
16 b
The Center for Infectious Disease Research, Seattle, 98109, WA, USA.
18 19
c
20
South Africa
EP
17
21
AC C
Department of Statistical Sciences, University of Cape Town, Cape Town, 7925,
22
*These authors contributed equally
23
$
Shared senior authors
24 25
Send correspondence to Thomas J. Scriba: [email protected], Tel: +27 021
26
406 6427
27
Keywords: biomarker, correlate of risk, diagnostic, whole blood, PBMC, mRNA
ACCEPTED MANUSCRIPT A staggering 23% of the global population is estimated to be infected with
29
Mycobacterium tuberculosis (M.tb) [1], representing an enormous reservoir of
30
individuals at risk of progressing to active tuberculosis (TB) disease. Identification of
31
those at risk of TB disease would allow targeted preventive therapy, potentially
32
curtailing transmission [2]. We recently identified and validated a 16-gene whole
33
blood transcriptomic correlate of risk (CoR) signature that can predict active TB
34
disease in M.tb-infected individuals more than a year before disease manifestations
35
and diagnosis [3]. The transcriptomic signature is measured by microfluidic
36
quantitative reverse transcription polymerase chain reaction (qRT-PCR), comprising
37
a set of 57 primer-probes (47 primer-probes detecting 47 transcripts representing 16
38
interferon (IFN) response genes, and 10 primer-probes used as a reference for
39
standardization). To improve high-throughput processing we sought to trim 9 primer-
40
probes from the signature such that the 48 primer-probes can be conveniently
41
accommodated in duplicate in a typical 96.96 Fluidigm gene expression platform.
M AN U
SC
RI PT
28
TE D
42
We selected the transcripts for removal based on 3 criteria: (1) those that showed the
44
least reproducibility in qRT-PCR assays, (2) those that yielded redundant signal to
45
the overall signature, and (3) when transcripts were ranked by the number of pairings
46
they form with other transcripts in the signature only those that form very few pairs
47
were removed (thus contributing little to the signature; data not shown). Based on
48
these criteria, we removed nine transcripts detected by 9 primer-probes, which
49
represented five IFN response genes.
50
The reduced 48-primer-probe signature comprised 38 primer-probes representing 11
51
IFN response genes, and the 10 reference primer-probes. The removed primer-
52
probes represented the genes, SEPT4, ANKRD22, APOL1, FCGR1A and GBP4. We
53
compared the prognostic performance of the original, 57 primer-probe signature to
AC C
EP
43
ACCEPTED MANUSCRIPT the reduced, 48 primer-probe signature (Adolescent Cohort Study (ACS) 11-gene
55
signature), measured by microfluidic qRT-PCR on samples collected from adolescent
56
progressors and controls that comprised the original training and test sets from the
57
ACS, in which the signature was discovered [3]. Ribonucleic acid (RNA) was isolated
58
with the PAXgene™ blood RNA kit (QIAGEN); complimentary DNA (cDNA) was
59
synthesised by reverse transcription and pre-amplified using PCR, as described [3].
60
Signature transcripts were quantified using TaqMan primer-probes on 96.96 Gene
61
Expression Chips (Fluidigm) on a Biomark HD multiplex microfluidic qRT-PCR
62
instrument. Signature scores were calculated as previously described [3]. Risk
63
scores of the original signature and reduced ACS 11-gene signature were essentially
64
identical (Spearman r=0.99, p=0.0 (Figure 1A)). Predictive performance of TB risk,
65
expressed as receiver operating characteristic (ROC) area under the curve (AUC)
66
also did not differ (Figure 1B). Thus, reduction of the CoR signature to 48 primer-
67
probes resulted in equivalent performance.
M AN U
SC
RI PT
54
TE D
68
Next, we sought to determine the performance of the 48 primer-probe whole blood
70
ACS 11-gene signature when measured in peripheral blood mononuclear cell
71
(PBMC) samples. Reliable measurement of the signature using RNA from PBMC
72
would allow testing and/or further validation using biobanked samples from historical
73
clinical studies. Because the signature score increases most proximal to the onset of
74
TB disease, we evaluated diagnostic performance using RNA from PAXgene whole
75
blood and cryopreserved PBMC samples (RNEasy mini kit, QIAGEN) collected in
76
parallel from 30 HIV-uninfected adults with newly diagnosed active TB disease
77
(sputum Xpert MTB/RIF-positive) and 30 healthy, M.tb-infected (QuantiFERON Gold
78
In-tube-positive, QIAGEN) adult controls.
79
We first compared messenger RNA (mRNA) expression differences for each IFN
80
response transcript in the reduced signature between TB cases and controls in whole
81
blood and PBMC. The difference in median mRNA expression between TB cases
AC C
EP
69
ACCEPTED MANUSCRIPT and controls (delta cycle threshold (Ct)) was estimated and 95% confidence intervals
83
(CI) of the median were obtained using the rank inversion method [4], with code from
84
the R package quantreg. All 38 IFN response transcripts were expressed at higher
85
levels in TB cases than in controls, whether measured in whole blood or PBMC
86
(Figure 1C). However, for transcripts including SCARF1, GBP2, GBP5, BATF2 and
87
TRAFD1, the relative differences in transcript expression between TB cases and
88
controls were greater in whole blood than in PBMC. This suggests a substantial
89
contribution of cell subsets present in whole blood but absent in PBMC, such as
90
granulocytes, to the disease-associated IFN response, as previously described [5].
91
Nevertheless, the ACS 11-gene signature classified samples from TB cases and
92
healthy controls with equivalent accuracy (Fig 1D). ROC-AUCs for whole blood and
93
PBMC were indistinguishable at 0.97 (95% CI 0.91-1.00) and 0.98 (0.95-1.00),
94
respectively. Risk signature scores were markedly lower in PBMC than in whole
95
blood regardless of disease status, again reflecting the comparatively weaker IFN
96
response transcriptomic signal in PBMC (Figure 1E). This necessitates the use of
97
independent, sample type-specific thresholds for classification of signature-positive
98
or negative samples.
99
The comparable performance of the CoR signature in discriminating active TB cases
100
from healthy controls in PBMC and whole blood demonstrates feasibility for applying
101
this signature to ascertain risk of TB in historical studies that have prospectively
102
collected cryopreserved PBMC.
SC
M AN U
TE D
EP
AC C
103
RI PT
82
104
In conclusion, the ACS 11-gene, 48-primer-probe transcriptomic signature has
105
diagnostic utility for TB with identical accuracy as the original 16-gene, 57- primer-
106
probe signature, and can be measured in cryopreserved PBMC.
107 108
Funding
ACCEPTED MANUSCRIPT This work was supported by the Strategic Health Innovation Partnerships (SHIP) Unit
110
of the South African Medical Research Council with funds received from the South
111
African Department of Science and Technology.
112
F.D. is supported by the Margaret McNamara educational grant for women in
113
developing countries.
RI PT
109
114 Conflicts of Interest
116
EF, MH, NB, MR, MF, SM, FD and MvR declare no conflicts of interest.
117
APN, DEZ, EGT and TJS report pending patent of the gene signature.
118
DEZ and TJS report receiving grants from the South African Medical Research
119
Council, National Institutes of Health and/or Bill and Melinda Gates Foundation
120
related to the gene signature during course of study.
M AN U
SC
115
121 122
References
123
[1]
TE D
Houben RMGJ, Dodd PJ. The Global Burden of Latent Tuberculosis Infection:
124
A Re-estimation Using Mathematical Modelling. PLoS Med 2016;13:1–13.
125
doi:10.1371/journal.pmed.1002152. [2]
Penn-Nicholson A, Scriba TJ, Hatherill M, White RG, Sumner T. A novel blood
EP
126
test for tuberculosis prevention and treatment. South African Med J
128
2016;107:4. doi:10.7196/SAMJ.2017.v107i1.12230.
129 130 131 132
AC C
127
[3]
al. A blood RNA signature for tuberculosis disease risk: a prospective cohort study. Lancet 2016;6736:1–11. doi:10.1016/S0140-6736(15)01316-1.
[4]
133 134
Zak DE, Penn-Nicholson A, Scriba TJ, Thompson E, Suliman S, Amon LM, et
Koenker R. Confidence intervals for Quantile Regression. Asymptomatic Stat. Proc. 5th Prague Symp., 1994.
[5]
Berry MPR, Graham CM, McNab FW, Xu Z, Bloch S a a, Oni T, et al. An
135
interferon-inducible neutrophil-driven blood transcriptional signature in human
136
tuberculosis. Nature 2010;466:973–7. doi:10.1038/nature09247.
ACCEPTED MANUSCRIPT 137 138 139 140
RI PT
141 Figure 1. Performance of the reduced ACS 11-gene signature of TB risk as a
143
classifier of TB disease from M.tb-infection in PBMC. (A) Correlation of the
144
signature scores generated from the original 57 primer-probe (16 genes) and the
145
reduced, 48 primer-probe (11-genes) qRT-PCR transcriptomic signatures of risk of
146
TB disease in progressors and controls from the Adolescent Cohort Study (ACS).
147
The Spearman correlation coefficient is shown. (B) Performance of the original and
148
reduced CoR signatures in classifying progressor and control samples from the
149
Adolescent Cohort Study. (C) Differences in transcript expression between TB cases
150
(Xpert MTB/RIF-positive) and healthy controls (QuantiFERON-positive) when
151
measured in whole blood or PBMC samples. Dots represent medians and the bars
152
95% CI for each transcript, identified by its TaqMan primer-probe set. (D) CoR
153
signature scores measured in whole blood or PBMC from active TB cases or healthy
154
QuantiFERON-positive persons. Horizontal lines represent medians, the box
155
represents the IQR and whiskers the range. (E) Receiver operating characteristic
156
curves illustrating diagnostic performance of the 48-primer-probe ACS 11-gene
157
signature in distinguishing individuals with active TB disease from healthy
158
QuantiFERON-positive controls when measured in RNA from whole blood or PBMC
159
samples.
160
AC C
EP
TE D
M AN U
SC
142
40 Primer Probes
D
100
TE D
C
RI PT
20
Sample Type Whole Blood PBMC
80
TB cases
SC
Sensitivity
B
100
80
40
0.2
Spearman r = 0.99 p = 0.0
0 20 40 60 80 57 primers: signature score (%)
20
0
QFT+ controls 100 0.0
E
Whole blood PBMC
0.4
0.2
0.0
0.2
0.2 ETV7.Hs00903228_m1 TAP1.Hs00897093_g1 TAP1.Hs00388675_m1 GBP5−j4 GBP5.Hs00369472_m1 BATF2.Hs00912736_m1 FCGR1C.Hs00417598_m1 TRAFD1.Hs00938765_m1
M AN U
48 primers: signature score (%) 60
Sensitivity
AC C
60 STAT1.Hs01013995_g1 STAT1.Hs01014002_m1 STAT1.Hs01013998_m1 STAT1.Hs01013993_m1 STAT1.Hs01013997_m1 STAT1.Hs01013996_m1 STAT1.Hs01013994_m1 STAT1.Hs01013992_g1 STAT1.Hs01013991_m1 STAT1.Hs01013989_m1 STAT1.Hs01014000_m1 GBP2.Hs00894837_m1 GBP2.Hs00894846_g1 GBP2−j1 GBP2.Hs00894842_g1 GBP2.Hs00894840_mH SERPING1.Hs00934330_m1 SERPING1.Hs00934329_m1 SERPING1.Hs00935959_m1 SERPING1.Hs00934328_g1 SERPING1.Hs00163781_m1 SCARF1.Hs01092483_m1 SCARF1.Hs01092485_g1 SCARF1.Hs01092482_g1 SCARF1.Hs00186503_m1 GBP1−j1 GBP1.Hs00266717_m1 GBP1.Hs00977005_m1 ETV7−j2 ETV7.Hs00903230_g1
Relative diference TB-LTBI (dCT) −4 −2 0 0
EP
Signature score (%)
A 1.0
0.8
ACCEPTED MANUSCRIPT
0.6
0.4 AUC (95% CI)
0.4 0.6 1 - Specificity
0.4 0.6 1 - Specificity
p-value
57 primers 0.83 (0.77-0.83) 1.1 x 10-14
48 primers 0.84 (0.77-0.83) 6.0 x 10-15
0.0 0.8
0.8
1.0
1.0
0.8
0.6
p-value AUC (95% CI) Whole blood 0.97 (0.91-1.00) 7.5 x 10-10 PBMC 0.98 (0.95-1.00) 2.1 x 10-10
0.0
1.0