C

Go

SC1

a

ARRA

KGSMMSBLC

1

wOao2ntsmdhlott

zf

0d

Journal of Neuroscience Methods 206 (2012) 7– 14

Contents lists available at SciVerse ScienceDirect

Journal of Neuroscience Methods

jou rna l h om epa ge: www.elsev ier .com/ locate / jneumeth

linical Neuroscience

ait analysis as a method for assessing neurological outcome in a mouse modelf stroke

usann Hetze1, Christine Römer1, Carena Teufelhart, Andreas Meisel ∗, Odilo Engelenter of Stroke Research Berlin, Neurocure Clinical Research Center, Department for Experimental Neurology, Charité Universitaetsmedizin Berlin, Charitéplatz 1,0117 Berlin, Germany

r t i c l e i n f o

rticle history:eceived 19 October 2011eceived in revised form 2 January 2012ccepted 1 February 2012

eywords:ait analysistroke

a b s t r a c t

Ameliorating stroke induced neurological deficits is one of the most important goals of stroke therapy. Inorder to improve stroke outcome, novel treatment approaches as well as animal stroke models predictivefor the clinical setting are of urgent need. One of the main obstacles in experimental stroke research ismeasuring long-term outcome, in particular in mouse models of stroke. On the other hand, assessingfunctional deficits in animal models of stroke is critical to improve the prediction of preclinical findings.Automated gait analysis provides a sensitive tool to examine locomotion and limb coordination in smallrodents. Comparing mice before and 10 days after experimental stroke (60 min MCAo) we observed a

CAoouse

troke modelehaviorong term functional outcomeatwalk

significant decrease in maximum contact area, stride length and swing speed in the hind limbs, especiallythe contralateral one. Mice showed a disturbed interlimb coordination represented by changes in regu-larity index and phase dispersion. To assess whether gait analysis is applicable to assess improvements byneuroprotective compounds, we applied a model calculation and approached common statistical prob-lems. In conclusion, gait analysis is a promising tool to assess mid- to long-term outcome in experimentalstroke research.

. Introduction

Ischemic stroke is the one of most frequent causes of deathorldwide especially in developed countries (World Healthrganisation (WHO), 2008) and, being the most frequent reason fordult disability (Johnston et al., 2009), puts a huge financial burdenn health care systems (Kolominsky-Rabas et al., 2006; Saka et al.,009). Three months after stroke, half of the survivors have stillot gained full recovery, and 25% of survivors are unable to masterheir daily life without professional care (Ward et al., 2005). Con-equently, improving the life quality of stroke patients is one of theain aims of stroke research. However, treatment options are to

ate very limited and many promising substances in stroke modelsave failed in clinical studies (Endres et al., 2008). Analyzing neuro-

ogical symptoms as gait impairments may improve the evaluationf neurological outcome both in humans and animals. In addition

hey can be suitable for the assessment of the effectiveness of newreatment concepts.

∗ Corresponding author at: Department of Neurology, Charité Universitaetsmedi-in Berlin, Charitéplatz 1, 10117 Berlin, Germany. Tel.: +49 30 450 560 026;ax: +49 30 450 560 915.

E-mail address: andreas.meisel@charite.de (A. Meisel).1 These authors contributed equally.

165-0270/$ – see front matter © 2012 Elsevier B.V. All rights reserved.oi:10.1016/j.jneumeth.2012.02.001

© 2012 Elsevier B.V. All rights reserved.

In humans, impairment of balance control after stroke is themain cause for gait deficits, preventing normal performance ofmovements that require complex coordination (e.g. pelvic and kneecontrol in stance phase, swing phase hip, knee and ankle flexion andextension) (Olney et al., 1989; Olney and Richards, 1995). Inabil-ity to perform normal movements after stroke results primarilyfrom an impaired swing initiation (Chen et al., 2005), which canbe explained by an inadequate leg propulsion and compensatorymechanisms (Nadeau et al., 1999; Peterson et al., 2010).

In contrast to the availability of proper tests in patients andample of tests available for evaluating motor as well as corti-cal impairment, animal behavior after cerebral ischemia remainsto be difficult to assess in particular in mouse models of stroke.Although tests for locomotion like rope walk (Carlini et al., 1967),grid walk (Kunkel-Bagden et al., 1993), ladder rung test (Metz andWhishaw, 2002) and foot print analysis (de Medinaceli et al., 1982)unravel distinct neurological impairments after stroke, their sensi-tivity is too low to distinguish subtle motor impairments. Contraryto patients, laboratory rodents, especially mice, show a quick func-tional recovery, which may be explained by their ecological nicheas prey animals, implying a need of compensatory mechanismsfor hiding sickness symptoms as a key for survival. (Tizard, 2008)

Assessing gait parameters with high speed cameras could unravelthese “hidden changes”, using as walkway either a glass plate forthe individual’s comfortable speed (Hamers et al., 2001) or a tread-mill to obtain current walking speed (Hampton et al., 2004).

8 roscie

so2wavmcptd

ttieasts

2

2

GeEcAedau

rfssGfr

2

(elantgtcWuB2

2

c

S. Hetze et al. / Journal of Neu

Several groups have investigated gait impairments in differenttroke models. Nevertheless, majority of this work has been carriedut in rats. In studies (Encarnacion et al., 2011; Vandeputte et al.,010; Wang et al., 2008) with rats similar altered gait parametersere found indicating that gait analysis is a feasible method for

nimal models of stroke. Contrary to the situation in the rat, onlyery little is known about gait changes after experimental stroke inice. This is a great disadvantage, since, to date, mice are the most

ommonly used animals in stroke research, despite the difficulty oferforming behavioral tests in them. Lubjuhn and coworkers foundwo parameters changed in a permanent distal MCAo model, buto not report on inter-limb coordination (Lubjuhn et al., 2009).

In this study we aim at describing gait changes in mice afterransient filament MCAo, modeling a severe infarction in the mediaerritory. We put special emphasis on long-term outcome, since thiss of urgent need for bench to bedside translation. To our knowl-dge, until today there are no studies that have used gait analysiss an outcome parameter for neuroprotection in mouse models oftroke. Hence our second goal is to establish gait analysis as validool for measuring mid- to long-term outcome after experimentaltroke in mice.

. Materials and methods

.1. Animals and housing

Male C57Bl6/J mice (Charles River Laboratories, Sulzfeld,ermany) were housed in cages lined with chip bedding andnvironmental enrichment (mouse tunnel and igloo; Plexx B.V.,lst, The Netherlands) on a 12 h light/dark cycle (change 7:00 o’lock) with ad libitum access to water and food (standard chow).t the time of the experiment, mice were 11–14 weeks old. Allxperiments were performed in accordance with the Europeanirective on the protection of animals used for scientific purposesnd approved by the relevant authority, Landesamt für Gesundheitnd Soziales, Berlin, Germany.

To compare consistence, but also due to protection of animalseasons, we included the control animals of different experimentsulfilling the following criteria 0.2 ml Saline immediately aftertroke (Placebo treatment), 60 min Middle cerebral artery occlu-ion with Bederson score of 2 after operation (see below) andait analysis before and 10 days after stroke. In total, 31 mice

rom 5 experiments were included. Inter experiment comparisonsevealed no significant or relevant difference in all gait parameters.

.2. Stroke model

The surgical procedure of Middle cerebral artery occlusionMCAo) was performed as described elsewhere in detail (Engelt al., 2011) according to the standard operating procedures of ourab (Dirnagl and Members of the MCAO-SOP Group, 2010). In brief,fter a ventral cervical midline incision a 12 mm silicon-coatedylon filament was introduced over the common carotid artery andhe internal carotid artery into the circle of Willis blocking the ori-in of the middle cerebral artery. After an ischemia time of 60 minhe filament was removed for reperfusion. Body temperature wasontrolled throughout the whole procedure and Isoflurane (Abott,

iesbaden, Germany) in a 1:2 mixture oxygen/nitrous oxide wassed for anaesthesia. We verified success of operation by using theederson score (Dirnagl and Members of the MCAO-SOP Group,010).

.3. Gait analysis

Gait analysis in mice was performed with an automatedomputer assisted method (CatWalkTM, Noldus Information

nce Methods 206 (2012) 7– 14

Technology, Wageningen, The Netherlands) according to manufac-tures instructions and published procedures (Hamers et al., 2001).The equipment was located in a dark and silent room (<20 lux ofillumination). In brief the system consists of an elevated 1.3 m longglass plate which is illuminated with a fluorescent light comingfrom the side and is internally reflected in the glass. When miceare passing the walkway and their paws contacting the glass plate,light leaves the glass plate due to the changed refractive index andis reflected downwards. A high-speed camera underneath the glassplate captures the images which are subsequently analyzed by theconnected computer program. Home cage was used as bait at theend of the walkway, and animals were trained in 3 sessions beforefirst measurement. We acquired a minimum of 3 compliant runs,which had to fulfill minimum run duration of 0.5 s, maximum runduration of 5 s, and a maximum speed variation of 60%. Runs, whereanimals turned or walked backwards, were excluded and the ani-mal got another try. Runs in which the software was unable tocalculate phase dispersion (Fig. 1) due to, for instance, anchor pawbeing undetected, were excluded from statistical analysis.

Gait parameters can be classified into three main categories(Hamers et al., 2006): Parameters related to (a) individual paws,like width and length of a paw print, duration of paw contact, thepressure caused by a paw, etc. (b) the position of footprints, forinstance relative print position, stride length and base of support(Fig. 1A) and (c) time dependent relationship between footprintssuch as phase dispersion and support formula (Fig. 1B and C). Thedefinitions for gait parameters we focused at are summarized inTable 1.

2.4. Magnetic resonance imaging

One day after MCAO surgery MRI was performed in a sub-group of animals on a Bruker 7 T PharmaScan® 70/16 with aBruker 98/38 mmRF Coil, operating on Paravision software plat-form (Bruker, Karlsruhe, Germany). Mice were anesthetized with1.5% isoflurane in an oxygen/nitrous oxide mixture, and bodytemperature was monitored with an MR-compatible physiologymonitoring unit and maintained within physiological limits usinga heated water jacket. They were then fixed using a stereotac-tic frame and positioned in the magnet bore. T2 weighted MRIwas achieved with a TurboRARE sequence (imaging parameters:256 × 256 in plane resolution, 20 slices with a thickness of 500 �m,FOV 28.5 mm, TR 3500 ms, TE 56 ms, acquisition time 6 min). Theaxial slices were chosen to cover the region between the olfac-tory bulb and the cerebellum. The size of the lesion apparent inthe T2 weighted imaging was determined semi-automatically usingAnalyze 5.0 software (AnalyzeDirect, Overland Park, KS, USA).

2.5. Statistical analysis

Data were analyzed with SPSS 14.0 for Windows (SPSS Inc.,Chicago, IL, USA). Normal distribution of variables was verified withKolmogorov–Smirnov test. A type I error (˛) of 0.05 and type II error(ˇ) of 0.2 was accepted. Parametric student’s t-test was applied tocalculate p values. Correlations between body weight and maxi-mum contact area were assessed using Spearman’s rho correlationcoefficient. P-value of less than 0.05 was considered statisticallysignificant (*p < 0.05; **p < 0.01; ***p < 0.001). Data are expressed asmean ± standard deviation (SD) and represented as Box plots withwhiskers Minimum to Maximum.

For calculating effect sizes and a priori sample sizes the pro-

gram G*Power 3.1.2 (Faul et al., 2007) was applied. Additionallywe used the comparison for effect sizes described by Matthewsand Altman (1996a) and standard procedures such as the standarderror of differences (Bland, 2000).

S. Hetze et al. / Journal of Neuroscience Methods 206 (2012) 7– 14 9

F Phase

p dispers missio

3

agMmSte

3

btt

ig. 1. Graphical representation of selected gait parameters (A) and calculation of

air (RF–RH) in relation to the step cycle time of the anchor limb (RF) (B) If Phase

upport; PD = phase dispersion; B and C modified from Kloos et al. (2005) (with per

. Results

Table 1 provides an overview of the investigated gait parametersnd their changes from before to ten days after MCAo. In a sub-roup of animals we measured brain lesions using MRI 1 day afterCAO surgery. Stroke affected both, striatum and cortex with aean infarct volume of 18.1% of the hemisphere (edema corrected,

D 4.9, n = 10; Fig. 2). Run duration did not differ relevantly betweenhe groups, despite animals walked slightly slower 10 days afterxperimental stroke (Fig. 3A).

.1. Body weight changes cannot explain gait changes after stroke

Mice subjected to MCAo had a body weight of 23.37 ± 1.53 gefore and 22.31 ± 1.64 g 10 days after surgery (Student’s t-test,(60) = 2.620, p = 0.011). Body weights did not correlate with any ofhe measured Catwalk gait parameters (data not shown).

dispersions (B,C). Phase dispersion is the timing between initial contacts of a limbsion exceeds 75%, the target paw is assigned to the next anchor (C) (BOS = base ofn of Elsevier Ltd., Oxford, UK).

3.2. Spatial and kinetic characteristics of individual paws

Maximal contact area of hind paws and right front paw decreasedin ischemic mice. Maximal contact areas of front (Student’s t-test, t(60) = 2.092, p = 0.041) and hind paws (Student’s t-test,t(60) = 5.710, p < 0.0005) on affected right side were decreased 10days after left-sided MCAo compared to baseline (before MCAo).Also the maximal contact area of the (left) hind paw, ipsilateralto the ischemic lesion, was decreased following MCAo (Student’st-test, t(60) = 4.084, p = 0.001) (Fig. 3B).

Stride length of hind paws decreases after MCAo. The dis-tance between consecutive steps with right (Student’s t-test,t(60) = 2.885, p = 0.005) as well as left hind paw (Student’s t-test,

t(60) = 2.883, p = 0.005) was shortened in mice 10 days after MCAocompared to the healthy condition of the same group of mice beforeMCAo (Fig. 4A). In contrast, the stride length of neither of the frontpaws was affected by cerebral ischemia.

10 S. Hetze et al. / Journal of Neuroscience Methods 206 (2012) 7– 14

Table 1Overview of stroke sensitive Gait parameters.

Parameter Definition Paw Mean ± SD before MCAo Mean ± SD after MCAo P d

(a) Spatial paw statisticsMax. contact Area of an individual RF * 25.00 ± 4.248 22.75 ± 4.231 0.041 0.54Area (mm2) paw print at moment of RH *** 23.43 ± 3.710 16.98 ± 5.086 <0.0005 1.46

its maximal contact LF 24.95 ± 4.055 23.53 ± 4.833 0.214 0.32with the ground LH *** 23.42 ± 5.462 18.41 ± 5.746 0.001 0.89

Max. intensity Intensity of maximal RF 114.7 ± 9.990 113.2 ± 10.39 0.554 0.15(%) contact of a paw with RH *** 142.9 ± 12.81 125.6 ± 22.68 <0.0005 0.98

the ground LF 115.2 ± 9.011 115.4 ± 11.81 0.942 0.01LH ** 141.0 ± 11.46 127.7 ± 23.61 0.007 0.70

(b) Kinetic paw statisticsRun duration Time for passing the * 2.439 ± 0.663 2.793 ± 0.553 0.026 0.62(s) walkway

Normalized Swing speed (speed of RF 1335 ± 171.4 1323 ± 198.0 0.798 0.06swing speed movement) RH *** 1311 ± 180.2 1102 ± 193.9 <0.0005 1.12(s) * run duration LF 1316 ± 157.3 1294 ± 112.4 0.532 0.16

LH ** 1368 ± 309.2 1180 ± 175.7 0.005 0.78

Stand (s) Time of one paw RF *** 0.132 ± 0.030 0.163 ± 0.033 <0.0005 0.94contact with the ground RH 0.133 ± 0.030 0.133 ± 0.035 0.996 0.01

LF *** 0.133 ± 0.032 0.161 ± 0.033 0.001 0.87LH 0.137 ± 0.034 0.145 ± 0.037 0.346 0.35

(c) Comparative paw statisticsStride length Distance between RF 57.58 ± 8.163 56.08 ± 7.541 0.456 0.02(mm) successive steps of the RH ** 54.92 ± 9.718 48.07 ± 8.973 0.005 0.78

same paw (Fig. 1) LF 57.42 ± 8.152 56.06 ± 7.588 0.500 0.13LH ** 55.08 ± 8.772 49.02 ± 7.744 0.005 0.72

Base of BOS in Fig. 1A RF-LF * 11.93 ± 7.297 8.226 ± 6.222 0.036 0.51support (mm) RH-LH * 22.07 ± 14.14 14.55 ± 10.52 0.021 0.62

Print positions Distance position of RF-RH 10.15 ± 4.431 12.45 ± 6.331 0.102 0.49hind paw to previous LF-LH 9.569 ± 5.045 10.73 ± 6.039 0.414 0.22position of ipsilateralfront paw

Duty cycle Represents relative RF * 53.34 ± 3.764 55.58 ± 3.380 0.016 0.72(%) stand phase as part of RH ** 53.66 ± 5.098 49.68 ± 5.527 0.005 0.79

step cycle LF * 53.00 ± 3.256 55.16 ± 3.773 0.019 0.56LH 55.25 ± 6.049 53.17 ± 5.074 0.147 0.41

(d) Coordination statisticsPhase Contact Target paw in RF > LH *** 1.606 ± 5.189 7.866 ± 6.321 <0.0005 1.12dispersion relation to step cycle of LF > RH *** 2.407 ± 5.371 10.32 ± 5.758 <0.0005 1.45(%) anchor paw LH > RH ** 43.05 ± 7.401 37.02 ± 7.581 0.002 0.82

(Fig. 1) LF > RF 48.71 ± 3.602 48.62 ± 3.895 0.931 0.01RF > RH 45.75 ± 4.199 44.80 ± 5.978 0.470 0.25LF > LH 45.96 ± 5.299 44.16 ± 5.601 0.199 0.31

Regularity Average of normal step ** 93.12 ± 5.965 87.40 ± 9.306 0.006 0.73index (%) sequence

P values smaller than 0.05 are set in bold.RF: right fore limb; RH: right hind limb; LF: left fore limb; LH: left hind limb; d = Cohen’s effect size.

*

mcddwsMtt

sppw

P < 0.05.** P < 0.01.

*** P < 0.001.

Normalized swing speed of hind paws is decreased in ischemicice. Swing speed describes the speed of a paw while it has no

ontact with a glass plate. In order to overcome a possible biaserived from animals’ individual run speed, we calculated theistance in mm characterizing how much an animal moves for-ard with a particular paw over the time of the whole run (swing

peed × run duration). Compared with the healthy mice (beforeCAo), ischemia reduces the swing distances with left (Student’s

-test, t(60) = 2.943, p = 0.005) and right hind paws (Student’s t-test,(60) = 4.420, p < 0.0005) (Fig. 4B).

Taken together, our data from gait analysis 10 days after left-

ided MCAo demonstrate that ischemia strongly affects both hindaws, with contralateral right more pronounced than the left hindaws. Paw print area, stride length and normalized swing speedere decreased by cerebral lesion.

3.3. Changes in gait coordination

Ischemia induces irregularities in step sequence patterns. Inter-paw coordination during gait is based on normal step sequencepatterns of an animal. The overall inter-paw coordination duringgait (gait regularity index of step sequence) is reduced 10 days afterstroke onset (Student’s t-test, t(60) = 2.879, p = 0.006) (Fig. 5C).

Inter-paw coordination during gait is impaired after MCAo. Avaluable parameter in assessing inter-paw coordination is alsophase dispersion which characterizes the placement of two paws(“target” and “anchor”) during the cycle of consecutive initial

contacts with an “anchor” paw (Kloos et al., 2005). The lowergroup size results from the exclusion criteria of non-detectedpaws as described above. Investigating diagonal limb pairs (LFto RH and RF to LH) 10 days after MCAo the time of hind paw

S. Hetze et al. / Journal of Neuroscie

F

cpbtLri(if

ieL

3

TtXem

Fc

ig. 2. Characteristic T2-weighted MRI of a mouse brain 1 day after a 60 min MCAo.

ontact succeeding respective front paw contact with a glasslate is prolonged compared to the healthy condition of the miceefore MCAo (Student’s t-test, t(54) = −5.311, p < 0.0005 for LFo RH; and Student’s t-test, t(56) = −4.141, p < 0.0005 for RF toH; Fig. 5A). Ischemia impairs the coordination between left andight hind paws, which was demonstrated by a shortened timenterval between right and left hind paw contact with a glass plateStudent’s t-test, t(60) = 3.170, p < 0.002). Inter-paw coordinationn ipsilateral pairs, i.e. LF to LH and RF to RH, as well as betweenront paws was not altered by MCAo (Fig. 5B).

In line with the alterations in individual paws, we observedrregularities in normal step sequence patterns as well as in gen-ral inter-paw coordination in diagonal pairs (LF to RH and RF toH) after MCAo.

.4. Analysis of applicability

Taken together, gait is altered after experimental stroke in mice.o test the applicability of gait analysis we applied a model calcula-

ion assuming an arbitrary neuroprotective agent called “substance”. On the one hand an improvement by 100%, so complete recov-ry, is rather an unachievable goal, on the other hand only ainor improvement may be not relevant for a clinical setting.

ig. 3. (A) Run duration is not relevantly altered 10 days following stroke, despite animals wontact area is significantly decreased in both hind paws and slightly in right front paw (S

nce Methods 206 (2012) 7– 14 11

Thus, we define that substance X should improve gait deficits atleast by 50%, which is in our experience a feasible and realisticvalue.

Here we use phase dispersions LF-RH as an example for exam-ining different statistical methods commonly used in analyzingbehavioral data. As described above, this parameter changes from2.4% (±5.4) before to 10.3% (±5.7) after experimental stroke, result-ing in a calculated mean phase dispersion of 6.5% (±5.5) amonganimals treated with substance X. The effect size (Cohen, 1988) forthree groups is f = 0.6, per definition a strong effect. Computing therequired sample size with the preconditions ̨ = 0.05 and ̌ = 0.8(Power) one gets a critical F of 3.24 and a required total sample sizeof 45, hence 15 per group. This is a commonly used group size forstroke studies.

However, this is only valid for the comparison between the threegroups. Usually neuroprotection studies aim to demonstrate a pro-tective effect for verum (e.g. substance X) compared to placebo (e.g.vehicle) treatment. Whereas the effect size comparing baseline andplacebo is relative large (d = 1.5), the effect size comparing placeboand substance X accounts with d = 0.75 only for a medium size effect.Hence one would need only 8 animals per group to show a sig-nificant difference between baseline and placebo ( ̨ = 0.05; ̌ = 0.8;two-tailed), but 30 per group to show a significant difference inStudent’s t-test between the both treatments.

To deal with the often smaller groups of animals, one may betempted to compare both treatments against baseline to show thatonly placebo treatment results in significant changes in gait. Thismay be misleading, since the p-value is dependent on the effect sizeand also on the precision of the effect’s estimation, so its standarderror (Matthews and Altman, 1996b). Hence without providing allinformation including effect size, sample size per group and distri-bution, this might result in false positive conclusions.

In our model calculation, we “observed” a large differencein effect size between treatment with substance X and placebo.Consequently we would need a statistical test that assumes asnull hypothesis that the true effect sizes are the same in bothgroups.

Matthews and Altman described a suitable procedure based onthe differences between the treatments and the standard error ofthe difference (Matthews and Altman, 1996a). Applying this, wedetect a significant difference between placebo and substance X(Table 2). To calculate the group sizes needed, first we take theformula for the test statistic

Z = x̄1 − x̄2

se21 + se2

2

(3.1)

alked slightly slower 10 days after experimental stroke (grey boxes). (B) Maximumtudent’s t-test; *p < 0.05; ***p < 0.001).

12 S. Hetze et al. / Journal of Neuroscience Methods 206 (2012) 7– 14

Table 2Example calculation for Phase dispersion LF-RH to compare effect sizes. Substance X is an arbitrarily chosen drug improving outcome by 50% (two sided test).

Group Placebo Substance X

Stroke Before After Before After

Mean 2.4 10.3 2.4 6.35SE 0.96 1.03 0.96 1.03N 31 31 31 31Effect 10.3–2.4 = 7.9 6.35–2.4 = 3.95SEdifference 1.41 1.41Difference between Placebo’s and Substance X’s differences 3.95Z 1.98 →P < 0.05

Fig. 4. Stride length (A) and normalized swing speed (swing speed × run duration; B)before (white boxes) and ten days after (grey boxes) experimental stroke. Both hindpaws exhibit deficits in the swing subtask (Student’s t-test; **p < 0.01; ***p < 0.001). Fig. 5. Inter-limb coordination is disturbed 10 days after MCAo (A, B) Phase disper-

sions are altered for the diagonal pairs and the coordination between the hind paws.(C) A lower step sequence regularity index reflects more irregular step patterns.

roscie

wtt

s

faa

n

Fsfr1iol(

4

tiiwiops(ma(pacdti

ibpsf

uatpobnidadppwc

S. Hetze et al. / Journal of Neu

here x1 and x2 represent the differences before and after stroke ofhe respective group and sei is the corresponding standard error ofhe difference. Replaced the x values by the differences and inserted

ediff =√

s21

n1+ s2

2n2

(3.2)

or the appropriate standard errors of the difference with s2 as vari-nce and n as group size, we can solve the equitation for n, if wessume that n is equal in all groups (n1 = n2 = n):

=s2

before+ s2

Placebo+ s2

before+ s2

X

((xbefore − xPlacebo) − (xbefore − xX )/z)2(3.3)

or significant values, per definition z has to be greater than 1.96 ormaller than −1.96 for two sided tests, other values can be insertedrom our experiments and the assumptions described above. As aesult we get a required sample size of 21 for a 50% improvement,6 for a 60% improvement and 12 for a 70% improvement. If only

mprovement is of interest, e.g. in testing neuroprotective drugs,ne could use this test as one-sided test (z would have then to bearger 1.65 or smaller −1.65), hence further reducing sample sizeBland and Altman, 1994).

. Discussion

Gait is a coordinated action of different subtasks. Hypothesizinghat severe cerebral ischemia in mice disturbs the well coordinatednterplay between the 4 limbs, we observed significant differencesn the swing speed and stride length of the hind paws. Hence

e conclude that the swing subtask in hind limbs is impairedn mid- to long-term outcome after stroke. A decreased intensityf foot prints in both hind paws may be an indicator of reducedropulsion by hind paws. In comparison, hemiparetic patients aftertroke show deficits in swing initiation and forward propulsionHall et al., 2011). In addition, we observed a decrease of maxi-

um contact area in both hind paws after experimental stroke,lso suggesting a less effective use in weight bearing and propulsionVrinten and Hamers, 2003). The strong effects of stroke in the hindaws, especially in the contralateral one, may reflect the neuro-natomical conditions. In addition, the impairment of one subtaskould explain partially the observed disturbance in inter-limb coor-ination. However, since we also see a worse coordination betweenhe hind limbs, we assume that also the overall coordination of gaits impaired.

In our study, post-stroke mice have an increase in stand timen both front paws. The duty cycle, meaning the fraction of weightearing in the whole step cycle, was slightly increased in both frontaws and decreased in hind paws. This may represent a compen-atory mechanism in which the front paws support the impairedunction of the hind paws.

In terms of interlimb coordination, we observed changes in reg-larity index, as well as in phase dispersions in the diagonal pairsnd in the back girdle pair. It is of great importance to consider,hat the change from walk to trot affects the lateral and diagonalhase dispersions, but not the girdle pairs (Hamers et al., 2006). Inur study, mice passed the walkway trotting, which was favoredy setting a short time limit for valid runs. Additionally, we didot observe the typical signs of walking in phase dispersion. Dur-

ng walk, diagonal pairs’ phase dispersion is around 75%, where weetected a change from 0 to 2% (typical for trot) to about 7–10%fter stroke. We did not observe the typical change in lateral phaseispersion switching from walk to trot, but untypically in girdle

airs’ phase dispersion. The not well timed placement of diagonalairs may affect the stability of the diagonal support phase in trot,hich may account for more irregular step patterns. Thus, interlimb

oordination is impaired in the MCAo model.

nce Methods 206 (2012) 7– 14 13

In line with data from ischemic rats, we observed significantchanges in the maximum contact area and inter-limb coordina-tion. Wang et al. reported in a rat model of distal Middle cerebralartery occlusion (MCAo) a decreased paw print intensity, maximalarea of paw contact and a disturbed limb coupling both 4 daysand 5 weeks after stroke. Interestingly, a treatment with excessiveenriched environment improved these deficits significantly (Wanget al., 2008). In a rat photothrombotic model, stroke decreasedintensities and contact areas of the contralateral hind paw. How-ever, these changes were observed at a very early time point of oneday after experimental stroke (Vandeputte et al., 2010). Similar toour observations, swing speed in the contralateral hind paw wasdiminished over 12 weeks in rat model of severe stroke. In addi-tion, for up to two weeks the duty cycle and paw print intensitywere decreased as well as the disturbed inter-limb coordinationwas disturbed (Encarnacion et al., 2011).

In the only published investigation in mice, Lubjuhn and col-leagues described several functional tests after permanent distalMCAo, among them three different systems to measure gait. In con-trast to our study differences in stance and brake duration two daysafter surgery were observed using automated systems with andwithout treadmill. In contrast, differences in stride length were notobserved (Lubjuhn et al., 2009). Since the model of permanent distalMCAo results in cortical lesions and ours affects cortex and stria-tum these differences might be explained by the different strokemodels. Furthermore, they do not report on the inter-limb coordi-nation or any other temporal parameters that could reveal deficitsin walking subtasks.

An important limitation of the Catwalk system is the depend-ency on walking speed, as gait velocity is an important influencefactor on many gait parameters, and consequently we either cor-rected for this or used parameters independent from walking speed.Although infarct volume does not necessarily correlate with func-tional impairment (Encarnacion et al., 2011; Jones et al., 2009), itis a limitation of our study that we did not correlate functionalimpairment and infarct volume due to our experimental setup.Furthermore, this study is only valid for the model and time pointwe have investigated. From the work of other groups (Encarnacionet al., 2011) we assume that differences are more pronounced atearlier time points. In human stroke research 3 month after strokeonset are accepted as time point for assessing long-term outcome.In contrast, generally accepted time points for measuring long-termoutcome in stroke models of mice have not been defined. Sincethe acute phase is completed within the first week after stroke(Dirnagl et al., 2003), measurement ten days after stroke reflectsmid- to long-term outcome, considering the short recovery time ofmice compared to humans after stroke. Longer follow up intervalsare needed to investigate whether gait analysis is appropriate forlong-term outcome in murine stroke research.

Analyzing the applicability of gait analysis in testing novel stroketreatments, there are only a few parameters which show an effectlarge enough to detect gait improvements. But even then, onewould need 15–20 animals per group to detect neuroprotective orregenerative effects at this late time point. This may be hamperedby the high mortality, caused by neurological reasons and medicalcomplications like infections (Meisel and Meisel, 2011). Especiallybetween day 3 and day 6 animals are at a high risk to develop pneu-monia with a high attributable mortality (Engel and Meisel, 2010),which may also influence the individuals motivation to walk. Takentogether gait analysis is a valuable tool to detect changes, but it isonly one piece of the puzzle in measuring neurological outcomeafter experimental stroke.

In conclusion, our data suggest that gait is similarly altered inmice and human stroke. Hence gait analysis – together with othertests – may be a good predictor for the effectiveness of potentialtherapeutic compounds in bench to bedside translation. However

1 roscie

irsi

C

A

imftE(mg

R

BB

C

C

C

d

D

D

E

E

E

E

F

H

Ward A, Payne KA, Caro JJ, Heuschmann PU, Kolominsky-Rabas PL. Care needs andeconomic consequences after acute ischemic stroke: the Erlangen Stroke Project.

4 S. Hetze et al. / Journal of Neu

t remains a challenging task to measure functional deficits in labo-atory rodents after experimental stroke. For further research, veryensitive measures for skilled walk may be a promising way toncrease the effect size and the sensitivity comparing treatments.

onflict of Interest

No conflict of interest to declare.

cknowledgements

The authors would like to thank Solveig Flaig for the mathemat-cal advice, Sonja Hochmeister for providing data from one experi-

ent and Sabine Kolodziej, Mareike Thielke and Verena Wörtmannor the excellent technical assistance. This work was supported byhe German Research Foundation (Exc 257), the Federal Ministry ofducation and Research (01 EO 08 01), the Helmholtz AssociationSO-022NG) and has received funding from the European Com-

unity’s Seventh Framework Programme (FP7/2007-2013) underrant agreement no 201024 (all given to AM).

eferences

land JM, Altman DG. One and two sided tests of significance. BMJ 1994;309:248.land M. An introduction to medical statistics. 3rd ed Oxford: Oxford University

Press; 2000.arlini EA, Teresa M, Silva A, Cesare LC, Endo RM. Effects of chronic administration of

beta-(3,4-dimethoxyphenyl)-ethylamine and beta-(3,4,5-trimethoxyphenyl)-ethylamine on the climbing rope performance of rats. Med Pharmacol Exp Int JExp Med 1967;17:534–42.

hen G, Patten C, Kothari DH, Zajac FE. Gait differences between individuals withpost-stroke hemiparesis and non-disabled controls at matched speeds. Gait Pos-ture 2005;22:51–6.

ohen J, editor. Statistical power analysis for the behavioral sciences. 2nd ed Hills-dale, NJ: Erlbaum; 1988.

e Medinaceli L, Freed WJ, Wyatt RJ. An index of the functional condition of ratsciatic nerve based on measurements made from walking tracks. Exp Neurol1982;77:634–43.

irnagl U, Members of the MCAO-SOP Group. Standard operating pro-cedures (SOP) in experimental stroke research: SOP for middle cere-bral artery occlusion in the mouse. Nature Precedings. Available onlinehttp://precedings.nature.com/documents/3492/version/2, 2010.

irnagl U, Simon RP, Hallenbeck JM. Ischemic tolerance and endogenous neuropro-tection. Trends Neurosci 2003;26:248–54.

ncarnacion A, Horie N, Keren-Gill H, Bliss TM, Steinberg GK, Shamloo M. Long-termbehavioral assessment of function in an experimental model for ischemic stroke.J Neurosci Methods 2011.

ndres M, Engelhardt B, Koistinaho J, Lindvall O, Meairs S, Mohr JP, et al. Improvingoutcome after stroke: overcoming the translational roadblock. Cerebrovasc Dis2008;25:268–78.

ngel O, Kolodziej S, Dirnagl U, Prinz V. Modeling stroke in mice – middle cerebralartery occlusion with the filament model. J Vis Exp 2011:e2423.

ngel O, Meisel A. Models of infection before and after stroke: investigating newtargets. Infect Disord Drug Targets 2010:9.

aul F, Erdfelder E, Lang AG, Buchner A. G*Power 3: a flexible statistical power

analysis program for the social, behavioral, and biomedical sciences. Behav ResMethods 2007;39:175–91.

all AL, Peterson CL, Kautz SA, Neptune RR. Relationships between muscle con-tributions to walking subtasks and functional walking status in persons withpost-stroke hemiparesis. Clin Biomech (Bristol, Avon) 2011.

nce Methods 206 (2012) 7– 14

Hamers FP, Koopmans GC, Joosten EA. CatWalk-assisted gait analysis in the assess-ment of spinal cord injury. J Neurotrauma 2006;23:537–48.

Hamers FP, Lankhorst AJ, van Laar TJ, Veldhuis WB, Gispen WH. Automated quan-titative gait analysis during overground locomotion in the rat: its applicationto spinal cord contusion and transection injuries. J Neurotrauma 2001;18:187–201.

Hampton TG, Stasko MR, Kale A, Amende I, Costa AC. Gait dynamics in tri-somic mice: quantitative neurological traits of Down syndrome. Physiol Behav2004;82:381–9.

Johnston SC, Mendis S, Mathers CD. Global variation in stroke burden and mor-tality: estimates from monitoring, surveillance, and modelling. Lancet Neurol2009;8:345–54.

Jones TA, Allred RP, Adkins DL, Hsu JE, O’Bryant A, Maldonado MA. Remodeling thebrain with behavioral experience after stroke. Stroke 2009;40:S136–8.

Kloos AD, Fisher LC, Detloff MR, Hassenzahl DL, Basso DM. Stepwise motor and all-or-none sensory recovery is associated with nonlinear sparing after incrementalspinal cord injury in rats. Exp Neurol 2005;191:251–65.

Kolominsky-Rabas PL, Heuschmann PU, Marschall D, Emmert M, Baltzer N, Neun-dorfer B, et al. Lifetime cost of ischemic stroke in Germany: results and nationalprojections from a population-based stroke registry: the Erlangen Stroke Project.Stroke 2006;37:1179–83.

Kunkel-Bagden E, Dai HN, Bregman BS. Methods to assess the development andrecovery of locomotor function after spinal cord injury in rats. Exp Neurol1993;119:153–64.

Lubjuhn J, Gastens A, von Wilpert G, Bargiotas P, Herrmann O, Murikinati S, et al.Functional testing in a mouse stroke model induced by occlusion of the distalmiddle cerebral artery. J Neurosci Methods 2009;184:95–103.

Matthews JN, Altman DG. Interaction 3: How to examine heterogeneity. BMJ1996a;313:862.

Matthews JN, Altman DG. Statistics notes. Interaction 2: Compare effect sizes not Pvalues. BMJ 1996b;313:808.

Meisel C, Meisel A. Suppressing immunosuppression after stroke. New Engl J Med2011;365:2134–6.

Metz GA, Whishaw IQ. Cortical and subcortical lesions impair skilled walking in theladder rung walking test: a new task to evaluate fore- and hindlimb stepping,placing, and co-ordination. J Neurosci Methods 2002;115:169–79.

Nadeau S, Gravel D, Arsenault AB, Bourbonnais D. Plantarflexor weakness as a lim-iting factor of gait speed in stroke subjects and the compensating role of hipflexors. Clin Biomech (Bristol, Avon) 1999;14:125–35.

Olney SJ, Colborne GR, Martin CS. Joint angle feedback and biomechanical gait anal-ysis in stroke patients: a case report. Phys Ther 1989;69:863–70.

Olney SJ, Richards C. Hemiparetic gait following stroke. Part I: Characteristics. GaitPosture 1995;4:136–48.

Peterson CL, Hall AL, Kautz SA, Neptune RR. Pre-swing deficits in forward propulsion,swing initiation and power generation by individual muscles during hemipareticwalking. J Biomech 2010;43:2348–55.

Saka O, McGuire A, Wolfe C. Cost of stroke in the United Kingdom. Age Ageing2009;38:27–32.

Tizard I. Sickness behavior, its mechanisms and significance. Anim Health Res Rev2008;9:87–99.

Vandeputte C, Taymans JM, Casteels C, Coun F, Ni Y, Van Laere K, et al. Automatedquantitative gait analysis in animal models of movement disorders. BMC Neu-rosci 2010;11:92.

Vrinten DH, Hamers FF. ‘CatWalk’ automated quantitative gait analysis as a novelmethod to assess mechanical allodynia in the rat; a comparison with von Freytesting. Pain 2003;102:203–9.

Wang Y, Bontempi B, Hong SM, Mehta K, Weinstein PR, Abrams GM, et al. Acomprehensive analysis of gait impairment after experimental stroke and thetherapeutic effect of environmental enrichment in rats. J Cereb Blood FlowMetab 2008;28:1936–50.

Eur J Neurol 2005;12:264–7.World Health Organisation (WHO), editor. The global burden of disease – update

2004. Geneva: WHO Library Cataloguing-in-Publication Data; 2008.