Optical forward-scattering for identification of bacteria within microcolonies

Optical forward-scattering for identification of bacteria within microcolonies.

Antoine CUERJoe-Loïc KODJA

Arthur LEFEBVREFlorian LICARIRobin LOUVET

Anil NARASSIGUIN

Mathieu DUPOYPierre MARCOUX

Frédéric MALLARD

33rdrd International Conference on Bio-Sensing Technology International Conference on Bio-Sensing Technology

© CEA. All rights reserved

P.R. Marcoux | Forward-scattering for bacterial identification | 13 May 2013 | 2

Introduction: How a bacterium species can be identified ?

Genomic analysis

Antigeniccharacteristics

ID based upon the composition of cellular

membrane

Molecular methods

Biochemical tests

Enzymatic activities

Morphological characteristics

Microscopy / staining

ID based upon theglobal cellular composition

Spectralfingerprint

API test

[P69] Non-invasive detection of [P69] Non-invasive detection of bacteria via the sensing of volatile bacteria via the sensing of volatile metabolites released by enzymatic metabolites released by enzymatic

activityactivity L.H. Guillemot, M. Vrignaud, P.R. Marcoux, T.-H. Tran-Thi.



Introduction: Rapid methods in diagnostic

• identification tests must be performed on a much smaller amount of cells (1-103 cells), so as to reduce the time dedicated to growth• identification tests must be faster

Reference method Rapid method under investigation

24h 6h

24h a few seconds

API tests (bioMérieux):Pathogen is identified

according to the results (+ or -) of a series a biochemical tests

Raman spectroscopy:Pathogen is identified according to its Raman fingerprint



1. Raman: inelasticinelastic scattering / Forward-scattering: elasticelastic scattering Elastic scattering yields much more photons:exposure time in Raman: 30 s (spectrum on a single cell) exposure time in forward-scattering: 1 ms (single-cell; microcolony)2. Raman: vibrational spectroscopyvibrational spectroscopy, a peak is linked with a particular vibration mode of a type of covalent bond: e.g. υ(C=O); υ(C−Η)…

Introduction: optical methods for identification

Diffraction (elastic scattering)

Raman (inelastic scattering)

Forward-scatteringForward-scattering is not a spectroscopy: it does not yield information about cell composition, but rather about morphological characteristics.



3. As for Raman spectroscopy and intrinsic fluorescence, forward-scattering is a label-free methodlabel-free method. Non invasiveNon invasive technique, requires little or no consumable, can be automatedcan be automated.

4. In direct space: the packing of bacteria cells within microcolony induces a

periodicperiodic modulation modulation of phasephase (refraction index) and absorbanceabsorbance

in reciprocal space: it yields diffraction

fringes.

Introduction: optical methods for identification



100806040200

15

10

5

0

-5

PC#1

PC#2

EC21HA4

strain

PCA plot for EC21 and HA4 scattering patterns

Elastic diffusion

(laser)

Projectiononto a basis

functions descriptor(n-component vector)

Multivariate statistics

classification(for ex. Principal

Component Analysis)

n ...1

Forward-scattering(543 nm)

H. alveiH. alvei

E. coliE. coli

Introduction: experimental process and setup

camera 2 camera 2 (acquisition of (acquisition of scattering scattering pattern)pattern)

camera 1 camera 1 (images in (images in direct space)direct space)

Microcolony on Microcolony on agaragar

Laser (534nm)Laser (534nm)

microcolony(6h of incubation)



1. Scattering patternspatterns: How are they formed ?What kind of information do they give ?

2. Image analysisImage analysis: How can we compare scattering patterns quantitatively ?

3. Results: A first database of Gram- species Gram- species at 6hon TSA (Tryptic Soy Agar)

4. First results on two CNS (Coagulase-Negative StaphylococciStaphylococci) at 6h on ChromID MRSA.

Introduction

Let’s investigate the possibility of using forward-scatteringforward-scattering as an identification method on microcolonies after 6 hours of incubationafter 6 hours of incubation

(37°C), directly on agardirectly on agar medium:



A scattering pattern contains complexcomplex phenotypicphenotypic information, information, which is a sum of various parameters, such as:

1. Refraction index:Refraction index: of nutrient agar medium, of bacteria, of extracellular matrix.

2. Cellular shapeCellular shape.

3. Geometry of bacteria stacking within microcolonystacking within microcolony (the scattering of planktonic cells, i.e. growing in liquid medium, yields a much less complex pattern).

4. Shape of the whole microcolonywhole microcolony: it acts as a micro-lens.

1. Scattering patterns : phenotypic information



4h50

Fringes at low angles low angles: corresponds to low spatial frequencies, the whole bacterial colony scatters. More light, but less complex shape. Available from the start.

HALA

Fringes at high angles high angles: corresponds to high spatial frequencies, scattering due to the stacking of cells. Much less photons (appears after several hours of incubation), but more complex shape. Seems to yield more discrimination.

Spatial periods: a few tens of µm.

Spatial periods: 1 µm and less.

1. Scattering patterns : two kinds of fringes

LA

HA



4h50

E. coli ATCC8739

2. Image analysis: four strains of the same species at t=6h

How can we quantitatively compare all these scattering patterns ?

E. coliE. coli ATCC25922 ATCC25922(EC10)(EC10)






4h50

Projection Anm of scattering pattern f(r,) onto Zernike polynomial Znm

= similarity coefficient between the image f and the basis function Znm

2. Image analysis: calculation of descriptorvector (descriptor)vector (descriptor)

V=(V1,V2,…,Vn) made of Anm projections:

imageimage(scatterogram)(scatterogram)

1

0

2

0

),(),(1

ddrrrZrfn

A nmnm

The more similar f and Znm look, the higher is Anm (Zernike moment).

ff((rr,,))

ZZnmnm((rr,,))

projection

projection

projection

projection Projections Anm (Zernike moments) are calculated for the first 120 Zernike

polynomials Znm



3. Results: a first database on Gram- species (t = 6h) Classified as

Imaged scatterogram

EC8 EC10 EC11 EC21 EC28 HA4 CF7 sum

EC8 48,8 7,9 0,0 11,0 17,3 0,0 15,0 100% (127)

EC10 6,4 49,1 14,5 0,0 25,5 2,7 1,8 100% (110)

EC11 0,0 7,3 89,1 0,0 2,7 0,9 0,0 100% (110)

EC21 12,9 0,0 0,0 81,9 0,0 0,0 5,2 100% (116)

EC28 20,5 24,2 1,5 0,8 42,4 0,0 10,6 100% (132)

HA4 0,8 0,0 0,0 0,8 0,0 86,0 12,4 100% (121)

CF7 7,8 0,0 0,0 1,7 0,0 3,5 87,0 100% (115)

Forward scattering on microcolonies (6h of

incubation, 37°C) growing on a thin layer (1mm) of TSA (Trypcase Soy Agar). Laser Laser

beam: 100µmbeam: 100µm on bacteria on bacteria.

Average classification rate (Naive Bayes) over the whole database: 69%.



3. Results: a first database on Gram- species (t = 6h)

Principal Component Analysis

Can we discriminate the different strains of the E. coli species ?

Supervised learning(Naive Bayes Continuous)

Classification rate = 82% on average



Second step: with commercial Petri dishes (5mm thick), scattering patterns are acquired without without opening lidsopening lids no risk of cross-contamination between samples

4. Results: distinguishing two species of Staphylococci

We chose ChromID MRSA (bioMérieux) as a nutrient medium: screening of Gram+ strains resistant to methicillinstrains resistant to methicillin.

Can we discriminate, after 6h of incubation, two species of discriminate, after 6h of incubation, two species of StaphylococciStaphylococci that grow on ChromID MRSA ?Study on Staphylococcus haemolyticus and Staphylococcus cohnii, two methicillin-resistant species (Coagulase-Negative Staphylococci).

As we obtain a significantly slower growth, we reduce the laser beam laser beam on bacteria down to 25µm on bacteria down to 25µm .



4. Results: distinguishing two species of Staphylococci

Principal Component AnalysisSupervised learning (Naive Bayes Continuous)

Classification rates: 92% on average92% on average

Forward scattering through the whole Petri dish (including lid). 6h of incubation (37°C). 2 species of methicillin-resistant Staphylococci

growing on ChromID MRSA (bioMérieux). Laser beam: 25µm on bacteria.

S. haemolyticusS. haemolyticus

S. cohniiS. cohnii

S. haemolyticusS. haemolyticus

S. cohniiS. cohnii



Identifying pathogenic species is not enough: a complete diagnosis must include AAntibiotic SSusceptibility TTesting (AST).

To guide the selection and modification of antimicrobial therapy

Conclusion

Currently under investigation…

Towards label-free, non invasive (without opening lids), non destructive, automated

methods.Less than 6h for identificationidentification + ASTAST



AcknowledgmentsMathieu DUPOY

Antoine CUER, Joe-Loïc KODJAArthur LEFEVBRE, Florian LICARI

Robin LOUVET, Anil NARASSIGUINCharles-Edmond BICHOT

Frédéric MALLARDFrédéric PINSTON

http://eric.univ-lyon2.fr/~ricco/tanagra/fr/tanagra.html

http://www.cs.waikato.ac.nz/ml/weka/

optical instrumentation

optical instrumentation

microbiology

microbiology

image analysis and data mining

image analysis and data mining

Optical forward-scattering for identification of bacteria within microcolonies

Health & Medicine

bacterial identification

scattering patterns

o c forwardscattering

scattering of planktonic

identification of bacteria

identification method

tests marcoux

growth identification