cerebral astrocyte response to micromachined silicon implants...however, brain implants must be...

rtmsnenlcTrwgtschc1ltsiowpdttlecscosaTitpi

Experimental Neurology 156, 33–49 (1999)Article ID exnr.1998.6983, available online at http://www.idealibrary.com on

Cerebral Astrocyte Response to Micromachined Silicon Implants

J. N. Turner,*,†,‡ W. Shain,*,† D. H. Szarowski,* M. Andersen,*,‡ S. Martins,§ M. Isaacson,§ and H. Craighead§*Wadsworth Center, New York State Department of Health, Albany, New York 12201-0509; †School of Public Health, The Universityat Albany, Albany, New York 12201-0509; ‡Biomedical Engineering, Rensselaer Polytechnic Institute, Troy, New York 12180-3590;

and §Applied and Engineering Physics, Cornell University, Ithaca, New York 14853-2501

Received June 2, 1998; accepted October 30, 1998

sadtp

dpfdnpHadmstimiaia

uftuonotmnhT1

The treatment of neurologic disorders and the resto-ation of lost function due to trauma by neuropros-hetic devices has been pursued for over 20 years. Theethodology for fabricating miniature devices with

ophisticated electronic functions to interface withervous system tissue is available, having been wellstablished by the integrated circuit industry. Unfortu-ately, the effectiveness of these devices is severely

imited by the tissue reaction to the insertion andontinuous presence of the implant, a foreign object.his study was designed to document the response ofeactive astrocytes in the hope that this informationill be useful in specifying new fabrication technolo-ies and devices capable of prolonged functioning inhe brain. Model probes fabricated from single crystalilicon wafers were implanted into the cerebral corti-es of rats. The probes had a 1 3 1-mm tab, forandling, and a 2-mm-long shaft with a trapezoidalross-section (200-mm base, 60mm width at the top, and30 mm height). The tissue response was studied byight and scanning electron microscopy at postinser-ion times ranging from 2 to 12 weeks. A continuousheath of cells was found to surround the insertion siten all tissue studied and was well developed but looselyrganized at 2 weeks. By 6 and 12 weeks, the sheathas highly compacted and continuous, isolating therobe from the brain. At 2 and 4 weeks, the sheath wasisrupted when the probe was removed from the fixedissue, indicating that cells attached more strongly tohe surface of the probe than to the nearby tissue. Theater times showed much less disruption. Scanninglectron microscopy of the probes showed adherentells or cell fragments at all time points. Thus, as theheath became compact, the cells on the probe and theells in the sheath had decreased adhesion to eachther. Immunocytochemistry demonstrated that theheath was labeled with antibodies to glial fibrillarycidic protein (GFAP), an indicator for reactive gliosis.he tissue surrounding the insertion site showed an

ncreased number of GFAP-positive cells which tendedo return to control levels as a function of time afterrobe insertion. It was concluded that reactive gliosis

s an important part of the process forming the cellular m

33

heath. Further, the continuous presence of the probeppears to result in a sustained response that pro-uces and maintains a compact sheath, at least par-ially composed of reactive glia, which isolates therobe from the brain. r 1999 Academic Press

INTRODUCTION

The treatment of central nervous system trauma,isease, and age-related degeneration with miniaturerosthetic devices has been a goal of a number of groupsor nearly three decades (13, 18, 19, 30). A number ofevices that stimulate and record from relatively largeerve bundles are becoming a practical reality in theeripheral nervous system (6, 7, 16, 17, 26–28, 46, 47).owever, brain implants must be fabricated differentlynd must be much smaller. Brain prostheses have beeneveloped to the point of clinical application, andillimeter-sized implants are being used for chronic

timulation of certain brain nuclei in humans to relievehe symptoms of Parkinson’s disease (3, 8, 11). Thesemplants, while producing encouraging and even dra-

atic results, are nearly as large as the brain regionnto which they are implanted. Thus, there is consider-ble interest in developing smaller devices that resultn less tissue damage and more focused stimulationnd have the capability of prolonged use.Nano- and microfabrication techniques are being

sed to develop implants that are capable of recordingrom and stimulating very small volumes of brainissue. Such implants have been used to record single-nit events (9, 12, 13, 20, 21, 25) and should be capablef stimulating small groups of functionally relatedeurons and possibly even single neurons. The technol-gy to fabricate such implants has been developed forhe manufacture of solid-state electronic devices, andicromachined neural prostheses based on silicon tech-ologies have been designed and fabricated at a veryigh level of sophistication (20–22, 25, 29–31, 45).hese probes can have cross-sections as small as 30 35 µm and can be as long as 1 cm. They can have
ultiple recording and stimulating electrode pads and
0014-4886/99 $30.00Copyright r 1999 by Academic Press

All rights of reproduction in any form reserved.

cacpnt

dfwiaswrdrtnoeeuiniwanc(wc

m1rgscaofsipItfdfmaici

pbipotstoustt

ctKdiclw2chitb

34 TURNER ET AL.

an even have on-board electronics for amplificationnd signal processing. Implants with multiple probesan even provide a three-dimensional array of electrodeads to independently record from or stimulate a largeumber sites within a fairly small volume of brainissue (9, 21, 25, 33, 45).

However, the use of these devices has been limitedue to the inability to effectively and chronically inter-ace them with the neurons of the brain. After a feweeks, the probe to tissue electrical resistance typically

ncreases to levels that render the devices unusable,nd conventional histology shows a compact cellularheath surrounding the insertion site (14, 21, 40, 42). Itas also shown that the mean density of neurons was

educed only in the immediate vicinity of the probe. Atistances .60 µm the density of neurons in the tissueeturned to control values (14). This indicates thathere is minimal depletion of the neural populationear the probe. The cellular sheath has also beenbserved in brain tissue as a result of inserting wirelectrodes which are larger and fabricated from differ-nt materials (1, 2). Polymer materials have also beensed to coat the probes to make them more biocompat-

ble, but a cellular sheath is still produced. The thick-ess of the cellular sheath was shown to vary depend-

ng on the surface material and whether or not a primeras used to improve adhesion between the polyimidend the silicon probes (42). Implantation of aralditeeedles also produced a cellular sheath and under someonditions a connective tissue capsule was also formed44). The most important point from this literatureith respect to the present study is that a compact

ellular sheath forms under all conditions.It is well known that the brain reacts to wounds byounting a glial cell-mediated protective response (10,

5, 24, 41). Thus, the central nervous system tissueeaction to implants is thought to be dominated byliosis, i.e., the activation of astrocytes. It has beenuggested, on the basis of standard histology, that theellular sheath is composed of glial cells, but this is notdefinitive result due to the highly unusual morphol-

gy of the cells in the sheath and the lack of resultsrom specific labels. Our preliminary results havehown the presence of astrocytes in the sheath asndicated by increased levels of glial fibrillary acidicrotein (GFAP) and the recruitment of microglia (4, 5).n order to obtain a better understanding of the pros-hetic probe–tissue interactions, we have designed andabricated model silicon probes to study the time-ependent responses of rat cerebral cortical tissueollowing probe insertion. The purpose of these experi-ents is to document changes in cortical astrocyte

ctivation, morphology, and distribution following thensertion of our model probes. The probes are microma-hined from single crystal silicon wafers and implanted
nto the cortex of young adult rats. For simplicity, our
robes did not have any electrode pads or electronics,ecause the vast majority of the surface of active probess the same material (SiO2), with only a very smallercentage being occupied by metal electrodes or anyther materials. In other words, we are assuming thathe tissue response is dominated by the silicon dioxideurface, the initial damage upon insertion, and possiblyhe continuous presence of the probe. The ultimate goalf our studies is to qualitatively and quantitativelynderstand the brain response to these types of probeso that future implants can be fabricated of materialshat will minimize the tissue’s response and optimizehe tissue–implant interface.

METHODS

Probe fabrication and preparation. Probes were mi-romachined from single-crystal silicon wafers (380 µmhick and 3 in. in diameter) by photolithography andOH etching. Figure 1 illustrates the fabrication proce-ure which is a standard methodology in silicon process-ng. These methods were developed by the integratedircuit industry and are well-established procedures. Aarge number of probes were fabricated into a singleafer as shown by the projection light micrograph (Fig.A). The probes had a 1 3 1-mm tab with a break-awayonnector which attached them to the wafer for easyandling. The tab was used to hold probes during

nsertion, and the 2-mm-long shaft was implanted intohe brain with the bottom of the tab resting on therain’s surface. Figure 2B is a higher magnification

FIG. 1. The processing steps in the fabrication of model probes.

imttsp

tctp

snTotoTtstDcamocsh

acsdidrsaawtmwambestTtsscwtao

awsdwa

sscaac

35ASTROCYTE RESPONSE TO MICROMACHINED SILICON IMPLANTS

mage of the probe obtained by scanning electronicroscopy (SEM). The cross-section of the shaft was a

rapezoid (base 200 µm; top 60 µm; height 130 µm). Theip of the shaft was tapered to form a wedge with aingle pointed end providing a sharp cutting surface forenetrating the pia and underlying brain tissue.Probe insertion device. Probes were inserted into

he brains of adult (100–125 g) Wistar rats using austom fabricated inserter (Fig. 3) that permitted con-rol of both insertion speed and depth of penetration. A

FIG. 2. The probes after fabrication, but before removal from theilicon wafer. (A) A projection light micrograph of a number of probeshowing the 2-mm-long shaft, 1 3 1-mm tap, and the break-awayonnector that attaches the probes to the wafer. (B) A SEM image ofn individual probe. The trapezoidal cross-section of the shaft ispparent. The sides etch at this angle to the top surface due to therystal orientation of the silicon.

air of brass locking forceps held the micromachined s

ilicon probes. The inserter’s table and drive mecha-ism are mounted on a standard stereotaxic device.he probe’s tab fits into a small slot milled into the tipf one side of the forceps. The top of the tab, oppositehe shaft, registers the probe against the back surfacef the slot aligning the shaft parallel to the forceps axis.he forceps clamps the probe which is removed from

he wafer by a slight twist to break the connector on theide of the tab (Fig. 2). The forceps is then clamped inhe precision drive table (MM-3M-EX-1; Automationevices, Inc.) which is driven by a precision lead screw

onnected to a motor through a 16:1 gearhead and anntibacklash coupling. An encoder is monitored by aicroprocessor that controls the motor and receives

perator input from the control pad. The device isapable of insertion speeds of 2 to 8 mm/s. An insertionpeed of 2 mm/s was used for all insertions reportedere.Probe insertion and preparation of brain tissue for

nalysis. Probes were inserted into the right cerebralortical hemisphere of adult rats near the motor–omatosensory border in the midstriatal region 2 mmistal from bregma. Animals were anesthetized byntraperitoneal injection of tribromoethanol (targetedose 23 mg/100 g body w). When animals no longeresponded to a tail pinch, they were placed in atereotaxic holder on which the insertion device waslso mounted, the head was swabbed with Betadyne,nd an incision was made to expose the skull. A hole,ith a diameter larger than the probe’s tab, was drilled

hrough the bone over the right hemisphere near theidline and the dura was pierced. The silicon probeas aligned above the brain. The probe tip was slowlydvanced, while being observed through a dissectingicroscope, until it just touched the surface of the

rain. It was removed a preset distance previouslyntered in the control unit and then driven at theelected speed toward the brain a distance equal tohe preset distance plus 2 mm (the length of the shaft).he forceps were released and withdrawn away from

he brain surface leaving the probe in place. A piece ofterile dialysis tubing was glued over the hole in thekull with ‘‘crazy glue,’’ the skin was closed with woundlips, an identifying ear clip was made, and the animalas returned to its cage. The tubing ensured that the

ab was not in contact with the skin. After recovery,nimals were returned to the animal facility andbserved on a regular basis.At 2, 4, 6, and 12 weeks following probe insertion

nimals were again anesthetized and perfusion fixedith 4% paraformaldehyde in physiological saline. The

kin on the skull was removed and the brain wasissected with the probe in place. In the few caseshere the bone grew back enough to contact the tab orn infection was noted, animals were removed from the
tudy. Brains were block-fixed for an additional 24 h in

tstfatU

cicfd

hdc

36 TURNER ET AL.

he same fixative and stored in Hepes-buffered Hanks’aline. Transverse cuts through the tissue revealedhat the probe shaft ran through the cortex with the tiprequently embedded in the corpus callosum (Figs. 4nd 5). All procedures were reviewed and approved byhe Wadsworth Center Institutional Animal Care and

FIG. 3. An image of the insertion forceps mounted in the stereotaead are below the ribbon cable, and the flexible coupling is below thrive unit. The mounting block for the forceps (AB) is attached to theonnects to our home built microprocessor control unit.

se Committee. b

Brains were prepared for histology and immunohisto-hemistry by blocking to include both the region contain-ng the probe and the contralateral side to act as aontrol. The probe was carefully withdrawn and storedor further analysis for cellular components and/orebris by light and scanning electron microscopy. The

holder. The precision drive unit is at the top. The dc motor and gearar head. The table moves on the precision shafts on each side of the

ble and the forceps (F) are mounted into the block. The ribbon cable

xice ge

ta

lock of tissue was sectioned into 100- and 300-µm

tsow(PeIttMm1diaihXHp(alpsptsptpw

bmS3lccawde

tirTitdaosbpstdtpmwbaaepadTit

S

saptio

sr


issue sections using a vibratome (Fig. 4). The 300-µmections were further processed for conventional histol-gy by embedding and cutting into thin sections (5 µm)hich were stained either with hematoxylin and eosin

H&E) to demonstrate histologic morphology or witherl’s Prussian blue reaction to demonstrate the pres-nce of iron resulting from local hemorrhages (43).ntact 100-µm-thick sections were used for immunohis-ochemistry and three-dimensional (3-D) image collec-ion by confocal laser scanning microscopy (Bio-RadRC600 mounted on an Olympus IMT-2 inverted lighticroscope) using 4X 0.13 NA, 10X 0.40 NA, and 40X

.0 NA objective lenses. These images provide a clearescription of the 3-D cellular architecture near thensertion sites. Tissue slices were stained using anntibody against GFAP. Sections were prepared formmunohistochemistry by treatment with sodium boro-ydride, solubilized using HBHS containing 2% Triton-100, incubated with 5% bovine serum albumin inBHS containing 1% Triton X-100 to block nonspecificrotein binding followed by rabbit anti-cow GFAP1:100 dilution) and rhodamine-labeled anti-rabbit IgG,nd mounted in glycerol saturated with n-propylgal-ate. The images were displayed as either maximumrojections or as stereo pairs using the ANALYZEoftware from the Mayo Foundation (38, 39). Stereoairs were prepared only from data sets collected withhe 40X objective. The stereo views provide a goodense of the 3-D structure and distribution of GFAP-ositive cells and cell processes. When samples werereated identically, but with the omission of the GFAPrimary antibody, no significant fluorescence signal

FIG. 4. The orientation of the 100- and 300-µm-thick tissueections relative to the probe. The sections are all transversely cutelative to the long axis of the shaft.

as observed. w

Preparation of probes. After removal from therains, probes were evaluated for adherent cellularaterial by confocal light microscopy and by SEM.taining with the fluorescent dye DiI (1, 18-dioctadecyl-,3,38,38-tetramethylindocarbocyanine), 0.125 mg/ml so-ution, in Hepes-buffered Hanks’ saline for 120 minontrasted any lipophilic material, and immunohisto-hemistry against GFAP demonstrated the presence ofstrocyte components. Probes, stained or unstained,ere air-dried, coated with gold/palladium in a plasmaischarge device, and observed in an ETEC scanninglectron microscope.

RESULTS

Two animals were removed from the study at au-opsy. One showed bone regrowth that was so extensivet contacted the tab thereby fixing it to the skull,esulting in motion between the probe and the brain.he second was found to have an infection at the site of

nsertion. Two other animals were removed later asheir tissue showed very extensive implant sites withamage extending along one direction for distancespproaching 1 mm. This appeared to be due to motionf the shaft cutting through the tissue. These animalshowed destruction of tissue including several smalllood vessels as well as microhemorrhages. It was notossible to definitively determine whether this exten-ive damage was produced at insertion or by motion ofhe probe while it was in the living animal. Moreetailed description of these tissues is not included ashese results appeared to be due to handling of therobes and the procedures used during insertion or toethods used to isolate the tab from the skull andound closure. These factors do not influence therains’ reaction to the long-term presence of the probend are beyond the scope of the present study. Inddition, the complication rate has decreased withxperience and refinement of surgical methods. Theresent study is, therefore, limited to tissues fromnimals in which the surgical and insertion proceduresid not appear to produce any gross complications.hus, there were six animals studied 2 weeks after

nsertion and three, four, and three animals, respec-ively, at 4, 6, and 12 weeks.

heath Production in Reaction to Implanted Probes:Conventional Histology

H&E sections were prepared from the 300-µm-thickections cut at three points along the length of the shafts indicated in Fig. 4. These samples are difficult torocess and were especially difficult to align such thathe subsequent thin sections (,5 µm) contained themplantation site oriented transversely to the long axisf the probe shaft. Of the 20 experimental animals, 11
ere studied by thin-section histology, and all were in

tapicmplpismesu

tad4ssad6df

alt

T to

sAp

38 TURNER ET AL.

he 2, 4, and 6 weeks postinsertion time points. All thenimals showed a well-developed sheath of cells com-letely surrounding the insertion site. The cells form-ng the sheath had hyperchromatic nuclei and formed aompact multilayered sheath extending around a lu-en that was similar to the shape and size of the

robe’s shaft (Figs. 6A–6C). At 2 and 4 weeks, thisumen sometimes contained ribbons of cells that ap-eared to be torn from the inside wall of the sheath,ndicating that the cells had attached to the probeurface and were dislodged when the probe was re-oved (Fig. 6A). The lumens otherwise appeared clear

xcept for 1 specimen whose lumen was larger than thehaft cross-section and was partially filled with annidentified amorphous material with a hallow center

FIG. 5. Image of a thick section of brain showing the placement ofhe shaft of the probe penetrates the entire cortex with the tip at or in

FIG. 6. (A–C) From H&E-stained conventional histology sectionsomewhat trapezoidal space and condensed sheath walling off the in–C, respectively, showing hemosiderin stained by Perl’s Prussian b
ostinsertion times, respectively.
he shape and size of the shaft. The sheath was compactnd was usually formed from four to six layers ofensely packed cells with small nuclei. At the 2- and-week time points (Figs. 6A and 6B), some areas of theheath had as few as two cell layers. By 6 weeks, theheath was well formed and compact (Fig. 6C). Occasion-lly, the sheath appeared locally disrupted with tissueamage extending into the surrounding tissue (Fig.C). From the H&E sections, it was not possible toetermine whether this resulted from damage in situ orrom specimen preparation.

The tissue showed rearrangement usually with few ifny large nuclei for a distance of 50 to 100 µm from theumenal or implant edge of the sheath. This distanceended to decrease with time, while the sheath also

probe at the motor–somatosensory border in the midstriatal region.the corpus callosum.

brains showing the probe insertion site which is characterized by ation site with a compact multilayer of cells. (D–F) Serial sections ofreaction. The rows from top to bottom represent 2, 4, and 6 weeks

the

ofserlue

bttnslwtiwocHsrlcat6hssiatoatpts

G

wislodlrw

igdsiFoc5it

afsic4wsiswsfafwTelpl6wac(ltc

Gs

atlit

40 TURNER ET AL.

ecome more compact. The large nuclei were assumedo correspond to neurons, and their distribution beyondhis distance did not appear to be altered. No necroticuclei were observed. The number of nuclei in theheath varied substantially, but this appeared to corre-ate more with position along the shaft and probablyith cortical anatomy than with time after implanta-

ion. No mitotic bodies were observed in the tissue,ncluding within the sheath. The specimens, showing aell-delineated implant site and well-formed continu-us sheath, did not show local hemorrhages as indi-ated by the presence of red blood cells (RBCs) in the&E sections. However, some of the thin sections were

tained for hemosiderin, using Perl’s Prussian blueeaction, which is indicative of excess iron due to theocal breakdown of RBCs. Most tissue sections, espe-ially at 2 and 4 weeks postinsertion, showed a smallmount of positive Prussian blue reaction even whenhe H&E sections did not reveal any hemorrhage (Figs.D–6F). This indicated that there had been someemorrhage probably at insertion of the probe. Hemo-iderin-positive material was observed within theheath and sometimes for a few hundred micrometersnto the tissue. When the latter occurred it was alwaysssociated with tissue disruption as in Fig. 6F. Thisends to indicate that the observed disruption did notccur during tissue processing, but was previous dam-ge associated with the probe either at insertion or dueo its continuous presence during the postinsertioneriod. The presence of this stain demonstrated thathe products associated with microhemorrhages weretill being resolved 6 weeks postinsertion.

FAP Immunoreactivity

GFAP-positive cells are presumed to be astrocytes andere scattered throughout the control tissues. The most

ntense labeling in the controls was seen on processesurrounding blood vessels. The amount and distribution ofabel was consistent with accepted anatomic distributionsf astrocytes. Other GFAP-positive cells were sparselyistributed throughout the tissue and exhibited the stel-ate morphology consistent with astrocytes (Fig. 7). Prepa-ation controls processed without primary antibody butith secondary antibody were negative for fluorescence.There is marked GFAP labeling in the tissue near the

nsertion site, but little if any increase at distancesreater than a few times the probe’s cross-sectionalimensions. Figure 8 shows representative insertionites for all time points. The GFAP labeling is mostntense in the region corresponding to the sheath inig. 6 indicating that astrocytes are a major componentf the sheath. An increased number of GFAP-positiveells were observed in the surrounding tissue as far as00 to 600 µm from the implantation site. This increasen the surrounding tissue was greatest at 2 weeks and
ended to decrease in terms of both the number of cells r
nd the distance away from the insertion site as aunction of postinsertion time. The intense label of theheath also tended to become more compact and morentense with time, corresponding to similar structuralhanges observed in the H&E sections (Fig. 6). At 2 and

weeks, astrocytes with typical stellate morphologyere easily distinguished in the tissue surrounding the

heath (Figs. 8A and 8B), and a large number ofntertwined astrocytic processes projected toward theheath with many being incorporated into it. Astrocytesere arranged around the perimeter of the insertion

ite forming the sheath with some processes extendingor considerable distances along the perimeter (Figs. 8And 8B). Although the sheath was becoming wellormed, individual processes were clearly distinguishedithin it, while cell bodies were not easily visualized.he sheaths’ inner surfaces appeared disorganized orven disrupted with processes partially filling theumens, which at 2 and 4 weeks were smaller than therobes’ cross-section, indicating that the sheath col-apsed when the probe was removed from the tissue. Atand 12 weeks, the labeling close to the insertion siteas very intense, forming a sheath so highly compactnd continuous that few individual processes and noell bodies could be distinguished within the structureFigs. 8C and 8D). The inner surface of the sheath waseft largely intact when the probe was withdrawn fromhe fixed tissue, forming a lumen mimicking the shafts’ross-section and essentially free of any material. This

FIG. 7. Field of control tissue immunohistochemically labeled forFAP. The general tissue labeling is low. The intensely labeled

tructures are astrocytes associated with blood vessels.

FIG. 8. High-magnification confocal microscope projection im-ges of the insertions sites at 2, 4, 6, and 12 weeks (A–D, respec-ively). The specimens are labeled with anti-GFAP antibodies. Thearge number of reactive astrocytes involved in response to thensertion and presence of the probe are obvious. (A and B) The earlyime points, showing disruption of the sheath structure due to probe
emoval; (C and D) a compact sheath with little disruption.

Tt

42 TURNER ET AL.

FIG. 9. Stereo views of the projection images in Fig. 8. They can be viewed in 3-D if the reader crosses his/her eyes or uses a stereo viewer.he continuous nature of the sheath structure and the surrounding astrocyte processes in the third dimension, i.e., through the volume of theissue samples, is demonstrated. (A) 2-week time point; (B) 4 weeks; (C) 6 weeks; and (D) 12 weeks.

iwss

stpr


ndicates that at 6 and 12 weeks the sheath was aell-formed integrated structure which is internally

tronger than the adhesion between cells on the probeurface and the cells forming the sheath.

FIG. 9—

Stereo pairs, in Fig. 8, show the 3-D structure of the i

heath and surrounding GFAP-positive cells. Theangled disordered arrangement of cells and theirrocesses are apparent in 3-D in Figs. 9A and 9Bepresenting the 2- and 4-week time points. The 3-D

ontinued
C
mage as viewed in stereo (by either crossing ones eyes

oatcpisopwtcawcaippicsmie

C

nthwwdwaeallt5tcawcTwmot

sat

tiWawsmsomtolirToT

tFmpccsidhsccioFooswd

C

wsiebfbicdpr

44 TURNER ET AL.

r using a stereo viewer) shows the astrocyte cell bodiesnd their processes projecting in all three dimensionsoward and into the sheath. A denser region of pro-esses corresponding to the sheath with some of therocesses being arranged around the perimeter of thensertion site is clearly seen to be an intertwined 3-Dtructure. Processes partially fill the space previouslyccupied by the probe in a disordered manner. Theserocesses may have been displaced when the probe wasithdrawn suggesting strong astrocyte–probe interac-

ions. A group of such processes projecting into theenter right portion of the lumen appear to end abruptlys opposed to the normal tapering suggesting that theyere broken at that point. At 6 and 12 weeks a

ontinuous, highly compact multilayered sheath ofstrocytes had developed completely surrounding thensertion site. The sheath is seen by stereo viewing toroject through the thick tissue section with numerousrocesses extending into the sheath from the surround-ng tissue. Much less disruption appears to have beenaused by withdraw of the probe indicating that theheath is well developed and that cell-to-cell attach-ent within the sheath is stronger than cell–probe

nteractions. The 3-D nature of the sheath is mostasily seen in Fig. 9D.

ells Strongly Attach to Probes after Insertion

The probes were studied by confocal light and scan-ing electron microscopy after removal from the fixedissue. The confocal light microscopic results (not shownere) indicated the presence of cellular material. Probesere stained with the lipophilic cyanine dye DiIC18(3),hich is lipophilic, and with anti-GFAP. The cyanineye labeling indicates the presence of cell membraneshile the GFAP labeling indicates the presence ofstrocytes. By SEM most of the probes showed clearvidence of adhered cells in tightly associated mono-nd multilayers. However, the distribution of the cellu-ar layers varied greatly. Some probes had thick multi-ayers of cells at the tip of the probes while other probeips were free of any adhered material. Most probes had0% or more of their surfaces covered with cells, buthree probes had little adhered material. High-magnifi-ation images of these latter surfaces showed whatppeared to be regions of very thin cytoplasm but nohole cells. Portions of the surfaces also appeared to be

oated with a extremely thin smooth amorphous layer.he nature of the latter could not be determined. Thereas no correlation between the nature or amount ofaterial coating the probes and the tissue reaction as

bserved by routine histology and GFAP immunoreac-ivity.

Low-magnification scanning electron micrographshow the shape of the probe, including the chisel tip,nd the sides and top of the trapezoidal cross-section of
he shaft (Figs. 10A and 10B). The shaft emerges from c
he tab in the bottom of the image (see Fig. 2 for anmage of a probe still attached to the silicon wafer).

hile about half of the shaft surface in Fig. 10Appears to be free of cells, there are large areas coveredith single or multiple cell layers. The right side

urface is partially covered with what appears to be aonolayer of cells starting about halfway down the

haft from the tip. A large area is seen about two-thirdsf the way along the shaft from the tip where theonolayer becomes a multilayer covering the side of

he shaft and extending over the top surface. The cellsn the top surface appear to be multilayered, while theeft surface is again covered by a monolayer. The inserts a higher magnification view of a portion of the upperight side surface which shows some attached RBCs.his was a 2-week postinsertion probe and was the onlyne that shows a significant number of attached RBCs.he tip of this probe was free of adhered material.Figure 10B is an SEM image from a 6-week postinser-

ion probe and is more heavily covered than the probe ofig. 10A. Most of the covered areas in Fig. 10B wereultilayers which were quite thick. The tip of this

robe, shown at higher magnification in Fig. 10C, wasovered with a thick multilayer of cells that completelyovered all detailed structure of the probe. The cells areo closely associated that it is difficult to distinguishndividual cells. The air-drying process also makes thisifficult as it distorts the cell structure. Figure 10D is aigher magnification image of the center portion of thehaft showing a very thick multilayer of cells, espe-ially on the left surface. The top surface is partiallyovered with cells; the lower edge of the cell multilayer,n the upper portion of the image, shows cells spreadingut over the surface and sending out fine processes.igure 10E shows how thinly the adhered cells spreadn the surface of the probes. A cell whose thick body wasn the top surface of the probe was observed to havepread onto the right side surface in a circular area,hich was so thin that its top edge is difficult toistinguish from the probe’s surface.

onclusions

We have demonstrated that rats can be implantedith our model silicon probes with essentially 100%

urvival rates for animals that did not suffer grossnsertion and/or surgical complications. After a recov-ry period of a few hours, animals exhibit normalehavior. The observed complications, that resulted inour animals being removed from the study, appear toe due to problems of insertion and/or incompletesolation of the probe’s tab from the skull bones. We willontinue to improve the surgical and insertion proce-ures, but it is important to note that experimentsresently being analyzed have a lower complicationate. Insertion of neuroprosthetic devices is a compli-
ated and difficult problem that will be the subject of

pmmsT


FIG. 10. A series of SEM images showing the nature of the cells adhered to the probe. (A) A 2-week probe; (B) a 6-week probe. The tworobes are partially covered with cells which are monolayers in some areas and multilayers in others. The insert in (A) shows an area at higheragnification demonstrating that red blood cells are attached to the probe. (C) The tip of the probe in (B) showing more clearly the thickultilayer of cells that cover the tip. (D) A region in which the cell coverage varies from very thick multilayers to single-cell edges that are

preading thinly over the surface and sending out fine processes. (E) A high-magnification image of a cell spreading over the surface of a probe.he thick cell body extends beyond the upper left corner while the cytoplasm is very thinly spread in a circular shape in the lower right.

fipsmmc3itatctpi

spTtfmaebgttfsttpldotupbcpiilspmtppassp

phafm4nlwEmatteatsnsptbartggtsb

ttecrtwmStatcpvawtptws

46 TURNER ET AL.

uture studies. The insertion speed of 2 mm/s using ournserter or hand insertion both work well with ourrobes as they easily penetrate the pia and appear to gotraight into the brain with little or no buckling. Thisay be due in part to the fairly large cross-sectionaking the probes stiffer with less tendency to buckle

ompared to others with a smaller cross-section (14, 29,0). In any event, the use of a precision mechanicalnsertion device is an advantage as it allows control ofhe insertion speed and orientation of the probe’s longxis with respect to the line of motion through theissue. Even a small misalignment can cause a signifi-ant increase in tissue damage (14). This is an effecthat requires further study, but the construction of ourrecision inserter is a first step to improving themplantation process.

The general tissue response to the implantation ofmall silicon (14, 21, 40, 42, 44) or metal (2, 49)rostheses has been studied by traditional histology.hese reports showed a sheath of cells that surroundedhe implanted probes independent of the materialsorming the probes or their surface coatings. Since ainimum volume of tissue was involved, electrical

ctivity could be recorded, and no particular adverseffects resulted, these probes are often described aseing biocompatible. This description is correct on aross tissue level, but if it is to be related to communica-ion with a smaller or specific population of neurons,he probes used to date, especially chronically, are inact not biocompatible. Previous studies all describe theheath as being a result of gliosis, but this determina-ion was made on the basis of routine histology withouthe use of cell-specific labels. The major results of theresent study were to show that the sheath was, ateast partially, composed of reactive astrocytes and toemonstrate how the sheath was formed as a functionf time. Most previous studies focused on fairly longime points as the implants were intended for chronicse in the brain. Thus, we chose 12 weeks as an endoint with the shortest time being 2 weeks. We showedy conventional histology that as early as 2 weeks aontinuous sheath of cells had formed around therobes and that the sheath was positive for GFAP asndicated by immunohistochemistry. Since GFAP is anndicator for reactive astrocytes, the sheath was ateast partially formed by this cell type. The fact that theheath formed so soon and so continuously around therobes’ shaft could explain the results of electricaleasurements showing that the probe to tissue resis-

ance substantially increases after a relatively shortostinsertion period. Thus, one major conclusion of theresent study is that to better understand the initiationnd formation of the sheath, the process must betudied at time points earlier than 2 weeks. In othertudies (4, 5), we have shown that microglia are also
resent in the sheath. s
The structure of the sheath developed as a function ofostinsertion time in the animal. By conventionalistology, it appeared continuous and densely formed atll times, but increased in thickness and density as aunction of time. The GFAP distributions showed aore varied structure. At the earlier time points (2 andweeks), the GFAP sheath was loosely organized withumerous spaces between the cell processes, but at the

ater times (6 and 12 weeks) it was very compact, and itas impossible to distinguish individual cell processes.ven at 2 weeks the sheath presented a significantechanical and electrical barrier, but by 6 weeks it

ppeared formidable. This may be a very importantime sequence since the functionality of neuropros-hetic devices depends on their capability of forminglectrical connections with neuronal dendrites andxons or at least have sufficiently good electrical connec-ion to the surrounding tissue to record field andingle-unit potentials. However, the regrowth of neuro-al processes into the implantation site is slower thanheath formation. Thus, the sheath would appear torevent regrowing neuronal processes from contactinghe implant and to form a high resistance barrieretween the probe and the brain. Pine and co-workersre developing neuroprosthetic devices which incorpo-ate neurons as part of the device. The somas arerapped in recesses in the probe shaft and the neuritesrow out into the tissue (35, 48). If the neurite out-rowth does not occur before the sheath is well formed,he neurons will never be able to grow through theheath and will never make contact with neurons in therain.The cellular response to the insertion and presence of

he probe is indicated by a combined interpretation ofhe GFAP-labeled images and the SEM images. At thearly time points the GFAP images indicated that theellular structure was disturbed when the probe wasemoved as cell processes appeared broken and filledhe space previously occupied by the probe at 2 and 4eeks postinsertion, but the sheath appeared to beinimally disturbed at 6 and 12 weeks. Furthermore,EM showed cells and/or cell fragments attached to allhe probes, indicating that soon after insertion cellsdhered strongly to the probes and to each other, buthat at later times the cell-to-cell adhesion, betweenells forming the sheath and those adhered to therobe, was weaker. Thus, by 6 weeks the probe wasery well isolated from the surrounding brain tissue,nd the cells on the probe were only weakly associatedith the sheath. There was significant variability be-

ween probes with respect to adhered material. Mostrobes at all time points showed whole cells attached tohe surface some of which were present in multilayershile others were single layers often with widely

pread thin regions of cytoplasm. Three probes did not
how whole cells, but did have what appeared by SEM

tcpcrrp

cilipafiniamcoiihoatmtTmsci

pfatctpoaoscitftadw

ss

nscnittdtTuilsntllcbsfictttWwal

CttCWHAc


o be portions of thinly spread cytoplasm and regionsoated with an extremely thin layer of smooth amor-hous material that was not cellular. There was noorrelation between the nature of the adherent mate-ial or the degree of coverage on the probes to tissueeaction as observed by routine histology or GFAP-ositive immunoreactivity.The general tissue response to the implantation and

ontinuous presence of the probe, as indicated by anncrease in the density of GFAP-positive cells, wasimited to a few hundred micrometers around thensertion site. The distance over which the GFAP-ositive cell density was increased changed with timend returned to control levels a few tens of micrometersrom the sheath by 12 weeks. These observationsndicate that the tissue response may have two compo-ents: first, the initial reaction which is dominated by

nsertion injury and involves the larger tissue region;nd second, the prolonged response which forms andaintains the compact sheath that is caused by the

ontinuous presence of the probe. However, the volumef tissue responding to the probe was very smallndicating that minimum damage resulted from thensertion and presence of the probe. The lack of micro-emorrhages particularly as demonstrated by the lackf hemosiderin, a breakdown product of RBCs, wasnother indication of minimum tissue damage. A thirdissue level response of note was that no increase initosis was observed at any time or at any location in

he tissue including within and nearby the sheath.hus, the cells forming the sheath appear to haveigrated in from the nearby tissue and probably repre-

ent activated astrocytes corresponding to the in-reased number of GFAP-positive cells in the surround-ng tissue.

It is clear that continued studies at shorter timeeriods are critical to our understanding of sheathormation. It is well established in the literature thatstrocytes respond to stab wounds of the brain, but thathe response decreases after 7 days as indicated byhanges in GFAP levels (34). Our results indicate thathere is a continuous GFAP response as long as therobe is in place. Thus, there appears to be a continu-us signal that maintains a chronic GFAP level andstrocyte response. It is also important to follow-up onur observation that microglia are also present in theheath (4, 5). It is obvious that neuroprosthetic devicesan only be successful if we understand how the sheaths formed and can control conditions in the affectedissue volume to minimize, slow, or prevent sheathormation. Another point that is being followed-up ishe shape of the probes. It has been reported that edgesnd sharp protrusions result in more extensive tissueamage (14). A study comparing the probes used in this
ork with others having smaller cross-sections and
moother surfaces with rounded edges will be reportedoon.We have demonstrated the formation and mainte-

ance of a robust cellular sheath surrounding ourilicon probes. The sheath forms before neuronal pro-esses can grow either to or away from, as in the case ofeurons included as part of the implant, the probe. This

s a major impediment to the practical application ofhis technology for restoration of brain function. Sincehe technology is potentially too powerful to be aban-oned, it is imperative that a better understanding ofhe sheath’s composition and formation be obtained.he present study clearly shows that part of thatnderstanding must be gained at short times after

nsertion and that a better understanding of the molecu-ar signaling and interaction with the immune re-ponse, especially microglia, is essential. If micro- andanofabricated devices are to function chronically inhe brain, sheath formation must be minimized byimiting the initial tissue damage and by slowing,imiting, or eliminating sheath formation. The formeran be approached by improved insertion methods andy the use of probes with smaller cross-sections andmoother, more rounded contours. The latter requiresundamental knowledge of the signaling events thatnitiate the brain’s initial reaction to tissue damageaused by implantation and its long-term response tohe probe’s presence. With this knowledge, interventionechniques based on surface biocompatibility and/orhe application of pharmaceuticals can be designed.ithout the use of this sophisticated technology oneill most likely be limited to acute-field measurementsnd/or stimulation as opposed to long-term cellularevel interaction required to replace lost function.

ACKNOWLEDGMENTS

This work was supported by NIH, NCRR RR-10957, and theornell Nanofabrication Facility, a node of the National Nanofabrica-

ion Users Network supported by the NSF. The authors acknowledgehe contributions of Ms. Diane Decker, supervisor of the Wadsworthenter’s histology laboratory, and Dr. W. Samsonoff, Director of theadsworth Center’s Electron Microscopy Core Facility. Mr. Alanershenroder and Mr. William Abbt of the Wadsworth Center’sutomation and Instrumentation department designed and fabri-ated the probe insertion device.

REFERENCES

1. Agnew, W. F., G. H. Yuen, R. H. Pudenz, and L. A. Bullara. 1975.Electrical stimulation of the brain. IV. Ultrastructural studies.Surg. Neurol. 4: 438–448.

2. Agnew, W. F., T. G. H. Yuen, D. B. McCreery, and L. A. Bullara.1986. Histopathologic evaluatin of prolonged intracortical elec-trical stimulation. Exp. Neurol. 92: 162–185.

3. Alesch, F., M. M. Pinter, R. J. Helscher, L. Fertl, A. L. Benabid,and W. T. Koos. 1995. Stimulation of the ventral intermediatethalamic nucleus in tremor dominated Parkinson’s disease and
essential tremor. Acta Neurochir. 136: 75–81.

1

1

1

1

1

1

1

1

1

1

2

2

2

2

2

2

2

2

2

2

3

3

3

3

3

3

3

3

3

3

4

48 TURNER ET AL.

4. Andersen, M. A., D. H. Szarowski, K. Stapleton, C. Castro, R.Davis, W. Shain, H. G. Craighead, M. Isaacson, and J. N. Turner.1997. Brain responses to insertion of nanofabricated siliconprobes. 27th Ann. Mtg. Soc. Neurosci. Abstr. 23(1): 35.11.

5. Andersen, M. D., D. H. Szarowski, S. Martins, H. G. Craighead,M. Isaacson, J. N. Turner, and W. Shain. 1998. Astrocyte andmicroglial responses to insertion of micromachined silicon neu-ral prostheses in immature rats. Submitted for publication.

6. Banwell, J. G., G. H. Creasey, A. M. Aggarwal, and J. T.Mortimer. 1993. Management of the neurogenic bowel in pa-tients with spinal cord injury. Urol. Clin. N. Am. 20: 517–526.

7. Bhadra, N., and J. T. Mortimer. 1997. Extraction forces andtissue changes during explant of CWRU-type intramuscularelectrodes from rat gastrocnemius. Ann. Biomed. Eng. 25:1017–1025.

8. Blond, S., and J. Siegfried. 1991. Thalalmic stimulation for thetreatment of tremor and other movement disorders. Acta.Neurochir. 52: 109–111.

9. Blum, N. A., G. Carkhuff, H. K. Charles, R. L. Edwards, andR. A. Meyer. 1991. Multisite microprobes for neural recording.IEEE Trans. Biomed. Eng. 38: 68–74.

0. Bovolenta, P., F. Wandosell, and M. Nieto-Sampedro. 1992. CNSglial scar tissue: A source of molecules which inhibit centralneurite outgrowth. In Progress in Brain Research (A. C. H. Yu,L. Hertz, M. D. Norenberg, E. Sykova, and S. G. Waxman, Eds.)pp. 376–379. Elsevier, Amsterdam.

1. Caparros-Lefebvre, D., S. Blond, P. Vermersch, N. Pecheux, J. D.Guieu, and H. Petit. 1993. Chronic thalamic stimulation im-proves tremor and levodopa induced dyskinesias in Parkinson’sdisease. J. Neurol. Neurosurg. Psych. 56: 268–273.

2. Carter, R. R., and J. C. Houk. 1993. Multiple single-unitrecordings from the CNS using thin-film arrays. IEEE Trans.Rehab. Eng. 1: 175–184.

3. Drake, K. L., K. D. Wise, J. Farraye, D. J. Anderson, and S. L.Bement. 1988. Performance of planar multisite microprobes inrecording extracellular single-unit intracortical activity. IEEETrans. Biomed. Eng. 35: 719–732.

4. Edell, D. J., V. V. Toi, V. M. McNeil, and L. D. Clark. 1992.Factors influencing the biocompatibility of insertable siliconmicroshafts in cerebral cortex. IEEE Trans. Biomed. Eng. 39:635–643.

5. Eng, L. F., A. C. H. Yu, and Y. L. Lee. 1992. Astrocytic response toinjury. In Progress in Brain Research (A. C. H. Yu, L. Hertz, E.Norenberg, E. Sykova, and S. G. Waxman, Eds.) pp. 353–365.Elsevier, Amsterdam.

6. Grill, W. M., and J. T. Mortimer. 1996. Quantification ofrecruitment properties of multiple contact cuff electrodes. IEEETrans. Rehab. Eng. 4: 49–62.

7. Grill, W. M., and J. T. Mortimer. 1997. Inversion of the current-distance relationship by transient depolarization. IEEE Trans.Biomed. Eng. 44: 1–9.

8. Hambrecht, F. T. 1990. The history of neural stimulation and itsrelevance to future neural prostheses. In Neural Prostheses(W. F. Agnew and D. B. McCreery, Eds.) pp. 2–23. Prentice HallInternational, Englewood Cliffs, NJ.

9. Heetderks, W. J., and F. T. Hambrecht. 1990. Applied neuralcontrol in the 1990’s. Proc. IEEE 76: 1115–1121.

0. Hetke, J. F., L. L. Lund, K. Najafi, D. Wise, and D. J. Anderson.1994. Silicon ribbon cables for chronically implantable microelec-trode arrays. IEEE Trans. Biomed. Eng. 41: 314–321.

1. Hoogerwerf, A. C., and K. D. Wise. 1994. A three-dimensional

microelectrode array for chronic neural recording. IEEE Trans.Biomed. Eng. 41: 1136–1145.

2. Ji, J., and K. D. Wise. 1992. An implantable CMOS circuitinterface for multiplexed microelectrode recording arrays. IEEEJ. Solid-State Circuits 27: 433–443.

3. Kitchener, P. D., J. P. Bourreau, and J. Diamond. 1993. NAKPH-diaphorase histochemistry identifies isolated endothelial cellsat sites of traumatic injury in the adult rat brain. Int. J. Dev.Neurosci. 53: 613–624.

4. Landis, D. M. D. 1994. The early reactions of non-neuronal cellsto brain injury. Annu. Rev. Neurosci. 17: 133–151.

5. Maynard, E. M., C. T. Nordhausen, and R. A. Normann. 1997.The Utah intracortical electrode array: A recording structure forpotential brain–computer interfaces. Electroencephol. Clin. Neu-rophysiol. 102: 228–239.

6. McCreery, D. B., W. F. Agnew, and J. McHardy. 1987. Electricalcharacteristics of chronically implanted platinum–iridium elec-trodes. IEEE Trans. Biomed. Eng. 34: 664–668.

7. McCreery, D. B., T. G. Yuen, W. F. Agnew, and L. A. Bullara.1997. A quantitative computer-assisted morphometric analysisof stimulation-induced injury to myelinated fibers in a periph-eral nerve. J. Neurosci. Methods 73: 159–168.

8. McCreery, D. B., T. G. Yuen, W. F. Agnew, and L. A. Bullara.1997. A characterization of the effects on neuronal excitabilitydue to prolonged microstimulation with chronically implantedmicroelectrodes. IEEE Trans. Biomed. Eng. 44: 931–939.

9. Najafi, K., and J. F. Hetke. 1990. Strength characterization ofsilicon microprobes in neurophysiological tissues. IEEE Trans.Biomed. Eng. 37: 474–481.

0. Najafi, K., J. Ji, and K. D. Wise. 1990. Scaling limitations ofsilicon multichannel recording probes. IEEE Trans. Biomed.Eng. 37: 1–11.

1. Najafi, K., K. Wise, and T. Mochizuki. 1985. A high-yieldIC-compatible multichannel recording array. IEEE Trans. Elec-tron Devices ED-32: 1206–1211.

2. Norenberg, M. D. 1994. Astrocyte responses to CNS injury. J.Neuropathol. Exp. Neurol. 53: 213–220.

3. Norman, R. A., P. K. Campbell, and W. P. Li. 1988. Silicon basedmicrostructures suitable for intracortical electrical stimulation.IEEE J. Solid-State Circuits 714–715.

4. Norton, W. T., D. A. Aquino, I. Hozumi, F. C. Chiu, and C. F.Brosnan. 1992. Quantitative aspects of reactive gliosis: A re-view. Neurochem. Res. 17: 877–885.

5. Pine, J., M. P. Maher, S. Potter, Y. C. Tai, S. Tatic-Lucic, and J. A.Wright. 1996. A cultured neuron probe. Proc. IEEE-EMBS Ann.Mtg.Amsterdam No. 421.

6. Ransohoff, R. M., and M. Tani. 1998. Do chemokines mediateleukocyte recruitment in post-traumatic CNS inflammation?TINS 21: 154–159.

7. Reitboeck, H. J. P. 1983. A 19-channel matrix drive withindividually controllable fiber microelectrodes for neurophysi-ological applications. IEEE Trans. Syst. Man Cybernet. SMC-13: 676–683.

8. Robb, R. A., and C. Barillot. 1988. Interactive 3-D image displayand analysis. Proc. SPIE Hybrid Image Signal Process 939:173–202.

9. Robb, R. A., and C. Barillot. 1989. Interactive display andanalysis of 3-D medial images. IEEE Trans. Med. Imaging 8:217–226.

0. Rousche, P. J., and R. A. Norman. 1992. A method for pneumati-cally inserting an array of penetrating electrodes into cortical
tissue. Ann. Biomed. Eng. 20: 423–437.

4

4

4

4

4

4

4

4

4


1. Schiffer, D., M. T. Giordana, P. Cavalla, M. C. Vigliani, and A.Attanasio. 1993. Immunhistochemistry of glial reaction afterinjury in the rat: Double staining and markers of cell prolifera-tion. Int. J. Dev. Neurosci. 11: 268–280.

2. Schmidt, S., K. Horch, and R. Norman. 1993. Biocompatibility ofsilicon-based electrode arrays implanted in feline cortical tis-sue. J. Biomed. Mat. Res. 27: 1393–1399.

3. Sheehan, D. C., and B. B. Hrapchak. 1980. Theory and Practiceof Histotechnology, 2nd ed. Mosby, St. Louis, MO.

4. Stensaas, S. S., and L. J. Stensaas. 1976. The reaction of thecerebral cortex chronically implanted plastic needles. ActaNeuropathol. (Berl.) 35: 187–203.

5. Tanghe, S. J., and K. D. Wise. 1992. A 16-channel CMOS neuralstimulating array. IEEE J. Solid-State Circuits 27: 1819–1825.

6. Tyler, D. J., and D. M. Durand. 1997. A slowly penetratinginterfascicular nerve electrode for selective activation of periph-eral nerves. IEEE Trans. Rehab. Eng. 5: 51–61.

7. Woodford, B. J., R. R. Carter, D. McCreery, L. A. Bullara, andW. F. Agnew. 1996. Histopathologic and physiologic effects ofchronic implantation of microelectrodes in sacral spinal cord ofthe cat. J. Neuropathol. Exp. Neurol. 55: 982–991.

8. Wright, J. A., S. Tatic-Lucic, Y. C. Tai, M. P. Maher, H. Dvorak,and J. Pine. 1996. Towards a functional MEMS neurowell byphysiological experimentation. Tech. Dig: ASME 1996 Int. Mech.Eng. Cong. Expo. 59: 333–338.

9. Yuen, T. G. H., and W. F. Agnew. 1995. Histological Evaluation ofPolyesterimide-Insulated Gold Wires in Brain. Biomaterials 16:951–956.

cerebral astrocyte response to micromachined silicon implants...however, brain implants must be...

Documents