All negative samples were correctly identified as negative, meaning that no contamination was introduced during purification. Four methods (phenol/chloroform, The DNeasy Tissue kit, Easy-DNA kit, MagNA Pure) were able to recover C. pneumoniae DNA from the lowest concentration sample (input concentration 10 copies/μl) with output concentrations ranging from 0.6–8 copies/μl (Table 1). For an input of 10³ copies/μl output concentrations ranged from 12–658 copies/μl and for an input of 10⁵ copies/μl output concentrations ranged from 728–7.5 × 10⁴ copies/μl (Table 1). For all input DNA concentrations the DNeasy Tissue kit showed the highest output concentration. Consequently, for the pure C. pneumoniae DNA samples the DNeasy Tissue kit had the best average recovery percentage (74.0%) followed by the MagNA pure with an average recovery percentage of 50.9%. All five methods showed approximately 10% intra-method difference between the highest and lowest recovery percentages.

All negative samples were correctly identified. Four methods (phenol/ chloroform + CTAB, The DNeasy Tissue kit, Easy-DNA kit, MagNA Pure) recovered C. pneumoniae DNA from the lowest input concentration sample (10 copies/μl) with output concentrations ranging from 1.3–6.3 copies/μl (Table 1) and the Easy-DNA kit had the highest output concentration. For an input of 10³ copies/μl output concentrations ranged from 13–285 copies/μl and the DNeasy Tissue kit had the highest output concentration. For an input of 10⁵ copies/μl output concentrations ranged from 1.4 × 10³-3.3 × 10⁴ copies/μl and the Easy-DNA kit had the highest output concentration. The range of recovery percentages obtained with one method was larger than for the pure DNA samples. For the spiked aorta samples the Easy-DNA kit had the best recovery percentage (40.6%) and the DNeasy Tissue kit the second best (32.6%).

All negative samples were correctly identified. No methods recovered C. pneumoniae DNA from the lowest EB dilution 1.2 × 10^-8 (~0.2 EB/μl), but all methods recovered C. pneumoniae DNA from 1.2 × 10^-6, 1.2 × 10^-5 and 1.2 × 10^-4 dilutions (~2000, 200, 20 EB/μl) (Table 2). Recovery percentages were calculated relatively to the method performing best. The MagNA Pure method performed best and showed the highest output concentrations for all dilutions. The DNeasy Tissue kit showed relative recovery percentages from 79.8–95.3%. The Easy-DNA kit showed the lowest relative recovery percentages from 1.5–5.3%. The phenol/chloroform+CTAB method showed the second lowest recovery percentages (11.4–35.8%) and the phenol/chloroform the third lowest recovery percentages (20.2–42.8%) (Table 2).

All negative samples were correctly identified. Two methods recovered C. pneumoniae DNA from the lowest dilution 1.2 × 10^-8 (~0.2 EB/μl); phenol/chloroform + CTAB and the MagNA Pure method (Table 2). From all other dilutions (~2000, 200, 20 EB/μl) (except one; 1.2 × 10–^-5, where the phenol/chloro-form+CTAB method showed higher recovery) the MagNA Pure method also showed highest output DNA concentrations and thereby the best performance. The phenol/chloroform+CTAB method showed from 18.0–110.5% relative recovery and the DNeasy Tissue kit showed between 58.7–73.7% relative recovery. The phenol/chloroform (44.5–53.2%) and Easy-DNA kit methods (12.3–26.1%) showed the lowest relative recovery percentages (Table 2). For the two best methods (MagNA Pure and DNeasy) the whole process from DNA extraction to PCR detection had a sensitivity of less than approximately 20 EB/μl (concentration in samples before DNA extraction) (Table 2).

When comparing recovery percentages for the pure DNA samples and the aorta samples spiked with DNA, it is seen that there was greater intra-method variation for the spiked aorta samples. For two methods (phenol/chloroform+CTAB and Easy-DNA kit) the recovery percentage was generally lower for the pure DNA samples (Table 1). When the overall recovery percentage was calculated for all samples for each method, the DNeasy Tissue kit had the best recovery percentage (53.3%) and MagNA Pure the second best (32.5%) (Table 1). On the pure EB samples and aorta samples spiked with EB, the MagNA Pure method performed best and the DNeasy Tissue kit second best (Table 2). For two methods the phenol/ chloroform+CTAB and Easy-DNA kit performance was best on aorta samples spiked with EB analogous to the situation on aorta samples spiked with pure DNA. On the basis of these combined results the DNeasy Tissue kit and the MagNA Pure methods were chosen for tests on patient samples.

The DNeasy Tissue kit and the MagNA Pure method were tested for purification of DNA from 78 atherosclerotic samples. None of the 2 × 78 DNA preparations were positive for C. pneumoniae and all samples were positive for human DNA. The average human DNA concentration in the samples purified with the DNeasy Tissue kit was 1.1 × 10⁴ copies/μl (SD = 1.6 × 10⁴, range 6.2 × 10² to 1.1 × 10⁵ copies/μl). The average human DNA concentration in samples purified with the MagNA Pure was 1.0 × 10⁴ copies/μl (SD = 1.1 × 10⁴ copies/μl, range 18 to 5.4 × 10⁴ copies/μl). As an inhibition check all samples purified with both methods were spiked with two different, known concentrations of C. pneumoniae genomic DNA (10³ copies/μl and 10 copies/μl). Subsequently, the spiked samples were subjected to real-time quantitative PCR. Here we found that both concentrations could be detected in all 78 samples using both extraction methods. The average C. pneumoniae concentration in the 78 samples purified with the DNeasy tissue kit and spiked with 10 copies/μl were 14.2 copies/μl (SD = 5.4) and in samples spiked with 10³ copies /μl it was 1.1 × 10³ copies/μl (SD = 185). The average C. pneumoniae concentration in the 78 samples purified with the MagNA Pure method and spiked with 10 copies/μl were 17.9 copies/μl (SD = 9.5) and in samples spiked with 10³ copies /μl it was 1.4 × 10³ copies/μl (SD = 287).

Two methods (phenol/chloroform+CTAB and Easy-DNA kit) performed better on the spiked homogenates than on pure C. pneumoniae DNA or EB (Table 1 and 2). It could reflect that in general these two purification methods perform better on the tissue samples as the purification kinetics is probably different in the two sample types. The Easy-DNA kit was used strictly as recommended by the manufacturers and not with the improvements done by Meijer et al. 12. They replaced the ethanol precipitation by further purification with silica particles and it was scaled up to enable DNA purification from 300 mg of tissue. This gave better sensitivity 12, so if we used their modification, the Easy-DNA kit might have performed better. Loss of DNA could be caused by precipitation, phenol/chloroform or chloroform extraction steps, as these steps are prone to DNA loss. This is also indicated by the fact that Meijer et al. 12 could improve the Easy-DNA method by replacing the precipitation step with silica-particles. The phenol/chloroform+CTAB method contains an additional precipitation step (CTAB) compared to conventional phenol/chloroform extraction, which may account for its relatively poor performance with pure DNA. The three remaining methods (phenol/chloroform, DNeasy Tissue kit and MagNA Pure) all showed reduced recovery for the homogenates. The DNeasy Tissue kit showed the best recovery on both pure DNA samples and aorta homogenates spiked with DNA. On purified EB and aorta homogenate spiked with EB the MagNA Pure method performed best. That two different methods were found to be best, probably reflects the different nature of the samples and underlines the importance of using whole bacteria and not just pure DNA for testing of extraction methods. The two best methods (The DNeasy Tissue kit and the MagNA Pure method) were chosen for use on patient samples. These methods possibly perform better because they contain few steps and because they do not contain any precipitation steps. A recent study comparing DNA extraction methods for extraction of C. pneumoniae DNA from vascular tissue also found a spin-column method (QIAamp DNA MiniKit) to give the best recovery 13. Another study tested recovery percentages of five different spin-column DNA extraction methods for purification of viral DNA from serum samples. Here recovery percentages were also determined with quantitative PCR in the LightCycler and the QIAamp Blood kit (uses same principle as DNeasy and is manufactured by Qiagen) showed a relative recovery percentage of 76% 14 comparable to our absolute overall recovery percentage of 53.3%. and overall average relative recovery of 66.5%.

From Tables 1 and 2 it is apparent that there is considerable variation in how much DNA could be recovered even within methods (3.0–53.3% overall average recovery and 8.8–100% overall average relative recovery). Real-time quantitative PCR uses a point (the threshold cycle or crossing point) at the beginning of the log-phase of the PCR to determine the original target DNA concentration. This gives a more realistic, though not perfect picture, of the actual DNA concentration in the sample tested. The variation would not have been as apparent if PCR products had been analysed by gel electrophoresis 1215 or time-resolved fluorometry (TRF) 6. When comparing PCR bands in a gel or PCR product amounts by TRF they represent the end points of the PCRs (i.e. at the plateau phase of the PCR) and they are often very similar even for very different initial target DNA concentrations, especially if many PCR cycles are performed. Therefore, previous studies have probably not been fully aware of the actual variation in output of DNA from their extraction methods and therefore our results may seem highly variable in comparison. In the study comparing different spin-column methods for DNA purification, they found relative recovery percentages between 2.2–76% 14. Therefore, it is apparent that even among methods based on similar principles there can be large differences in recovery of DNA 14.

Another reason for the apparent variability could be that the number determined by the real-time PCR depends not only on how much DNA is actually present, but also on the amount of inhibitors left in the sample, which will add further variation. It could be speculated that methods with low recovery percentages do not remove inhibitors sufficiently and are prone to DNA loss during purification, which in combination may give rise to the low recovery as measured by real-time PCR. In the early days of real-time PCR an estimation of the inhibition in the sample was obtained by calculating PCR efficiency for each sample using a dilution curve 16. This was based on the assumption (and also other assumptions not mentioned here) that PCR efficiency is constant during the entire run, which stems from the PCR kinetics model currently used in the LightCycler software R_n = R₀ *(1+E)ⁿ (R_n: Fluorescence at cycle number n, R₀: initial amount of target, E: PCR efficiency). However, recent publications have indicated that this may not reflect the real PCR kinetics in a given sample as the efficiency varies form cycle to cycle and because other parameters in a sigmoidal model are probably better suited to describe the kinetics 1718. As it is still unclear how PCR kinetics is best described and how much trivial factors like variation in reaction components, thermal cycling conditions and mispriming events contribute to deviations in PCR kinetics we have not incorporated PCR kinetics in the interpretation of our data. However, investigation of PCR kinetics under different conditions would clearly be helpful in selecting DNA extraction methods suitable for samples to be used in real-time PCR, as it possibly could be used to estimate the level of inhibition in the samples. This should obviously be a subject for future studies.

The rationale for using purified C. pneumoniae DNA in some of the samples was that input and output DNA concentrations were known or could be determined. However, this does not test whether the DNA extraction methods are capable of recovering DNA from whole EB. Therefore a test using tissue spiked with purified EB was also performed. It should be noted that this will not entirely reflect reality either as normally the EB are present inside human cells. Interestingly, the DNeasy Tissue kit has been used on lungs from mice experimentally infected with C. pneumoniae and here the kit was able to recover C. pneumoniae DNA amplifiable by real-time PCR 10. Therefore, this kit is actually able to recover C. pneumoniae DNA from a real infection where the bacteria are present inside cells. Our study in combination with the study by Apfalter et al. 8 indicate that more work is needed to optimise DNA extraction methods. Automated DNA extraction methods like the MagNA Pure method used in this study may be the future choice, as these methods may be able to produce more reproducible results, but this remains to be examined more carefully.

The DNeasy Tissue kit and the MagNA Pure method were used on 78 atherosclerotic tissue samples and subsequently amounts of C. pneumoniae DNA and human DNA were determined with quantitative real-time PCR. We found no samples positive for C. pneumoniae DNA. All samples were positive for human DNA. As there was great variation in the amount of human DNA among samples, it could be suspected that low human DNA numbers are caused by inhibitors, as the tissue input in mg should be approximately equal for the samples. However, atherosclerotic tissue shows a very complex and variable composition which may contribute to the variability observed. Some tissue samples may contain high levels of calcium, which will decrease the amount of human DNA, or a high number of inflammatory cells, which will increase the amount of human DNA. The high variability in human DNA recovery from the patient samples can lead to the speculation that there may also be a high variability in recovery of C. pneumoniae DNA, especially as the bacterium is obligate intracellular. Therefore, as it has been proposed by a recent Australian study 19, it may be necessary to analyse more samples (up to 15 samples according to 19) from different artery locations in each patient in order to obtain the true number of C. pneumoniae positive patients. This is according to this study 19 advantagous because the distribution of C. pneumoniae over the tissue is patchy and bacteria are present in low numbers. It should be noted though that using single random sections from 10 patients the Australian study found from 35.6–41.6% percent positives 19, whereas we find no positives in 78 patients using two samples purified with two different DNA extraction methods from the same site. Therefore, we will probably not observe as high numbers of positive patients even though more samples per patient were analysed. Especially in this experimental context where the PCR sensitivity is 2 genomes in the purified sample amount loaded directly into the PCR tube 9 and the sensitivity of the whole process of DNA extraction followed by PCR is less than 20 EB/μl (concentration in samples before DNA extraction).

It is highly probable that there are larger amounts of human DNA than of C. pneumoniae DNA present in the samples. Accordingly, inhibition will have a greater effect on the C. pneumoniae PCR results and consequently the human DNA PCR cannot be used as an inhibition control. Therefore, inhibition was tested in the samples by spiking them with two concentrations of C. pneumoniae DNA before PCR. Hereby, we found that the C. pneumoniae DNA could be detected in all samples. Therefore, it seems that the DNeasy Tissue kit and the MagNA Pure method did remove inhibitors sufficiently.

To study whether there were methodological differences between studies with positive and negative results we compared all the negative studies with a selection of representative positive studies (Table 3). The positive studies were selected so that they represented studies both with high and low prevalence of C. pneumoniae and the entire span of years of C. pneumoniae direct detection in tissue by PCR. Furthermore, different countries were represented. Among the positive studies The QIAamp Tissue Kit and variations of phenol/chloroform extractions were mostly used. Most of the positive studies declare that they attempt to control contamination at some level, at least with reagent negative controls in the PCR runs. The early negative studies 252627 only used Proteinase K as DNA extraction method, which may not be sufficient as, in most cases, it will leave behind many potential inhibitors. However, one of the positive studies did use Proteinase K only for extraction 32. The Chelex [100 method used in one of the negative studies 7 may not be sufficient for purification either as it only removes multivalent cat ions and lyse cells 41. The rest of the negative studies used DNA extraction methods optimised by their own laboratory 15 or methods also used by positive studies 6202122. Two negative studies used the modified version of the Easy-DNA kit 2324. This means that at least 7/11 negative studies used a DNA extraction method that should be suitable for purification of DNA from tissue. Four of these studies did also check sufficiently for PCR inhibition 6212224, but it is unclear whether the three other studies did 231520. Not surprisingly, some of the negative studies did use a DNA extraction method which might have been insufficient (4/11 negative studies). Three of these studies did check for inhibitors, by either addition of an internal processing control to the samples 7, spiking samples with C. pneumoniae DNA before PCR 26 or with a β-actin-specific PCR on the samples 25, however, as discussed above, testing for presence of human DNA may not be a sufficient inhibition control. This means that with regard to methodology, 9/11 negative studies did use sufficient DNA extraction methods or did check for inhibitors.

When comparing PCR methods used it is seen that almost the same number of negative and positive studies used a nested PCR assay (5/11 negative vs. 6/13 positive studies used nested PCR). The reason for higher positivity rates in studies using nested PCR could be either because nested PCRs are more sensitive or because they are more susceptible to contamination, but in the selected studies (table 3) there is no clear indication of this. A large number of positive studies used a PCR based on the PstI fragment (1/10 negative vs. 6/13 positive studies used a PstI-fragment based PCR). It could be speculated that either the PstI fragment-based PCRs are either more prone to unspecific amplifications or they are more sensitive. This needs to be investigated in more detail.

To date two Danish studies 67 and the present study have been conducted by three different Danish research units, and in no study has C. pneumoniae DNA been detected in atherosclerotic tissue. Negative results were obtained despite the use of optimised diagnostic tests and relatively large numbers of patient samples. The frequency of atherosclerosis in Denmark is not lower than in e.g. USA, England and Italy, where C. pneumoniae has been found in atherosclerotic tissue (Table 2) with high prevalence. These facts argue against a correlation of C. pneumoniae and cardiovascular disease.

A number of circumstances could explain why C. pneumoniae has not been demonstrated in atherosclerotic samples from Danish patients. From other studies it seems probable that there are great inter-laboratory differences in the handling of both the PCR and DNA extraction methods 8. Several multicenter comparison trials have also shown that some laboratories may have contamination problems 842. Consequently, it could be argued that high detection rates of C. pneumoniae DNA in atherosclerotic tissue, at least in part, may be caused by contamination. A recent study has for example shown that the nested PCR is highly unreliable for C. pneumoniae detection because it is very prone to contamination 43. On the other hand it could be argued that the negative studies used insensitive or inappropriate techniques. However, the present study and most other studies with negative results have attempted to optimise methods to ensure optimal sample preparation or inhibition control and they used a sensitive PCR technique. However, as discussed above it may be necessary to optimise methods further and to include more samples from each patient 19.

It is possible that divergence between studies is caused by periodical presence of C. pneumoniae in blood vessels dependent on the local epidemic conditions and C. pneumoniae is just transiently present during epidemics. The C. pneumoniae seroprevalence is not as high in Denmark as in many other countries where seroprevalence is often over 50%. Among Danish blood donors without respiratory tract symptoms it was approximately 40% 44. In Denmark there has apparently not been a C. pneumoniae epidemic since the early eighties 45. One study found 1.3% (28/ 2219 samples) prevalence of C. pneumoniae by culture and PCR in Danish respiratory tract samples sent to the Chlamydia laboratory at Statens Serum Institut in Denmark from 1993 to 1995. (Most routine detection of C. pneumoniae in Denmark is performed at Statens Serum Institut) 46. This low prevalence of C. pneumoniae in respiratory tract samples may explain why C. pneumoniae DNA is not found in the Danish atherosclerotic tissue as the vascular C. pneumoniae infections are likely to originate from the respiratory tract infections. In table 3 it is seen that the prevalence of C. pneumoniae in atherosclerotic tissue from different countries varies between 0–58%, which indicates variations that may be caused by local epidemic conditions, even though not all positive studies published are represented in the table. However, no countries at present maintain sentinel surveillance of C. pneumoniae infections and results from local health authorities and research reports have not yet provided evidence that different epidemic conditions could cause the differences.

C. pneumoniae was cultivated and EB purified as described previously 47. C. pneumoniae DNA was purified from EB with the Qiagen Blood and Cell culture DNA Midi Kit as instructed by the manufacturer and concentration was measured with a spectrophotometer as previously described 910.