Special sample types: CSF

Many infectious agents of clinical significance infect the central nervous system (CNS) and are thus of interest to the diagnostic laboratory. Examples can be found from every imaginable pathogen type; some of the better known suspects include viruses such enteroviruses, various herpesviruses, and JC virus; bacteria such as C. pneumonia, N. meningitidis, T. pallidum, and H. influenzae; yeast and fungi including C. neoformans and Candida spp.; and parasites like Acanthamoeba spp., Balamuthia spp., and T. gondii.

Many tests for infectious agents can make use of relatively easily obtained sample types, from direct swabs of accessible surfaces to expelled samples (urine, stool, sputum) to peripheral blood samples. Access to the CNS compartment is significantly more challenging, as direct tissue biopsy isn’t a simple option, and the presence of the blood-brain barrier negates any possible use of peripheral blood as a reliable sample type.

The best option left is cerebrospinal fluid (CSF), a protein-rich liquid which serves functions including cushioning of the brain, nutrient diffusion, and maintenance of intracranial pressure. An average adult CNS total volume is estimated at approximately 150 ml; however, only relatively small volumes (1-2 ml) are generally available for safe collection via lumbar puncture. CSF is an intrinsically sterile sample type, meaning that almost any pathogen detection is both pathogenic and significant, although exceptions can occur due to contamination during sample collection or may result in context of a systemic, non-CNS infection if there are transient breakdowns in the blood-brain barrier.

Challenges of CSF

Two challenges are common in the use of CSF as a diagnostic sample. The first is that titers of organisms may be quite low, which is one of the reasons why highly sensitive molecular methods may be the test modality of choice in this sample type. (By comparison, one study of CSF viral culture for HSV showed only a four percent positive detection rate in biopsy proven cases of HSV encephalitis).1 Unfortunately, the second problem is that CSF may often contain numerous substances inhibitory to nucleic acid amplification tests (NAATs) such as polymerase chain reaction (PCR) and in particular is rich in RNases, which can be problematic for efficient recovery of RNA targets such as enteroviruses. Optimal use of CSF as an MDx diagnostic specimen type thus requires both efficient recovery of low titer nucleic acids present and effective removal of a range of possible inhibitors, possibly along with RNase inactivation.

While rapid, crude sample preparation methods can work well for providing PCR template in some settings, this does not appear to hold true when the specimen is CSF. One representative study of this by Alfonso and coworkers2 reported on the relative efficacy of four CSF sample preparation methods in context of recovery of T. gondii DNA. In this example, all four approaches began by obtaining a cell pellet from the CSF by a brief low speed centrifugation. The least effective method was to digest this pellet with proteinase K followed by phenol-chloroform extraction, alcohol precipitation, and resuspension; the authors report a limit of detection (LOD) of 117 tachyzoites. Direct boiling of the cell pellet in sterile water was a better approach, with an estimated LOD of 16 tachyzoites. This was still significantly improved upon by an approach of proteinase K digestion in a cell lysis buffer followed by centrifugation to remove debris and testing of supernatant, with a reported LOD of two tachyzoites. Finally (and reassuringly, if your lab has invested in commercial nucleic acid extraction systems), a commercial extraction method based on the now common chaotropic lysis/silica adsorption/wash/elute approach was the clear winner, with a reported LOD of a single tachyzoite.

Two further observations might be made with regard to these results. First, it’s sometimes suggested that PCR inhibition observed with CSF samples is likely due to high protein content. That the poorest results of this study were obtained with phenol–chloroform extraction (a potent protein removal approach) casts some doubt onto this claim and suggests that CSF inhibition likely arises from non-protein constituents, at least in the context studied here. A second indirect observation is that three of the sample preparation approaches described here would likely have little effect on inactivating endogenous RNases; they are notoriously robust against proteinase digestion and extended boiling. Only the fourth (commercial) method, with use of a chaotropic lysis agent, would have much effectiveness at inactivating RNases. Thus, if the assay target were an RNA species such as an enterovirus, an even bigger bias in favor of the fourth method would likely have been observed. In fact, it’s notable that one of the early references for the use of guanidine thiocyanate (chaotrope)-based extraction methods3 was expressly in the context of CSF samples.

Two takeaways

A takeaway from this, then, is that CSF extraction is probably best done by commercially available, chaotropic lysis-based methods (which include most of the common paramagnetic particle-based methods and common spin-column methods). As another takeaway, recall that we observed above that target organism titers can be low in CSF. This suggests that inclusion of a carrier or coprecipitant molecule such as tRNA during the extraction process is also likely a wise choice. (Briefly, these act as a sacrificial material to intrinsic losses in the process; that is, if X ng of nucleic acid is normally lost in the protocol through adherence to plasticware, escape in aqueous waste phase, and other reasons, and all your input is meaningful target nucleic acid but on the order of X ng, you may lose all of it. If, however, you add in 9X ng of a carrier at the start for a total of 10X ng input, now the X ng lost consists of 0.9X ng carrier and 0.1X ng target material, meaning you still have 0.9X ng target material coming out, or 90 percent of your input.) This second point—frequently low target content—is also why functional removal of NAAT inhibitors by simple dilution of extracts may not be a viable option for CSF-derived samples, even though it often works with other, more target-rich specimen types.

As with most other MDx specimen types, fresh samples are ideal; however, freshly frozen CSF samples have been successfully extracted with recovery of at least some intact nucleic acids following prolonged storage. DNA targets are generally more stable than RNA ones for recovery after frozen storage, and in either case use of a sample which has been through multiple freeze-thaw cycles is highly undesirable.4 Post extraction, the eluted nucleic acids can be stored equivalently to any other MDx extracts (i.e., -20 °C for DNA, or RNA short term; -80 °C recommended for RNA longer term).

While the invasive collection nature of CSF can be outweighed by its ability to sample the otherwise hidden CNS compartment, this value proposition in its collection only holds true when it can be extracted and analyzed in a manner that maximizes its diagnostic utility. Less than ideal laboratory practices that may allow acceptable diagnostic accuracy on less demanding sample types cannot always be expected to perform well when CSF is being examined. Starting with the largest possible input volume of fresh sample and using best practices in extraction methods are crucial in making the most of this sample type.


  1. Nahmias AJ, Whitley RJ, Visintine AN, Takei Y, Alford CA Jr. Herpes simplex virus encephalitis: laboratory evaluations and their diagnostic significance. Collaborative Antiviral Study Group. J Infect Dis. 1982;145(6):829-836.
  2. Alfonso Y, Fraga J, Cox R, et.al. Comparison of four DNA extraction methods from cerebrospinal fluid for the detection of Toxoplasma gondii by polymerase chain reaction in AIDS patients. Med Sci Monit. 2008;14(3):MT1-6.
  3. Casas I, Powell L, Klapper PE, Cleator GM. New method for the extraction of viral RNA and DNA from cerebrospinal fluid for use in the polymerase chain reaction assay. J Virol Meth. 1995;53(1):25-36.
  4. DeBiasi RL, Tyler KL. Molecular methods for diagnosis of viral encephalitis. Clin Microbiol Rev. 2004;17(4):903–925.

John Brunstein, PhD, is a member of the MLO Editorial Advisory Board. He serves as President and Chief Science Officer for British Columbia-based PathoID, Inc., which provides consulting for development and validation of molecular assays.