Analysis of clinical isolates of Propionibacterium acnes by optimised RAPD

Random amplification of polymorphic DNA (RAPD) was evaluated as a genotypic method for typing clinical strains of Propionibacterium acnes . RAPD can suffer from problems of reproducibility if parameters are not standardised. In this study the reaction conditions were optimised by adjusting template DNA concentration and buffer constituents. All isolates were typeable using the optimised RAPD protocol which was found to be highly discriminatory (Simpson’s diversity index, 0.98) and reproducible. Typing of P. acnes by optimised RAPD is an invaluable tool for the epidemiological investigation of P. acnes for which no other widely accepted method currently exists.


Introduction
Propionibacterium acnes is a Gram-positive, non-sporeforming, anaerobic bacillus, which is commonly present as part of the normal £ora of the skin, oral cavity, large intestine and the external ear [1]. Historically P. acnes is considered to be of low virulence; however, in recent years it has been found as the aetiological agent in various pathologies. It is most notably implicated in the condition acne vulgaris [2] but is also associated with endophthalmitis [3], endocarditis [4], osteomyelitis [5], sarcoidosis [6] and prosthetic hip infections [7]. Recently P. acnes has been isolated from intervertebral disc material from patients with severe sciatica [8]. It is proposed that the microorganism might be causing a chronic low-grade infection in the lower vertebral discs of patients with severe sciatica [8].
Molecular typing of P. acnes and other cutaneous pro-pionibacteria has not been widely applied in the clinical situation. Dairy propionibacteria have been analysed previously by random ampli¢cation of polymorphic DNA (RAPD) and restriction endonuclease analysis [9], and pulsed ¢eld gel electrophoresis has been used to compare strains of P. acnes isolated from cases of endophthalmitis [10]. However, in our experience utilisation of these methods has provided data which are less than satisfactory, time-consuming and expensive to implement. Progress in molecular typing of P. acnes has been hindered by the lack of an e¡ective cell lysis technique; however, this has now been overcome to some extent by the use of penicillin in the lysis method [11]. RAPD is a rapid method which uses a single short primer allowing for the detection of DNA sequence polymorphisms. We investigated the potential of RAPD to genotype strains of P. acnes isolated from clinical samples. Ampli¢cation-based DNA ¢ngerprinting methods can be sensitive to changes in reaction conditions resulting in changes in pro¢les and therefore careful optimisation of typing protocols is required [12]. In this study we describe an optimised RAPD protocol for the epidemiological typing of P. acnes from clinical sources. Such a method will provide invaluable epidemiological data for P. acnes which is currently lacking.

Bacterial strains
Clinical strains of P. acnes (n = 46) were obtained from the University Hospital Birmingham NHS Trust and iden-ti¢ed by analytical pro¢le index (bioMe ¤rieux). P. acnes NCTC 737 and 10390 were included as reference strains and laboratory strains of Propionibacterium granulosum (n = 1) and Propionibacterium avidum (n = 3) as related (outlying) species.

DNA extraction
Strains were grown in 20 ml brain heart infusion (Oxoid, UK) in universal bottles and incubated at 37 ‡C without shaking for 72 h. The cultures were then supplemented with 20 Wl penicillin G (20 mg ml 31 ) and incubated for a further 3 h. Genomic DNA was prepared by sodium dodecyl sulfate (SDS) lysis and ethanol precipitation as follows: cells were harvested from 6 ml of culture (13 000Ug, 5 min) and the pellet was resuspended in 270 Wl TrisÊ DTA (10 mM Tris, 1 mM EDTA, pH 8.0). The cells were heated at 75 ‡C for 10 min to inactivate DNases and lysed by addition of 30 Wl SDS (100 mg ml 31 ). The lysate was treated with 3 Wl proteinase K (10 mg ml 31 ) and the suspension incubated at 65 ‡C for 3 h. The suspension was diluted to 600 Wl with sterile distilled water and an equal volume of phenol:chloroform:isoamyl alcohol (25:24:1) was added. After brief vortexing and centrifugation (13 000Ug, 1 min) the upper aqueous phase containing the DNA was recovered and precipitated by addition of 0.1 vol 3 M sodium acetate pH 5.2 and 2 vol ice-cold ethanol. The suspension was incubated at 320 ‡C for 20 min and the DNA pelleted at 13 000UgU20 min. Following a 70% (v/v) ethanol wash the DNA pellet was dried at 40 ‡C for 10 min. The pellet was redissolved in 30 Wl polymerase chain reaction (PCR) grade water and allowed to rehydrate overnight at 4 ‡C. DNA samples were checked for purity and quanti¢ed by spectrophotometric measurement (optical density ratio 260/280 nm). All preparations were stored at 320 ‡C until required.

RAPD optimisation
The optimum bu¡er composition for PCR reactions was determined using the Opti-Prime1 bu¡er matrix (Stratagene, USA) consisting of 12 bu¡ers varying in MgCl 2 and KCl concentrations and pH ( Table 1). The bu¡er supporting ampli¢cation of pro¢les with suitable discrimination was selected for further testing. A range of DNA concentrations (3^40 ng Wl 31 ) were subjected to RAPD analysis to establish the optimum concentration of template.

Optimised RAPD reaction
PCR cycling conditions were performed as described by Hilton et al. [13] using the selected bu¡er and DNA concentration. PCR was carried out in a 25 Wl reaction containing 1UPCR bu¡er, 200 WM dNTPs (Promega, UK), 100 pmol primer 1254 (5P-CCGCAGCCAA-3P, MWG biotech, Ebersberg, Germany), 1.25 U Taq polymerase (Promega, UK) and 2 Wl of DNA template. The ampli¢cation procedure comprised of one cycle for 4.5 min at 94 ‡C followed by ¢ve cycles of 30 s at 94 ‡C, 2 min at 20 ‡C and 2 min at 72 ‡C, and 35 cycles of 30 s at 94 ‡C, 1 min at 32 ‡C and 2 min at 72 ‡C (PTC-100 Peltier Thermal Cycler, MJ Research, Inc., USA). The ampli¢cation was concluded with a ¢nal extension step of 5 min at 72 ‡C and the reactions stored at 320 ‡C. Ampli¢cation products were separated by electrophoresis in 2% agarose and visualised by ethidium bromide staining. RAPD ¢ngerprints were analysed using GelCompar (Applied Maths, Belgium) with the band matching coe⁄cient of Dice and a dendrogram generated using UPGMA clustering. To con-¢rm intra-and inter-reproducibility using the selected optimum quantity of DNA and PCR bu¡er, RAPD was performed on DNA samples in duplicate and on samples of DNA template prepared from separate cell cultures. The discriminatory power of the typing method was calculated using Simpson's diversity index (DI) [14].

RAPD optimisation^bu¡er composition
Pro¢les from PCR reactions using 12 bu¡ers varying in pH and MgCl 2 and KCl concentrations are shown in Fig. Table 1 Opti-Prime1 bu¡er matrix [19] Tris^HCl ( 1. At each pH, a high concentration of MgCl 2 (3.5 mM) with a low concentration of KCl (25 mM) did not support product ampli¢cation (lanes 3, 7 and 11). However, a decrease in MgCl 2 concentration to 1.5 mM with a low concentration of KCl improved ampli¢cation (lanes 1, 5 and 9). A high concentration of KCl (75 mM) with a low or high concentration of MgCl 2 supported product formation but gave smeared pro¢les with low discrimination (lanes 2, 4, 6, 8, 10 and 12). Increasing the pH appeared to have little e¡ect on the pro¢les obtained relative to the e¡ect of changing the concentrations of KCl and MgCl 2 . Similar e¡ects of bu¡er composition upon RAPD pro¢le were found for all strains examined, with bu¡er 9 giving the clearest and most discriminatory pro¢les. These results demonstrate the importance of the optimisation of PCR reactions. As demonstrated by other workers [15], variation in the magnesium concentration resulted in marked alterations of pro¢les. Magnesium promotes and stabilises primer^template interactions, has an e¡ect on denaturation of the template DNA and is required for enzyme activity and ¢delity. High concentrations of magnesium may inhibit ampli¢cation due to inadequate denaturation of the template DNA and can also lead to the accumulation of non-speci¢c ampli¢cation products. By contrast, insu⁄cient magnesium ions will reduce yield as primers are unable to anneal e⁄ciently to the template DNA [16]. Potassium chloride can also a¡ect PCR speci¢city as it facilitates primer annealing [16] and can directly a¡ect Taq polymerase [17]. A lower concentration of KCl in the PCR bu¡er was optimal for the primer^template combination used in this study. Pro¢les produced using a bu¡er with a high concentration of potassium chloride resulted in smeared pro¢les. This may have occurred due to extension of one primer without extension of a primer on the opposite strand. The pH of the PCR bu¡er had little e¡ect on the pro¢les obtained ; however, if the pH is too low, non-speci¢c reactions can occur and if too high, yield is reduced. Bu¡er 9 (1.5 mM MgCl 2 , 25 mM KCl, pH 9.2) was selected for use in further RAPD reactions.

RAPD optimisation^template DNA concentration
When optimising RAPD the template concentration can be a critical factor to consider. The concentration of DNA can in£uence the number of products resulting in di¡erent ¢ngerprints, therefore standardisation of the template concentration is important for reproducibility. Excess template can result in suppression of the ampli¢cation process due to competition between template DNA and ¢rstround amplicons and relative shortage of primers. Fig. 2 shows the RAPD pro¢les obtained using bu¡er 9 with template concentrations ranging from 3^40 ng Wl 31 . The pro¢les remained constant throughout this concentration range and a DNA concentration of 10 ng Wl 31 with bu¡er 9 was chosen as the optimised RAPD system for analysis of clinical isolates.

Optimised RAPD of P. acnes isolates
Strains of P. acnes isolated from various sources were examined using the optimised RAPD method. Fig. 3 shows the pro¢les from four strains chosen to illustrate the di¡erent pro¢les, containing six to eight bands in the 200^1500 bp range. All isolates were typeable by this method and, on repeated testing on two separate occasions three months apart, all pro¢les could be assigned  Table 1.  to the same pattern as that previously generated giving 100% intra-and inter-reproducibility (Fig. 3). The optimised RAPD protocol was highly discriminatory (DI, 0.98), in keeping with the requirement for an index greater than 0.9, which is desirable if the results of a typing scheme are to be interpreted with con¢dence [14].
The relationship between P. acnes pro¢les was determined by UPGMA cluster analysis (Fig. 4). Clinical isolates of P. acnes were represented in a major cluster at 65% similarity, which further subdivided into two distinctive pro¢le types. One of these distinctive pro¢le types contained NCTC 737, a P. acnes serotype I strain [18], and the other contained NCTC 10390, of serotype II (S. Patrick, personal communication). It will be interesting to determine whether the clinical strains in the two major RAPD groups belong to serotype I or II and hence if this typing system distinguishes between the two serotypes. In particular, the cluster containing the serotype I strain, NCTC 737, was characterised by the presence a band of 200 bp (e.g. strain a in Fig. 3). Studies are currently in progress to sequence this band as a guide to the genetic di¡erences between the two serotypes of P. acnes. A DNA probe based on this sequence might provide an alternative approach to the identi¢cation of type I and type II strains.
The RAPD system di¡erentiated between P. acnes and the related Propionibacterium species, P. granulosum and P. avidum, which gave markedly di¡erent pro¢les with less than 40% similarity to those of P. acnes. Strains isolated from di¡erent clinical sources were distributed amongst the RAPD pro¢le types within the P. acnes clusters. Fu- Fig. 4. UPGMA dendrogram analysis of RAPD pro¢les from culture collection strains of P. acnes (NCTC 737 and 10390), P. granulosum and P. avidum, and clinical strains of P. acnes labelled according to clinical source : blood culture (blood), skin strains (skin), prosthetic hip infections (hip) and microdissectomy tissue isolates (sciatica). ture work with more strains from a larger variety of clinical sources will enable us to determine whether certain genotypes are associated with speci¢c clinical conditions and phenotypic properties.