Easy Identification of Pathogens via Binding to Glycosylated Materials
Lucienne Otten and Matthew I. Gibson
l.c.otten@warwick.ac.uk
@smrlam
www2.warwick.ac.uk/go/gibsongroup
@LabGibson
Background and current challenges •  The prevalence of protein-­‐carbohydrate interac6ons in important biological processes means that they are readily exploited by bacteria, bacterial toxins, viruses and immune system escape of malaria infected red blood cells1 •  Their importance in pathogen adhesion makes carbohydrates an aYrac6ve target for iden6fica6on and detec6on. However care needs to be taken to ensure that detec6on is specific and not influenced by binding of contaminants from biological samples •  A carbohydrate array can be designed that is highly specific for the toxin but this can be challenging and expensive as it requires the purifica6on of oligosaccharides •  An ideal array would u6lise monosaccharides, but this is also a challenge due to the promiscuity of lec6ns.2 The use of a powerful data classifica6on algorithm can aid in the improvement of this system for iden6fica6on and detec6on purposes. Surface modifica6on Data Analysis Classifica6on of an unknown sample ( ) using (OH) 4
Hydrazide func6onalised 96-­‐well plates were further A B A in the procedure 30 mM
the ini6al data set is challenging as it could func6onalised with various glycans sugar
/0.5 % aniline
O
belong to summarised in A and Func6onalisa6on was confirmed NH 2 pH 5.5, 50 °C NHeither group (A). Linear discriminant NH
3
analysis NH(LDA) transforms the data set to using a modified drop shape analysis technique (B). 30 mM sugar
/0.5 % aniline
NH 2
NH
O
pH 5.5, 50 °C NH
NH
B
100
Fluorescence (RFU)
A
% Surface Spreading of Dye
(OH) 4
80
60
40
20
0
Hydrazide
Galactose
100
80
60
40
20
1500
and y plane thus aiding in the classifica6on of 1500
samples (B). AEer LDA the unknown sample can 1000
be classified as belonging to the green class. Glyceraldehyde
1E-3
0.01
0.1
Concentration of FITC-BSA (mg/ml)
Case 2: Bacteria Case 1: Toxins Hydrazide Surface
Galactose
Mannose
Galactose
An6bio6c resistance remains a global health concern with deaths as a result of an6microbial resistance due to outnumber those from cancer by 2050. Current bacterial iden6fica6on techniques are oEen 6me consuming, expensive or challenging and as such development of facile point-­‐of-­‐care diagnos6c tools remains an important research goal. •  Samples of 5 bacterial species were fluorescently labeled prior to incuba6on with several sugar surfaces before washing of unbound sample 1000
Cholera is caused by the secre6on of CTx from the bacteria Vibrio cholerae and is common in war-­‐torn countries. Infec6on causes symptoms similar to those seen in Concentration of FITC-BSA (mg/ml)
Surface
Ricin poisoning and thus toxin iden6fica6on is crucial in determining appropriate treatment. •  Binding profiles of fluorescently labeled cholera toxin, RCA120 (a non-­‐toxic deriva6ve of ricin) and several environmental lec6ns of plant origin to 4 sugar surfaces was determined B
A
Mannose
3000
Galactose
Glucose
2500
PNA
50% Gal: 50% Man
CTx
2000
SBA
RCA120
1500
500
SBA
DBA
0
CTx
DBA
PNA RCA120 SBA
B
LD1
A Lectin
PNA
CTx
DBA
SBA
RCA120
RCA120
LDA SBA
PNA
DBA
LD1
LD1
•  Linear discriminant analysis of the original binding data produced a model in which all toxins were well resolved from each other. Hydrazide
500
Glyceraldehyde
Otten, L., and Gibson, M.I. RSC Advances, 2015
500
1E-3
0.01
0.1
RCA120
5
0
LD2
10
DBA
0
-4
-2
LD2
2
4
Fluorescence (RFU)
15
0
2000
Hydrazide Surface
Galactose
Mannose
2000
Surface
C
Fluorescence (RFU)
% Surface Spreading of Dye
Cminimise the overlap between classes in the x B
A
LDA -60
-40
-20
5
10
15
20
25
30
3000
Otten, L., and Gibson, M.I. RSC Advances, 2015
Conclusion: The model was able to correctly iden.fy blind samples of all lec.ns and was able to detect the presence of a toxin even in the presence of another lec6n. Background References
1.  Holgerrson et al., Immunology and Cell Biology, 2005. 2.  hYp://www.func6onalglycomics.org 3.  Godula and Bertozzi, J. AM. CHEM. SOC., 2010 Acknowledgements
•  Dr Ma0hew Gibson • Dr Elizabeth Fullam • The Gibson Group Top10
30
Surface
Man
0
Lac
25
1000
Glyc
20
Top10
E. coli
GlcNAc
15
smeg
M. smegmatis
Glc
10
Putida
P. putida
Gal
5
K12
M. marinum
Mar
Dex
-20
E. coli
K12
Cel
-40
group
Ara
-60
2000
LD2 5
0
-80
Fluoresence (RFU)
10
LD2
0
-4
-2
LD2
2
4
15
-80
PNA
•  Linear discriminant analysis produced a model in which all species analysed LD1
showed dis6nct profiles. LD1 Otten, L., Fullam, E., Gibson, M.I. in preparation
Conclusion: The model was able to correctly iden.fy all bacterial species with 92 % accuracy and was able to iden.fy a completely blind sample of one of the bacterial species.