www.elsevier.com/locate/ybbrc

Biochemical and Biophysical Research Communications 336 (2005) 1268–1277

BBRC

Hyperphosphorylated neurofilament NF-H is a serum biomarkerof axonal injury

Gerry Shaw a,b,g,*, Cui Yang a,b, Rebecca Ellis a,b,e, Kevin Anderson b,c, J. Parker Mickle a,b,d,Stephen Scheff f, Brian Pike a,b,1, Douglas K. Anderson a,b,d,e, Dena R. Howland a,b,e

a Department of Neuroscience, University of Florida College of Medicine, Gainesville, FL, USAb McKnight Brain Institute, University of Florida College of Medicine, Gainesville, FL, USA

c Department of Physiological Sciences, University of Florida College of Medicine, Gainesville, FL, USAd Department of Neurosurgery, University of Florida College of Medicine, Gainesville, FL, USA

e Malcom Randall VAMC, Gainesville, FL, USAf Department of Anatomy and Neurobiology, University of Kentucky, Lexington, KY, USA

g EnCor Biotechnology Inc., Alachua, FL 32615, USA

Received 6 August 2005Available online 12 September 2005

Abstract

Several lines of reasoning suggest that the phosphorylated axonal form of the neurofilament subunit NF-H is likely to be releasedfrom damaged and diseased neurons in significant amounts. Detection of this protein in serum or CSF might therefore provide infor-mation about the presence and degree of neuronal loss. We therefore developed a sensitive NF-H ELISA capable of detecting picogramquantities of phosphorylated NF-H (pNF-H). This assay showed that soluble pNF-H immunoreactivity is readily detectable in the seraof adult rats following various types of experimental spinal cord injury (SCI) and traumatic brain injury (TBI), but is undetectable in thesera of normal animals. Here we describe details of the time course and extent of serum pNF-H expression following experimental SCIand TBI. Following SCI, serum pNF-H showed an initial peak of expression at 16 h and a second, usually larger, peak at 3 days. Fol-lowing TBI, lower levels of serum pNF-H were detected with a peak at 2 days post-injury. We also show that the higher levels of pNF-Hreleased from injured spinal cord as compared to brain are in line with the �20-fold higher levels of pNF-H present in spinal cord. Thesefindings suggest that serum levels of pNF-H immunoreactivity may be used to conveniently monitor neuronal damage and degenerationin experimental and presumably clinical situations.� 2005 Elsevier Inc. All rights reserved.

Keywords: Neurofilaments; NF-H; pNF-H; Hyperphosphorylated NF-H; Biomarker; Traumatic brain injury; Spinal cord injury; ELISA; Axonal bio-marker; Asonal injury biomarker; Neuronal cytoskeleton; Neuronal injury biomarker

Much interest has been focused on the detection of spe-cific marker substances in serum, CSF or other body fluidsthat may indicate the presence of specific kinds of disease,pathology or trauma. The detection of so-called biomark-ers allows the monitoring of animal models of damageand disease states, and may provide quick and simple clin-

0006-291X/$ - see front matter � 2005 Elsevier Inc. All rights reserved.

doi:10.1016/j.bbrc.2005.08.252

* Corresponding author. Fax: +1 352 392 8347.E-mail address: shaw@mbi.ufl.edu (G. Shaw).

1 Present address: National Institutes of Health—NIGMS, 45 CenterDr., MSC 6200, Room 3AN18, Bethesda, MD 20892-6200, USA.

ical diagnosis and prognosis in humans. We have beeninterested in specific biomarkers of neuronal injury andhave attempted to predict which neuron specific proteinsmight be detectable in the serum and CSF of animals suf-fering neurodegeneration following trauma or pathology.Neurofilaments are the most abundant protein componentsof neurons and should therefore be released from damagedand dying neurons in large amounts. Due to their abun-dance it should be relatively easy to detect them and, be-cause they are only found in neurons, detection of theseproteins points unambiguously to neuronal damage.

G. Shaw et al. / Biochemical and Biophysical Research Communications 336 (2005) 1268–1277 1269

Neurofilaments consist predominantly of three subunits,namely NF-L, NF-M, and NF-H. NF-H contains unusualtandemly repeated peptides centered on the sequence ly-sine-serine-proline (KSP). These KSP repeats are verynumerous, up to 60 in some mammalian species, and inaxonal neurofilaments essentially all of the serine residuesare phosphorylated [1]. Several studies have shown thatpNF-H is more resistant to calpain and other proteasesthan either NF-L or NF-M [2–7]. These findings suggestthe hypothesis that pNF-H is released into serum andCSF in significant amounts following CNS injury, and,being resistant to degradation, may be readily detectedwith appropriate ELISA or other assays. We developed apNF-H ELISA, described here for the first time, whichwas used to test this hypothesis. Since blood is much moreeasily and conveniently obtained than CSF, we concentrat-ed on the detection of pNF-H in serum. Our ELISA dem-onstrates that pNF-H can in fact be detected in the sera ofexperimentally injured animals in surprisingly large quanti-ties, but is not detected in the sera of control animals with-out CNS injuries. We also provide a description of the timecourse and extent of the pNF-H serum signal following avariety of experimental CNS injuries. Finally, we usedthe ELISA to quantify the levels of pNF-H in different re-gions of the rat CNS.

Materials and methods

Antibody production, characterization, and immunocytochemistry. Bo-vine tissue was obtained from a local slaughterhouse, transported on ice,desheathed of meninges, and stored at �70 �C. Neural 10 nm filamentrich gels were prepared using a glycerol polymerization protocol basedon the method of Delacourte et al. [8] as outlined previously [7]. Thefilament preps were dissolved in 6 M urea in 10 mM phosphate buffer,1 mM EDTA, and 0.1% b-mercaptoethanol, pH 7.5, and applied to aDEAE cellulose column equilibrated in the same buffer. The column wasdeveloped with a gradient from 0 to 0.25 M NaCl, and a single cleanNF-H protein band, with an apparent SDS–PAGE molecular weight of220 kDa, eluted at about 0.05 M NaCl. This material was dialyzedagainst PBS and used for antibody production in rabbit and chicken.Polyclonal antibodies were raised by standard procedures and one rabbitwas exsanguinated and sera were collected for affinity purification. Forchicken antibodies, eggs were taken and IgY preparations were preparedby chloroform delipidation followed by polyethylene glycol precipitationessentially as described previously [9]. Both antibodies are availablecommercially from EnCor Biotechnology (Alachua, FL) and are referredto as CPCA-NF-H (chicken anti-pNF-H) and RPCA-NF-H (rabbit anti-pNF-H). Both the rabbit sera and IgY preps were then affinity purifiedusing purified bovine NF-H coupled to cyanogen bromide activatedSepharose 4B (Sigma–Aldrich, St. Louis, MO). Eluted antibodies weredialyzed overnight against PBS and quantified using the Pierce micro-BCA assay. The affinity purified preparations are hereafter referred to asAP-CPCA-NF-H and AP-RPCA-NF-H. Monoclonal antibodies SMI33and SMI35 were obtained from Sternberger Immunocytochemicals (Bal-timore, MD). Western blotting was performed using standard methods asdescribed previously [7] using PVDF membranes (Bio-Rad, Hercules,CA). Immunocytochemistry was performed using formalin/cold methanalas outlined previously [10]. Mixed neuron/glia cultures were derived fromp18 rat cortex and grown for 10 days in tissue culture. Secondary anti-bodies were commercial preparations of extensively cross-absorbed goatanti-mouse, anti-rabbit, and anti-chicken ALEXA 488 and 594 conju-gates, and were used at 0.5 lg/ml (Molecular Probes, Eugene, OR).

ELISA. One hundred microliter of AP-CPCA-NF-H at a final con-centration of 0.01 mg/ml in 50 mM sodium bicarbonate buffer at pH 9.5was applied to each well of 96-well Maxisorb ELISA plates (Fisher Sci-entific, Pittsburg, PA). Plates were incubated overnight and then non-specific binding was blocked with 200 ll ELISA buffer (5% Carnationinstant non-fat dried milk in Tris buffer saline/Tween (TBST, 150 mMNaCl, 10 mM Tris/HCl, pH 7.5, plus 0.1% Tween 20)) per well for 1 h.Blocked plates could be stored for several weeks in TBS plus 5 mM azideat 4 �C. Fifty microliter of ELISA buffer was applied to each well of theplate, and protein samples made up to 50 ll total volume with ELISAbuffer were applied to the first row of each plate. This material was typ-ically serially diluted down the dish, leaving a final volume of 50 ll in eachassay well. After a 1 h incubation with shaking at room temperature, theplate was extensively washed in TBST using a Bio-Rad microtiter platewasher. 100 ll of AP-RPCA-NF-H was dissolved in ELISA buffer andapplied to each well at a final concentration of 1 lg/ml, and incubated for1 h at room temperature with shaking. After again washing with TBST,each well was incubated with 100 ll ELISA buffer with 1:2000 goat anti-rabbit alkaline phosphatase (Sigma–Aldrich, St. Louis, MO). After 1 h atroom temperature with shaking, the plates were washed for a final timeand developed with 100 ll/well of 0.1 M glycine, 1 mM Mg2+, and 1 mMZn2+at pH 10.4 containing 1 mg/ml p-nitrophenyl phosphate (Sigma–Aldrich, St. Louis, MO). After 1 h, the reaction was stopped with 50 ll/well of 2 M NaOH, and results were quantified on a Tecan SpectrafluorPlus ELISA plate reader at 405 nm absorbance. Data were normalized bysubtraction of the background from each determination and division ofthe resulting signal by the OD difference between the background and thesignal from full saturation of the assay.

Experimental injuries. For experimental spinal cord injuries, femaleLong–Evans Hooded rats weighing 220–250 g were obtained from Har-lan Labs (Indianapolis, IN). Females were used because bladder carecomplications post-SCI occur less frequently than in males. Surgicalprocedures were performed under aseptic conditions on a heating pad.Rats were anesthetized with sodium pentobarbital (40–60 mg/kg, IP).Half of rats received a contusion injury which mimics a common form ofspinal cord injury in humans that result in bilateral spinal damage andmotor-sensory loss. Contusion injuries were performed using the NYUimpactor device [11,12]. For the contusion injury, a T12, partial T11laminectomy was made without disrupting the dura, and a 10-g rod wasdropped 25 mm onto the exposed spinal cord. Sham controls receivedlaminectomies and were placed into the injury device, but did not receivea spinal injury. The second group of rats received a spinal hemisectioninjury. This type of injury causes only unilateral damage and mimics apartial severing of the cord due to a bullet or knife wound in humans.For this procedure, a T11 laminectomy was made and the dura slitlongitudinally with a size 11 scalpel blade. A lateral spinal hemisectionwas then made using iridectomy scissors. Following all surgeries, thelesion site was covered with dura film and then gelfoam. Incisions wereclosed in layers, and animals recovered in heat-controlled incubatorswith food and water ad libitum. Bladders were manually emptied at least2 times/day until voluntary bladder function recovered. Fluids wereadministered immediately after injury and each time a blood sample wastaken (3–5 ml lactated ringers, SQ). Approximately 100 ll of blood wastaken from the tip of the tail several days prior to injury and then at thespecified times post-injury. Blood was typically collected at 5 min, 2 h,8 h 16, and 24 h post-injury, and then every 24 h up to 10 days. Inseveral animals blood samples were collected out to 21 days.

For experimental TBI, male Sprague–Dawley rats (200–225 g Harlanlabs, Indianapolis, IN) were subjected to severe unilateral cortical contu-sion (1.5 or 2 mm penetration see [13–15]). The cortical contusion used inthese experiments results in severe behavioral deficits, significant loss ofcortical tissue, blood–brain barrier disruption, and loss of hippocampalneurons [15–18], mimicking the sequelae of human closed-head injury.Animals were anesthetized with isoflurane (2%) and placed in a stereotaxicframe (Kopf Instruments, Tujunga, CA) prior to injury. Using sterileprocedures, the skin was retracted, and a point was identified midwaybetween bregma and lambda, and midway between the central suture andthe temporalis muscle laterally. This location, which was consistent

1270 G. Shaw et al. / Biochemical and Biophysical Research Communications 336 (2005) 1268–1277

between all animals used in this study, constituted the central location ofthe 6 mm diameter circular craniotomy made with a hand-held Micheletrephine (Miltex, York, PA). The skull cap was carefully removed withoutdisruption of the underlying dura. Prior to the injury the head of theanimal was angled in a medial to lateral plane so that the impacting tipwas perpendicular to the exposed cortical surface. This was accomplishedby rotating the entire stereotaxic frame in the transverse plane whileleaving the nose bar at �5.0. The exposed brain was injured using apneumatically controlled impacting device (PSI, Fairfax, VA) with a 5-mmbeveled tip, which compressed the cortex to a depth of 1.5 or 2 mm dis-placement at 3.5 M/s, corresponding to moderate or severe unilateralinjury, respectively [17,19]. Control animals received a craniotomy andwere placed in the stereotaxic frame, but did not receive a cortical con-tusion. In both groups, Surgicel (Johnson and Johnson, Arlington, TX)was laid upon the dura and the skull cap replaced. A thin coat of dentalacrylic was then spread over the craniotomy site and allowed to dry beforethe wound was stapled closed. During all surgical procedures and recov-ery, the core body temperature of the animals was maintained at 36–37 �Cusing heating pads. Blood was typically collected before injury, at 1, 2, 4,6, 12, and 24 h post-injury, and 2, 3, 4, 5, 7, 14 and in some cases 21dayspost-injury.

To collect blood, all rats were restrained and the tail was cleaned withan antiseptic solution. A sterile scalpel blade was used to make a trans-verse section through the long axis of the tail approximately 0.5–1.0 mmfrom the tip immediately prior to the first collection. 100–200 lL of bloodwas collected in a sterile microcentrifuge tube and sterile gauze held to thetail for 1–3 min to facilitate hemostasis. For subsequent blood collection,the clot/scab was removed, the tail gently massaged, and blood collectedas above. Following collection, blood was kept on ice and then stored at�80 �C. Once the entire series of samples for a particular animal had beencollected, all were thawed out and centrifuged at 14,000 rpm in anEppendorf centrifuge at room temperature. The plasma supernatant wasthen used for ELISA. All animal procedures were approved by theappropriate Animal Care and Use Committees at the Universities ofFlorida and Kentucky.

Quantification of protein levels in normal CNS regions. Female Long–Evans Hooded rats (220–250 g, Harlan Labs, Indianapolis, IN) weredeeply anesthetized (sodium pentobarbital 40–60 mg/kg, IP), thendecapitated, and CNS tissue samples were quickly dissected out, weighed,frozen in liquid nitrogen, and stored at �80 �C. Samples of different re-gions of the CNS were homogenized at 10 mg/ml wet weight in 4 M urea,1 mM EGTA, 1 mM EDTA, 0.2 mM PMSF, and 10 mM Tris/HCl, pH7.2, as described by Hashimoto et al. [20], and then centrifuged for 5 minat 14,000 rpm in an Eppendorf centrifuge. The supernatants, here de-scribed as the urea soluble protein fraction, were used for SDS–PAGE andELISA. The small pellet which presumably contains extracellular matrixand other insoluble proteins was discarded. Protein gels were run out at100 lg wet weight per lane and stained with Coomassie Brilliant BlueR250. Blots from these gels were generated by standard blotting tech-niques and probed with antibodies to pNF-H. The extracts were alsoapplied to ELISA plates as serial dilutions as described for the serumsamples. Protein levels of pure preparations of pNF-H were quantified bygel scanning and using the Pierce micro-BCA assay.

Results

Antibody specificity and ELISA development

In pilot experiments, we found that an affinity purifiedchicken NF-H antibody very effectively captured pureNF-H, and that an affinity purified rabbit NF-H antibodywas effective in the detection role. The rabbit antibody wasitself detected with a commercial goat anti-rabbit alkalinephosphatase conjugate. A series of experiments tested dif-ferent concentrations of these antibodies to optimize the

signal to noise ratio, and it was clear that the assay reliablyand sensitively detected NF-H in various kinds of solution.Having developed a workable ELISA we then character-ized the capture and detection antibodies in detail. Theseantibodies were developed by EnCor Biotechnology (Alac-hua, FL) and are referred to as CPCA-NF-H and RPCA-NF-H. The immunogen was the �220 kDa apparentmolecular weight form of NF-H isolated from bovinespinal cord, which corresponds to the hyperphosphorylat-ed, axonal form of this protein. On Western blots of wholerat nervous system homogenates both antibodies stronglystain the �200 kDa band corresponding to phosphorylatedrat NF-H, which is known to have a slightly higher mobil-ity in SDS–PAGE than the homologous bovine protein[21]. In both cases, there was some staining under the major200 kDa NF-H band, which may represent in vivo break-down products of NF-H, less phosphorylated forms ofNF-H or phosphoisotypes of NF-M, to which many NF-H antibodies cross-react [22]. However, following affinitypurification on the immunogen, both antibodies revealeda much greater specificity for the hyperphosphorylated200 kDa NF-H protein band (Fig. 1A), with no obviousstaining of NF-M or other lower molecular weight materi-al. In addition, both affinity purified antibodies failed torecognize enzymatically dephosphorylated NF-H or a ser-ies of recombinant fusion proteins including the non-phos-phorylated C-terminal tail regions of rat NF-H and NF-M[23]. The immunocytochemical properties of these two anti-bodies are very similar to those of SMI31, SMI35, RT97,NE14, and many other monoclonal antibodies specific forphosphorylated forms of NF-H [24]. In sections of nervoustissue and neurons in tissue culture, both antibodies stainedaxonal profiles strongly and specifically, showing no stain-ing of neuronal perikarya, which however can be stainedstrongly with antibodies such as SMI33, which recognizesa non-phosphorylated form of NF-H. Fig. 1B shows stain-ing, in the red channel, of rat cortical neuron/glia cultureswith AP-CPCA-NF-H (upper panel) and AP-RPCA-NF-H (lower panel). The specimens are also stained, in thegreen channel, with SMI33, which binds to a phosphateindependent epitope on NF-H. Neither affinity purifiedantibody stains perikaryal or dendritic NF-H, which areclearly and strongly visualized with SMI33. Instead, bothstain the longer finer axonal profiles in these cultures. Weconclude that both antibodies are directed against thehyperphosphorylated, axonal forms of NF-H, and thatan assay based on these two reagents will capture and de-tect only this axonal form of NF-H, which we here referto as pNF-H.

Having obtained a convincingly specific signal with ourprototype ELISA, we ran out samples of known amountsof bovine serum albumin (BSA) on SDS–PAGE gels alongwith samples of pure pNF-H. We quantified the relativelevels of the two proteins using both gel scanning and thePierce micro BCA protein assay. For the prototype pNF-H ELISA we standardized on 1 h incubation times in eachantibody reagent and on a 1 h chromagen development

Fig. 1. Antibody specificity. (A) Western blots of AP-CPCA-NF-H and AP-RPCA-NF-H antibodies on homogenates of rat spinal cord. Left lane (CBB)shows 100 lg wet weight of Coomassie brilliant blue stained protein extract from adult rat brain stem. Middle lane (Chk) is a blot of same material stainedwith AP-CPCA-NF-H (1:5000 dilution of �40 lg/ml). Right lane (Rab) is a corresponding blot stained with AP-RPCA-NF-H (1:2000 dilution of �65 lg/ml). (B) Mixed neuron/glial cultures derived from newborn rats, grown for 10 days in tissue culture, were double labeled with SMI33 (green channel) andAP-CPCA-NF-H (red channel, top figure) or SMI33 (green) and AP-RPCA-NF-H (red channel, bottom figure). Both specimens were also counterstainedwith DAPI (blue). Field of view of both images is 287 by 216 lM.

G. Shaw et al. / Biochemical and Biophysical Research Communications 336 (2005) 1268–1277 1271

time which gave us a saturating signal with 50 ng/ml ofpure pNF-H (Fig. 2A). These standard curves are the aver-age of 16 separate determinations, and error bars indicate 1standard deviation. Since we routinely assay a total of 50 llsample per well, 50 ng/ml corresponds to an absoluteamount of 2.5 ng pNF-H. The assay gives a typical timedependent sigmoidal ELISA dilution profile with goodreproducibility, the error bars showing one standard devi-ation (Fig. 2A). The data obtained after a 1 h developmenttime were normalized against the average background (typ-ically 0.2 OD units or below) and saturation level (�1.6OD units) of the assay and are plotted in Fig. 2B. Thesedata show a linear detection range from 0 to about 5 ng/ml, the assay producing a 10% maximum signal with only1.25 ng/ml pNF-H. The inset in Fig. 2B shows the responseat very low levels of pNF-H. The sensitivity of the assaycan be conservatively represented as the first level ofpNF-H which produces a signal above the signal fromthe blank standard plus 3 standard deviations from thatstandard, which correspond to �0.4 ng/ml pNF-H, or anabsolute amount of 20 pg pNF-H (Fig. 2B, inset). The nar-row error bars indicate excellent reproducibility, and repre-sent about 10% variability throughout the range, withsomewhat greater variability at the lowest levels and lessas the assay approaches saturation.

Studies of spinal cord and traumatic brain injury

We had hypothesized that pNF-H would be releasedinto sera following CNS injury, but not be present undernormal circumstances. In fact, no pNF-H was detected in

the sera of untreated rats, while pilot experiments showedthat the prototype ELISA detected pNF-H in sera obtainedpost-mortem from rats given severe experimental spinalcord injuries (SCI) 1–7 days earlier as part of other studies.Post-mortem sera from sham controls (given laminecto-mies but no CNS injury) and normal animals withoutCNS injury showed no detectable levels of serum pNF-Hin these pilot experiments.

We then performed more detailed studies on a series of25 animals given experimental spinal contusion and knife-cut lesion injuries. Because we had detected no serum pNF-H signal in any of the sham animals studied previously, noshams were included in this series of experiments. Addi-tionally, because pilot experiments showed the presenceof strong serum pNF-H signals we standardized on only10 ll of plasma in the first row of the ELISA well, whichfollowing serial dilution became 5, 2.5, and 1.25 ll in thefirst, second, and third wells, respectively. We obtainedtime course data from 8 animals which were given spinalcontusion injuries and 13 animals that received a spinalcord hemisection. Interestingly, almost every animal, irre-spective of the experimental treatment, revealed a biphasicpNF-H serum signal; pNF-H levels rose rapidly in the firstfew hours post-injury and produced a sharp peak at about16 h post-injury. The levels then declined somewhat androse again to produce a second broader and usually higherpeak of pNF-H serum expression with a maximum at 3days. The level of serum pNF-H then declined slowly downto baseline levels over the following few days, but frequent-ly showed later smaller peaks of pNF-H expression.Fig. 3A shows the average pNF-H serum signal from 2 h

Fig. 2. Prototype ELISA characterization. (A) Standard curves obtainedwith 1:1 serial dilutions of pure pNF-H, starting at 50 ng/ml, applied toELISA plates coated with optimized concentrations of AP-CPCA-NF-H,and binding of pNF-H detected with AP-RPCA-NF-H. Data showaverage of 16 determinations, and error bars show one standard deviation.(B) Shows normalized data for one hour development time, and the insetshows the data obtained at very low levels of pNF-H.

Fig. 3. Experimental SCI results. (A) Shows average results from threerats given T11 hemisections at time 0. Blood samples were processed forELISA at the indicated times post-injury, and error bars are standarddeviations. (B) Shows data from an individual rat, not part of the data setused to generate (A), which was given a T11 hemisection and bloodsamples taken out to 18 days post-injury (each data point average of twodeterminations). (C) Shows a rat given a spinal contusion injury withserum samples taken out to 19 days. In both (B,C), 5 ll of serum was usedto produce the ‘‘100,’’ plot, and the ‘‘50’’ and ‘‘25’’ plots correspond todata from serial dilutions of this material, i.e., 2.5 and 1.25 ll of serum.

1272 G. Shaw et al. / Biochemical and Biophysical Research Communications 336 (2005) 1268–1277

to 7 days post-injury from three representative animalswith T11 hemisection injuries. The clearly biphasic natureof the signal is obvious. The error bars indicate standarddeviations and show that by 8 h the average signal is al-ready significantly above background levels. The pNF-Hsignal had not returned to baseline by 7 days, thus in laterexperiments blood was collected out to 3 weeks post-injury.Fig. 3B shows serum levels of pNF-H from 2 h to 18 dayspost-hemisection (rat not part of the data set used inFig. 3A). The data are similar to those in Fig. 3A, butthe extended time line shows that by 18 days the pNF-Hserum signal was nearly down to baseline. For this experi-ment we also plotted the data from the first, second, andthird wells of the ELISA, indicated by 100, 50, and 25, cor-responding to 5, 2.5, and 1.25 ll of plasma, respectively.The similar shape of the three profiles builds confidencein the sensitivity and reliability of the ELISA. Fig. 3Cshows similar data from a rat with a spinal cord contusioninjury. Again the biphasic nature of the signal is clear, as is

the steeper nature of the first peak compared to the second.This particular specimen is a little unusual in that the firstpeak is higher than the second peak. We also noted latersmaller peaks of serum pNF-H >9 days after the originalinjury in many of the animals. These later peaks are partic-ularly obvious in the data shown in Fig. 3C. By compari-son with appropriate standards we calculated that themaximum level of pNF-H detected in these SCI experi-ments was as high as 250 ng/ml pNF-H in the 3–5 daypost-injury period following injury.

G. Shaw et al. / Biochemical and Biophysical Research Communications 336 (2005) 1268–1277 1273

We also performed a series of studies on rats subjectedto a controlled cortical impact (CCI), a model of TBI.Since, unlike the SCI experiments described above, wehad no pilot studies to work with, we also performed aseries of sham craniotomy controls. Nine animals weregiven relatively mild 1.5 mm CCI injuries and four morewere given sham craniotomies (Fig. 4A). Since pilot exper-iments showed that the serum pNF-H signals were muchweaker than those obtained from the SCI experiments,the first lane of the ELISA was incubated, following serialdilution, with 25 ll of serum rather than 5 ll as in the SCIexperiments. In the experimental group, we noted a rapidincrease in the average level of serum pNF-H in the fewhours following injury, coupled with a less rapid increasepeaking at 2 days post-injury. Levels of pNF-H declinedslowly back to baseline over the following few days. Therewas some variability in the response of individual animals,as indicated by the relatively large standard deviations ob-tained. The sham controls gave, in aggregate, a tiny pNF-H serum signal in the first few hours after injury, perhapsdue to section of peripheral nerves during the fairly inva-sive craniotomy procedure, and perhaps also due to the

Fig. 4. Experimental TBI results. (A) Nine rats were given 1.5 mm CCIcompared to four sham craniotomy controls. Error bars representstandard deviations. No serum pNF-H was detected prior to theoperation, but levels of pNF-H rapidly increased immediately followinginjury, the CCI animals showing significantly stronger pNF-H signals. (B)Shows the serum pNF-H response from an animal given a 2 mm CCI,revealing a biphasic response reminiscent of the SCI data shown in Fig. 3.25 ll of serum was used to produce the ‘‘100’’ plot, and the ‘‘50’’ plot wasproduced by a 1:1 dilution of this material.

technical difficulty of generating shams without damagingthe brain. In fact, the tiny peak noted is largely due to theresponse of a single animal, which may have receivedinadvertent cortical damage. Despite this, the animals giv-en this CCI injury still had pNF-H serum levels whichwere statistically significantly above the sham craniotomycontrols by 12 h post-injury. The average peak level ofexpression of serum pNF-H observed was about 20 ng/ml, much lower than that seen in the SCI experiments.Although a biphasic serum pNF-H response was not asobvious as seen with the SCI animals, the shape of the ser-um response profile was consistent with the superimposi-tion of two peaks, and in some individual animals twopeaks were clearly visible. Fig. 4B shows data from an ani-mal given a more severe 2 mm CCI, revealing a clear peakof serum pNF-H expression at 12 h post-injury and a sec-ond distinct peak at 2 days post-injury. This animal alsohad a peak level of serum pNF-H of about 30 ng/ml sug-gesting that more severe injuries result in greater levels ofserum pNF-H.

pNF-H expression in different CNS regions

A possible explanation for the much stronger serumpNF-H signal following SCI compared to TBI is thatpNF-H is simply more abundant in spinal cord than brain.Fig. 5A shows Coomassie brilliant blue stained protein gelsof equivalent amounts of urea soluble protein from theindicated regions of the CNS; there is an obvious proteinband running at �200 kDa in the spinal cord and brainstem extracts, though this band was less obvious in cerebel-lum and not distinct in the cortical extract. The ELISAshowed that the levels of pNF-H in spinal cord and brainstem were at least an order of magnitude higher than thatin cortical regions, with cerebellum containing intermediatelevels (Fig. 5B). These CNS samples contained between 7.1and 0.58 ng of pNF-H per 25 lg of wet weight of materialapplied to the first well. The BCA protein assay showedthat rat cortex contains about 5% wet weight of urea solu-ble protein, while brain stem and cerebellum are each about3.7% urea soluble protein. Spinal cord, presumably due tothe higher relative lipid content, is only about 2.8% ureasoluble protein by weight. We can conclude that about1% of the urea soluble protein in rat spinal cord is pNF-H. In brain stem, cerebellum, and cortex, the percentagecontent is about 0.54%, 0.23%, and 0.05%, respectively.These figures are consistent with but are likely to be farmore accurate than quantitative scans of the gel lanes inFig. 5A.

Discussion

Informative biomarkers detectable in serum are conve-nient for researchers and clinicians since blood is routinelyand much more easily obtained than CSF. Since neurofila-ments are found only in neurons, the detection of releasedneurofilament subunits indicates unambiguously that

Fig. 5. Levels of NF-H in different rat CNS tissues. (A) Tissues were weighed, dissolved in urea buffer, 100 lg wet weight of material was run per lane andstained with Coomassie brilliant blue R250. The 200 kDa band corresponding to the hyperphosphorylated axonal form of NF-H (pNF-H) is clearly visiblein the spinal cord and brain stem, but is relatively much less abundant in cerebellum and especially cortex. (B) Levels of pNF-H as quantified by ELISAshow higher levels of pNF-H in spinal cord and brain stem than cerebellum and especially cortex, in excellent agreement with (A).

1274 G. Shaw et al. / Biochemical and Biophysical Research Communications 336 (2005) 1268–1277

neurons have been damaged or destroyed. Here, we de-scribe a pNF-H ELISA and use it to show that pNF-Hcan be readily detected in surprisingly large amounts inthe sera of rats given experimental SCI and, in smallerbut still significant amounts, in the sera of rats given exper-imental TBI. Since the hyperphosphorylated form of NF-His found only in axons, the assay must be measuring axonalinjury in a novel and sensitive fashion. The current assayuses two affinity purified polyclonal antibodies made intwo different, distantly related, species. Polyclonal antibod-ies are expected to be effective both in protein capture anddetection roles [25], and this may account for the high sen-sitivity of the assay. The unusual protein chemical proper-ties of pNF-H, in particular the up to 60 potentialphosphoepitopes comprising the KSP phosphorylationsites, also make pNF-H a particularly attractive candidatebiomarker. It seems reasonable to propose that the amountof pNF-H detected will correlate with the amount of axon-al injury, degeneration, and neuronal death, so that moreserious CNS injury will result in a relatively greater serumpNF-H signal. The level of serum pNF-H associated with aparticular type of injury may therefore have diagnostic andprognostic value, and we are currently performing experi-ments to test this hypothesis.

We also show details on the time course of serumpNF-H expression following two forms of experimentalSCI. These data reveal two peaks of pNF-H expression;a sharper one at about 16 h and another less steep peakat about 3 days. The animals given experimental TBI alsoappeared to have a biphasic response, although the timecourse was shorter, with one peak at about 12 hoursand the second at about 2 days. Possibly the first peakis derived from axons acutely damaged by the injury,and the second corresponds to pNF-H release as a

consequence of secondary axonal degeneration. FollowingSCI we noted some later peaks in the second and eventhird week after injury, presumably indicating the degen-eration of significant numbers of axons well after the ori-ginal insult. We are currently performing experimentsaimed at trying to understand the significance of thesevarious peaks.

We were initially surprised at the much higher levels ofserum pNF-H detected following SCI compared to TBI.However, the rat spinal cord is dominated by bundles oflarge diameter projection axons, expected to be rich inpNF-H. In contrast, cortical gray matter has narroweraxonal profiles in a background rich in neuronal cellbodies, dendrites, glia, and vasculature. Larger diameteraxons in myelinated bundles are found in the brain, butthey tend to be concentrated in the deeper subcortical whitematter. In line with this, our ELISA and SDS–PAGE anal-ysis showed that pNF-H was more than one order of mag-nitude more abundant in rat spinal cord than brain, so thata comparable brain injury would be expected to release sig-nificantly lower amounts of pNF-H than a spinal cord inju-ry, as we in fact observed. We also note that use of theassay to measure pNF-H levels in tissue samples will beuseful in future studies of the progression of neuronal dif-ferentiation, maturation, axon elongation, intoxication,and disease, since pNF-H levels are expected to alter inall these states.

We noticed some variability in the serum response toTBI as revealed by the relatively large standard deviationsshown in Fig. 4A. A contributory factor to this may be thatpNF-H levels are highest in white matter deep within thebrain, which is likely to show variable degrees of damagefollowing the type of surface cortical injury used here. Itis also possible that the particular TBI paradigm used

G. Shaw et al. / Biochemical and Biophysical Research Communications 336 (2005) 1268–1277 1275

results in a variable degree of blood–brain barrier damageand that this may contribute to a variable degree of pNF-Hrelease into the blood. We also noted small serum pNF-Hsignals from the averaged sham TBI animals, though, asnoted in the results section, the size of even this small peakwas exaggerated by the results from a single animal. Thesham paradigm is invasive in itself and may lead to inad-vertent cortical damage. However, the craniotomy mustalso cause the severing of some peripheral axons in the skinand skull, which are potential sources of small amounts ofpNF-H. We are currently performing other experimentsaimed at clarifying these issues.

The small calcium binding protein S100b, neuron spe-cific enolase (NSE), and microtubule associated proteintau have all been proposed as biomarkers of various kindsof neuronal injury, with the greatest attention to datebeing focused on TBI [26,27]. Although some S100b isfound in neurons, it is much more heavily concentratedin astroglia in the CNS and is also found in many othercell types outside the nervous system. The widespreadexpression of S100b means that elevated levels of this pro-tein in CSF or serum can have many possible causes. Forthis reason the specificity and usefulness of S100b as amarker of neuronal injury has recently been subject to de-bate [28]. Tau is more specifically expressed in neurons,but is also found in astroglia [29], and certain other celltypes throughout the body [30]. Zemlan et al. [31] havemeasured levels of c-tau, a proteolytic fragment derivedfrom tau, in both CSF and serum of TBI patients. Levelsof c-tau are elevated in CSF following nervous systeminjury and various types of neurological disease. The onlystudy of serum levels of c-tau after neuronal injury re-vealed amounts in the low nanogram range, significantlyless than the peak levels of pNF-H detected here [32]. Sev-eral previous studies have examined NF-L expression inCSF, but not as of yet in serum (see [33] and referencestherein). Other potential biomarkers of neuronal injuryinclude neuron specific enolase, myelin basic protein,and PGP9.5/ubiquitin C-terminal hydrolase 1 [26,27].These proteins also have not been extensively examinedin serum, and, where data have been presented, the levelsdetected are significantly lower than we have demonstrat-ed with pNF-H. In summary, the present study is the firstto demonstrate large and easily detectable amounts of aneuron specific protein in serum following experimentalCNS injury.

There is considerable literature on changes in the level ofpNF-H, generally detected immunocytochemically, in avariety of damage and disease states [34,35], but relativelyfew ELISA studies. One group used ELISA to look at theexpression of pNF-H in human brain and found a specificincrease in the levels of pNF-H in samples from Alzhei-mer�s patients [20]. More recently Hu et al. [36] showed thatpNF-H was detected at elevated levels in the CSF of Alz-heimer�s patients. After we had initiated the studies de-scribed here, Petzold et al. [37] described a very sensitiveNF-H ELISA developed using the commercially available

monoclonal antibody SMI35 as the capture antibody.SMI35 recognizes preferentially the phosphorylated formsof NF-H and is very similar in specificity to the twopNF-H antibodies described here. As the work reportedhere was being completed this group used their assay to de-tect pNF-H in the plasma of a group of individuals withoptic neuritis [38]. This recent paper from this group is,to our knowledge, the first report showing pNF-H inblood, although in a different context from the work pre-sented here. Both ELISAs produce extremely similar re-sults on human CSF samples from patients with a varietyof brain disorders (Petzold and Shaw in preparation).These and other findings [36] show that certain specifickinds of human neurological disease are associated withgreatly elevated levels of CSF and possibly also serumpNF-H. The examination of serum and CSF pNF-H istherefore likely to be fruitful for future studies of bothexperimental and clinical disease states. Clearly also manyother questions are raised by the present findings. Whatform of pNF-H is being detected, a proteolytic fragmentor intact molecules? What is the half-life of pNF-H in ser-um? What is the significance of the two peaks of serumpNF-H expression? Are other neurofilament subunits orfragments of them also detectable in serum samples frominjured animals? Work to address all these issues isunderway.

The present prototype assay is non-invasive, requiring50 ll or less of blood, and is quick to perform, the entireELISA described here requiring less than 4 h. The currentassay can be used to quantify serum pNF-H in an individ-ual animal in an ongoing experiment, as shown here. Itseems likely that the information will be particularly usefulin experimental studies aimed at ameliorating neuronaldeath and axonal injury by therapeutic strategies; if the sec-ond broad peak of serum pNF-H expression correspondsto secondary neuronal death, drugs which reduce the levelof this peak are presumably working by reducing second-ary neuronal loss. There are many possible means to makethe assay described here more rapid and sensitive, forexample, by using appropriate avidin–biotin conjugatesand/or fluorescence-based detection methods. In addition,recently described methods using capture and detectionantibodies bound to appropriate metallic nanoparticlesare claimed to increase sensitivity by several orders of mag-nitude over that obtained by ELISA [39–41]. Antibody re-agents like those here also should allow the development ofcolorimetric or filter based methods that would permit ra-pid and convenient detection of pNF-H in serum. Suchapproaches could also potentially be useful in diagnosisof human accident victims. In particular, the degree of neu-ronal injury in humans is difficult to determine using X-rays, CAT scans, MRI or other current imaging methods,and in some situations these instruments may not be avail-able. The present work may therefore be the first step to-wards the development of a diagnostic kit allowing therapid determination of the degree of neuronal injury fol-lowing trauma.

1276 G. Shaw et al. / Biochemical and Biophysical Research Communications 336 (2005) 1268–1277

Acknowledgments

This work was supported by grants from the Departmentof Veterans Affairs, the C.M. and K.E. Overstreet Endow-ment, NIH NS39828, and by private funding from EnCorBiotechnology Inc. Gerry Shaw holds equity in EnCor Bio-technology Inc., a company commercializing the antibodiesand potentially the ELISA used in this study, and may ben-efit by receiving royalties or equity growth. The use of neuro-filament proteins as biomarkers of neuronal injury isintellectual property described in the pending US patent20040241762, submitted by Gerry Shaw and Brian Pike.We thankYongWang for excellent technical assistance, Sar-ah Sumner, andWilbur O�Steen, for assistance with injuries,animal care, and sample collection, and Colin Sumners andMohan Raizada for rat cortical neuron/glia cultures.

References

[1] M.J. Strong, W.L. Strong, H. Jaffe, B. Traggert, M.M. Sopper, H.C.Pant, Phosphorylation state of the native high-molecular-weightneurofilament subunit protein from cervical spinal cord in sporadicamyotrophic lateral sclerosis, J. Neurochem. 76 (2001) 1315–1325.

[2] W.W. Schlaepfer, C. Lee, V.M. Lee, U.J. Zimmerman, An immuno-blot study of neurofilament degradation in situ and during calcium-activated proteolysis, J. Neurochem. 44 (1985) 502–509.

[3] M.E. Goldstein, N.H. Sternberger, L.A. Sternberger, Phosphoryla-tion protects neurofilaments against proteolysis, J. Neuroimmunol. 14(1987) 149–160.

[4] H.C. Pant, Dephosphorylation of neurofilament proteins enhancestheir susceptibility to degradation by calpain, Biochem. J. 256 (1988)665–668.

[5] G.V. Johnson, J.A. Greenwood, A.C. Costello, J.C. Troncoso, Theregulatory role of calmodulin in the proteolysis of individual neurofil-ament proteins by calpain, Neurochem. Res. 16 (1991) 869–873.

[6] J.A. Greenwood, J.C. Troncoso, A.C. Costello, G.V. Johnson,Phosphorylation modulates calpain-mediated proteolysis and cal-modulin binding of the 200-kDa and 160-kDa neurofilament proteins,J. Neurochem. 61 (1993) 191–199.

[7] G. Shaw, C. Yang, L. Zhang, P. Cook, B.R. Pike, W.D. Hill,Characterization of the bovine neurofilament NF-M protein andcDNA sequence and identification of in vitro and in vivo calpaincleavage sites, Biochem. Biophys. Res. Commun. 325 (2004) 619–625.

[8] A. Delacourte, G. Filliatreau, F. Boutteau, G. Biserte, J. Schrevel,Study of the 10-nm-filament fraction isolated during the standardmicrotubule preparation, Biochem. J. 191 (1980) 543–546.

[9] A. Polson, Isolation of IgY from the yolks of eggs by achloroform polyethylene glycol procedure, Immunol. Invest. 19(1990) 253–258.

[10] J.G. Evans, C. Sumners, J. Moore, M.J. Huentelman, J. Deng, C.H.Gelband, G. Shaw, Characterization of mitotic neurons derived fromadult rat hypothalamus and brain stem, J. Neurophysiol. 87 (2002)1076–1085.

[11] J.A. Gruner, A monitored contusion model of spinal cord injury inthe rat, J. Neurotrauma 9 (1992) 123–126 [discussion 126–128].

[12] M.L. Lemons, D.R. Howland, D.K. Anderson, Chondroitin sulfateproteoglycan immunoreactivity increases following spinal cord injuryand transplantation, Exp. Neurol. 160 (1999) 51–65.

[13] P.G. Sullivan, J.N. Keller, W.L. Bussen, S.W. Scheff, Cytochrome crelease and caspase activation after traumatic brain injury, Brain Res.949 (2002) 88–96.

[14] S.W. Scheff, P.G. Sullivan, Cyclosporin A significantly amelioratescortical damage following experimental traumatic brain injury inrodents, J. Neurotrauma 16 (1999) 783–792.

[15] P.G. Sullivan, J.N.Keller,M.P.Mattson, S.W. Scheff, Traumatic braininjury alters synaptic homeostasis: implications for impaired mito-chondrial and transport function, J. Neurotrauma 15 (1998) 789–798.

[16] S.A. Baldwin, T. Gibson, C.T. Callihan, P.G. Sullivan, E. Palmer,S.W. Scheff, Neuronal cell loss in the CA3 subfield of the hippocam-pus following cortical contusion utilizing the optical disector methodfor cell counting, J. Neurotrauma 14 (1997) 385–398.

[17] S.A. Baldwin, S.W. Scheff, Intermediate filament change in astrocytesfollowing mild cortical contusion, Glia 16 (1996) 266–275.

[18] R. Beer, G. Franz, A. Srinivasan, R.L. Hayes, B.R. Pike, J.K.Newcomb, X. Zhao, E. Schmutzhard, W. Poewe, A. Kampfl,Temporal profile and cell subtype distribution of activated caspase-3 following experimental traumatic brain injury, J. Neurochem. 75(2000) 1264–1273.

[19] P.G. Sullivan, A.G. Rabchevsky, R.R. Hicks, T.R. Gibson, A.Fletcher-Turner, S.W. Scheff, Dose-response curve and optimaldosing regimen of cyclosporin A after traumatic brain injury in rats,Neuroscience 101 (2000) 289–295.

[20] R. Hashimoto, Y. Nakamura, I. Tsujio, H. Tanimukai, T. Kudo, M.Takeda, Quantitative analysis of neurofilament proteins in Alzheimerbrain by enzyme linked immunosorbent assay system, PsychiatryClin. Neurosci. 53 (1999) 587–591.

[21] G. Shaw, E. Debus, K. Weber, The immunological relatedness ofneurofilament proteins of higher vertebrates, Eur. J. Cell Biol. 34(1984) 130–136.

[22] M.J. Carden, W.W. Schlaepfer, V.M. Lee, The structure, biochemicalproperties, and properties, and immunogenicity of neurofilamentperipheral regions are determined by phosphorylation state, J. Biol.Chem. 260 (1985) 9805–9817.

[23] J. Harris, C. Ayyub, G. Shaw, A molecular dissection of the carboxyterminal tails of the major neurofilament subunits NF-M and NF-H,J. Neurosci. Res. 30 (1991) 47–62.

[24] V.M. Lee, L. Otvos Jr., M.J. Carden, M. Hollosi, B. Dietzschold,R.A. Lazzarini, Identification of the major multiphosphorylation sitein mammalian neurofilaments, Proc. Natl. Acad. Sci. USA 85 (1988)1998–2002.

[25] E. Harlow, D. Lane, Antibodies: A Laboratory Manual, Cold SpringHarbor Laboratory Press, New York, 1988.

[26] T. Ingebrigtsen, B. Romner, Biochemical serum markers for braindamage: a short review with emphasis on clinical utility in mild headinjury, Restor Neurol. Neurosci. 21 (2003) 171–176.

[27] T. Ingebrigtsen, B. Romner, Biochemical serum markers of traumaticbrain injury, J. Trauma 52 (2002) 798–808.

[28] R.E. Anderson, L.O. Hansson, O. Nilsson, R. Dijlai-Merzoug, G.Settergren, High serum S100B levels for trauma patients without headinjuries, Neurosurgery 48 (2001) 1255–1258 [discussion 1258–1260].

[29] T. Togo, D.W. Dickson, Tau accumulation in astrocytes in progres-sive supranuclear palsy is a degenerative rather than a reactiveprocess, Acta Neuropathol. (Berl) 104 (2002) 398–402.

[30] Y. Gu, F. Oyama, Y. Ihara, Tau is widely expressed in rat tissues, J.Neurochem. 67 (1996) 1235–1244.

[31] F.P. Zemlan, E.C. Jauch, J.J. Mulchahey, S.P. Gabbita, W.S. Rosen-berg, S.G. Speciale, M. Zuccarello, C-tau biomarker of neuronaldamage in severe brain injured patients: associationwith elevated intra-cranial pressure and clinical outcome, Brain Res. 947 (2002) 131–139.

[32] G.J. Shaw, E.C. Jauch, F.P. Zemlan, Serum cleaved tau protein levelsand clinical outcome in adult patients with closed head injury, Ann.Emerg. Med. 39 (2002) 254–257.

[33] W.J. Van Geel, L.E. Rosengren, M.M. Verbeek, An enzymeimmunoassay to quantify neurofilament light chain in cerebrospinalfluid, J. Immunol. Methods 296 (2005) 179–185.

[34] A. Al-Chalabi, C.C. Miller, Neurofilaments and neurological disease,Bioessays 25 (2003) 346–355.

[35] Q. Liu, F. Xie, S.L. Siedlak, A. Nunomura, K. Honda, P.I. Moreira,X. Zhua, M.A. Smith, G. Perry, Neurofilament proteins in neurode-generative diseases, Cell Mol. Life Sci. 61 (2004) 3057–3075.

[36] Y.Y. Hu, S.S. He, X.C. Wang, Q.H. Duan, S. Khatoon, K. Iqbal, I.Grundke-Iqbal, J.Z. Wang, Elevated levels of phosphorylated

G. Shaw et al. / Biochemical and Biophysical Research Communications 336 (2005) 1268–1277 1277

neurofilament proteins in cerebrospinal fluid of Alzheimer diseasepatients, Neurosci. Lett. 320 (2002) 156–160.

[37] A. Petzold, G. Keir, A.J. Green, G. Giovannoni, E.J. Thompson, Aspecific ELISA for measuring neurofilament heavy chain phospho-forms, J. Immunol. Methods 278 (2003) 179–190.

[38] A. Petzold, K. Rejdak, G.T. Plant, Axonal degeneration andinflammation in acute optic neuritis, J. Neurol. Neurosurg. Psychiatry75 (2004) 1178–1180.

[39] E.E. Loomans, A.M. van Doornmalen, J.W. Wat, G.J. Zaman, High-throughput screening with immobilized metal ion affinity-based

fluorescence polarization detection, a homogeneous assay for proteinkinases, Assay Drug Dev. Technol. 1 (2003) 445–453.

[40] J.M. Nam, C.S. Thaxton, C.A. Mirkin, Nanoparticle-based bio-barcodes for the ultrasensitive detection of proteins, Science 301 (2003)1884–1886.

[41] D.G. Georganopoulou, L. Chang, J.M. Nam, C.S. Thaxton, E.J.Mufson, W.L. Klein, C.A. Mirkin, From the cover: nanoparticle-based detection in cerebral spinal fluid of a soluble pathogenicbiomarker for Alzheimer�s disease, Proc. Natl. Acad. Sci. USA 102(2005) 2273–2276.