Identification of Respiratory Tract Bacteria Isolated from Sputum of Cystic Fibrosis Patients by FT-IR Spectroscopy

ALEJANDRO MIÑÁN; ALEJANDRA BOSCH; CECILIA VESCINA; JOSE DEGROSSI; JUERGEN SCHMITT; BLANCA GATTI; MIRTA FRANCO; DIETER NAUMANN; OSVALDO YANTORNO

Heidelberg Germany

Workshop; Spec 2006. Shedding Light on Disease:; 2006

Lung disease in cystic fibrosis (CF) is characterized by vicious cycle of inflammation and infection, and chronic bacterial colonization, which is the major cause of morbility and mortality in these patients. Although the most common organisms isolated from sputum samples are Pseudomonas aeruginosa, Staphylococcus aureus, and Haemophilus influenzae, during the last decades an increasing fraction of CF patients has been colonized by other non-fermenting gram-negative bacilli as Burkholderia spp, Stenotrophomonas maltophilia, Ralstonia spp, Acinetobacter spp. and Achromobacter spp. Identification of these bacteria at genus and species level in hospitals and CF centres are normally performed by time-consuming laborious and sometimes unreliable biocemical methods. Futhermore, the taxonomic complexity in some of These genera, as bacteria belonging to the B. cepacia complex, also contributes to misidentifications. At least 9 distinct species of closely related bacteria (genomovars), nearly indistinguishable by commercially avariable systems, comprise this complex. Accurate identification of multiply antibiotic-resistant bacilli isolated from CF patients is critical in epidemiological studies, intra-hospital outbreaks, patient treatment and therapeutic options. In this work we develop a rapid, sensitive and specific method based on FT-IR spectroscopy and artificial neural networks (ANNs) for the discrimination and identification of nonfermenting and multi-resistant gram-negative bacilli isolated from respiratory samples of CF patients at the genus, and in the case of B. cepacia complex down to the subtype level. A total of 12 reference strains and 100 clinical isolates from 90 patients at the La Plata Chilren ´s Hospital and Buenos Aires Clinical Hospital were used in this study. Bacteriaa were isolated by routine clinical procedures, and phenotypic identification of Pseudomonas, Stenotrophomonas, Ralstonia, Acinetobacter and Achromobacter was carried out by API20NE system. Bacteria of B. cepacia complex were also identified by PCR-RFLP technique. FT-IR spectroscopic measurements were performed with cells grown on ATS medium. As some of these genus produce different quantities of cell-bond-pili and reserved substances as poly-hydroxy-butyrate acid (PHB), culture conditions and sample preparation were previously optimised to eliminate the corresponding spectral signals that interfere in the FT-IR differention and identification. First derivativesof original spectra, vector normalized in the whole spectral range measured, were processed by statistical methods based on pattern recognition using cluster analysis (CA) and ANNs. A hierarchical database consisting of a top level and two subsequent sub-classification ANNs was created. The top-level ANN was optimized for the differentiation and identification of strains from the six above-mentioned genera. Then, a more dedicated network was developedto differntiate the 9 genomovars (I-IX) belonging to B. cepacia complex. Finally, for B. cenocepacia (genomovar III), which is recognised as the species with the highest virulence spreading capacity of the complex, a third level in the hierarchical classification scheme was trained to differentiate the two subtypes IIIa and IIIb. Our results demonstrate a high relianbility and a strong potential of ANN-based FT-IR spectrum analysis for accurate differentiation and identification of isolates from sputum of CF patients in a short time (less than 10 h after bacterial isolation), and without complicated sample handling.