ARTICLE

DOI: 10.1038/s41467-018-05727-y OPEN

Implementation and benchmarking of a novel

analytical framework to clinically evaluate
tumor-specific fluorescent tracers
Marjory Koller1, Si-Qi Qiu1,2, Matthijs D. Linssen3,4, Liesbeth Jansen1, Wendy Kelder5, Jakob de Vries 1,
Inge Kruithof6, Guo-Jun Zhang 7, Dominic J. Robinson8, Wouter B. Nagengast3, Annelies Jorritsma-Smit4,
Bert van der Vegt 9 & Gooitzen M. van Dam1,10,11
1234567890():,;

During the last decade, the emerging field of molecular fluorescence imaging has led to the
development of tumor-specific fluorescent tracers and an increase in early-phase clinical
trials without having consensus on a standard methodology for evaluating an optical tracer.
By combining multiple complementary state-of-the-art clinical optical imaging techniques, we
propose a novel analytical framework for the clinical translation and evaluation of tumor-
targeted fluorescent tracers for molecular fluorescence imaging which can be used for a
range of tumor types and with different optical tracers. Here we report the implementation of
this analytical framework and demonstrate the tumor-specific targeting of escalating doses of
the near-infrared fluorescent tracer bevacizumab-800CW on a macroscopic and microscopic
level. We subsequently demonstrate an 88% increase in the intraoperative detection rate of
tumor-involved margins in primary breast cancer patients, indicating the clinical feasibility
and support of future studies to evaluate the definitive clinical impact of fluorescence-guided
surgery.

1 Department of Surgery, University Medical Center Groningen, University of Groningen, Groningen 9700 RB, The Netherlands. 2 The Breast Center, Cancer

Hospital of Shantou University Medical College, Shantou 515000 Guangdong, China. 3 Department of Gastroenterology and Hepatology, University Medical
Center Groningen, University of Groningen, Groningen 9700 RB, The Netherlands. 4 Department of Clinical Pharmacy and Pharmacology, University Medical
Center Groningen, University of Groningen, Groningen 9700 RB, The Netherlands. 5 Department of Surgery, Martini Hospital, Groningen 9700 RM, The
Netherlands. 6 Department of Pathology, Martini Hospital, Groningen 9700 RM, The Netherlands. 7 Changjiang Scholar′s Laboratory of Shantou University
Medical College, 515000 Shantou, Guangdong, China. 8 Erasmus Medical Center Rotterdam, Rotterdam 3015 GD, The Netherlands. 9 Department of
Pathology, University Medical Center Groningen, University of Groningen, Groningen 9700 RB, The Netherlands. 10 Department of Nuclear Medicine and
Molecular Imaging, University Medical Center Groningen, University of Groningen, Groningen 9700 RB, The Netherlands. 11 Department of Intensive Care,
University Medical Center Groningen, University of Groningen, Groningen 9700 RB, The Netherlands. These authors contributed equally: Marjory Koller, Si-Qi
Qiu. Correspondence and requests for materials should be addressed to G.Dam. (email: [email protected])

NATURE COMMUNICATIONS | (2018)9:3739 | DOI: 10.1038/s41467-018-05727-y | www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05727-y

M
olecular fluorescence imaging enables visualization of this novel analytical framework and confirm the tumor-specific
tumor-specific, upregulated proteins and biological targeting of the near-infrared (NIR) fluorescent tracer
processes involved in oncogenesis and allows real-time bevacizumab-800CW in escalating doses on a macroscopic and a
imaging of tumor tissue with a high resolution for various clinical microscopic level. We subsequently observed an 88% increase in
applications, such as image-guided surgery, endoscopy, and the intraoperative detection of tumor-involved margins, thus
pathology. During the past decade, several tumor-specific fluor- indicating clinical feasibility in support of future studies to eval-
escent tracers have been developed and validated in animal uate the definitive clinical impact of fluorescence-guided surgery
models, leading more recently to a substantial increase in early- in primary breast cancer patients.
phase clinical trials evaluating molecular fluorescence imaging1,2.
Despite the increasing activity in the field, several critical factors Results
to ensure translation of optical tracers to clinical applications Summary of study design and patient demographics. The study
remain insufficiently established. No widely accepted analytical was designed as a clinical dose escalation trial investigating four
framework or standard evaluation methodology serves as a gold doses of bevacizumab-800CW (4.5 mg, 10 mg, 25 mg, and 50 mg)
standard for determining the efficacy of a fluorescent tracer in in patients with invasive T1–T2 primary breast cancer scheduled
clinical applications. for breast cancer surgery. Bevacizumab-800CW was injected
The majority of these early-phase clinical studies have been intravenously three days prior to surgery (Fig. 1a). A step-up
executed in image-guided surgery applications. In oncological approach was used in which three patients per dose group were
surgery, it is important to remove the tumor completely without included, followed by expansion of the two best-performing dose
any residual disease, since incomplete resections are inevitably groups to a total of ten patients each (Supplementary Fig. 1).
associated with higher rates of re-operations, increased rates of Twenty-six patients with invasive primary breast cancer were
recurrent disease and lower overall survival3. Intraoperatively, enrolled between 12 October 2015 and 2 February 2017. Three
surgeons are mainly dependent on visual inspection and palpa- patients received 4.5 mg, ten patients 10 mg, ten patients 25 mg,
tion alone to distinguish cancer tissue from benign tissue, a and three patients 50 mg of bevacizumab-800CW. No serious
method with unknown accuracy. The available intraoperative adverse events, allergic or anaphylactic reactions, were reported in
techniques for margin assessment have not yet been adopted this trial. Two adverse events were reported, one patient from the
universally. Frozen section analysis and imaging techniques like 4.5 mg group experienced nausea till 30 min after tracer injection,
specimen radiography are time consuming and lack diagnostic another patient from the 25 mg group had hot flushes after tracer
accuracy4. Anatomical imaging modalities like CT and MRI have administration that recovered spontaneously. None of the
been adapted for use in the operating theatre, but cannot be used patients felt any burden of the infusion three days prior to
in real-time and have limited tumor specificity. Consequently, surgery.
there is an unmet need for real-time tumor-specific imaging that In 19 patients, histopathological analyses showed an invasive
is compatible with the workflow in the operating theatre. This carcinoma of no special type (NST); in five patients a lobular
might be provided by using sensitive optical imaging techniques carcinoma, in one patient a mucinous carcinoma and in one
combined with tumor-specific fluorescent tracers; this approach patient a papillary carcinoma. In four patients, there was a tumor-
is currently under investigation in early-phase clinical trials1. involved surgical margin of the invasive primary tumor; in four
As there is no consensus on a standard evaluation methodol- other patients the surgical margin of unexpected additional
ogy for fluorescence imaging, we implemented a novel analytical carcinoma in situ was positive adjacent to a completely removed
framework for data collection, and fluorescence image analyses primary tumor. This resulted in a total positive-margin rate of
based on our experience in molecular fluorescence imaging in 30% according to the most recent SSO-ASTRO guidelines9
surgery and endoscopy5–8. In the present study we implement (Table 1).
Macroscopic imaging Microscopic imaging
a Tracer injection Intraoperative Fresh specimen Fresh tissue slice FFPE block 10-µm-thick section 4-µm-thick section
High
b d f h j l
Serially Paraffin
slicing embedding
3 days

25 µm

800 CW Hoechst 480 nm

c e g i k
m

25 µm Low

Safety Intraoperative imaging Specimen imaging Macro-segmentation Micro-segmentation Fluorescence microscopy

potential clinical impact potential clinical impact tumor-to-background ratio biodistribution cellular distribution

Fig. 1 The clinical analytical framework enabling correlation of intraoperative fluorescence signals with histopathology, from macroscopic to microscopic
levels. a Intravenous administration of bevacizumab-800CW three days prior to surgery. b, c Color image and corresponding fluorescence image obtained
in vivo during surgery to determine potential clinical value. d, e Imaging of the fresh surgical specimen, followed by serially slicing. f, g Imaging of the fresh
tissue slices to determine tumor-to-background ratio based on macro-segmentation, followed by paraffin embedding. h, i Imaging of formalin-fixed
paraffin-embedded (FFPE) blocks to determine heterogeneity of tracer uptake within a tumor. j, k Imaging of 10-µm-thick tissue sections for micro-
segmentation to reveal microscopic biodistribution and correlation with fluorescence signals from the macroscopic to microscopic level. l, m Fluorescence
microscopy to determine tracer distribution on a cellular level. Scale bars represent 1 cm, in l, m the scale bar represents 25 µm

2 NATURE COMMUNICATIONS | (2018)9:3739 | DOI: 10.1038/s41467-018-05727-y | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05727-y ARTICLE

Table 1 Demographics of study patients

4.5 mg 10 mg 25 mg 50 mg
n=3 n = 10 n = 10 n=3
Patient characteristics
Age (years/range) 72 (68–77) 61 (50–69) 57 (49–69) 63 (53–70)
Clinicopathological parameters (number)
Tumor type invasive primary tumor
Invasive carcinoma of no specific type 2 9 6 2
Lobular caricnoma 0 1 4 0
Mucinous carcinoma 0 0 0 1
Papillary carcinoma 1 0 0 0
Tumor size (cm/range) 1.5 (1.4–1.7) 1.3 (0.5–2.4) 1.8 (0.7–3.2) 0.9 (0.8–1.1)
Tumor grade (modified Bloom-Richardson)
Grade I 0 4 0 0
Grade II 3 4 7 0
Grade III 0 2 3 2
n/a — — — 1
Estrogen receptor positive (>10%) 3 9 8 3
Progesterone receptor positive (>10%) 2 8 7 2
HER2 receptor positive
(IHC 2 + or 3 + with positive FISH) 1 1 1 1
Carcinoma in situ present 1 6 9 3
Safety data (number)
Adverse events 1 0 1 0
Serious adverse events 0 0 0 0
Surgical resection margin status (number)a
Primary tumor
Free 3 7 9 3
Not free 0 3 1 0
Additional Carcinoma in situ component
Free 1 5 7 2
Not free 0 1 2 1
aDenotes according to ASTRO guidelines
HER2 human epidermal growth factor receptor 2, IHC immunohistochemistry, FISH fluorescence in situ hybridization

The analytical framework. Breast-conserving surgery consists of modality are shown in Fig. 2a–x. A complete overview per patient
intraoperative assessment of the margins by the surgeon using is shown in Supplementary Fig. 2. Sodium dodecyl sulfate poly-
visual inspection and palpation and evaluation of the excised acrylamidegel electrophoresis (SDS-PAGE) of tumor lysates
specimen by standard histopathology. Therefore, the data collec- demonstrated that the complete compound bevacizumab-800CW
tion procedure within the analytical framework to determine the was intact and present in the human primary breast tumor, as
tumor-specific targeting of bevacizumab-800CW should fit the confirmed by comparing the height of the band of the tumor
standard workflow and needs to provide qualitative and quanti- lysates with the lane containing diluted bevacizumab-800CW
tative data on fluorescence related to the standard of care. As such, (Supplementary Fig. 3).
the data collection procedure within the analytical framework We used fluorescence images of all fresh tissue slices obtained
consisted of: (i) qualitative in vivo intraoperative macroscopic by a light-tight macroscopic imaging device for quantitative
imaging to determine the potential clinical value of fluorescence- macro-segmentation analyses to determine the tumor-specific
guided surgery (Fig. 1b, c), (ii) qualitative ex vivo imaging of the targeting of bevacizumab-800CW on a macroscopic level by
fresh whole surgical specimens (Fig. 1d, e), (iii) quantitative calculating the tumor-to-background ratio (TBR). Freshly excised
ex vivo imaging of the fresh tissue slices of the fresh specimens to tissue represents the in vivo situation the best for the calculation
determine the tracer distribution on a macroscopic level (Fig. 1f, of the TBR, as it has not yet been processed with formalin or
g), (iv) quantitative multi-diameter single-fiber reflectance/single- embedded in paraffin and the conditions of imaging are the most
fiber fluorescence (MDSFR/SFF) spectroscopy of the fresh tissue optimally standardized; i.e., the tumors within the slices are all on
slices to determine the intrinsic fluorescence intensities, (v) the surface without overlaying tissue, the distance from stage to
quantitative fluorescence flatbed scanning of formalin-fixed par- camera is equal in all patients, and no ambient light is influencing
affin-embedded (FFPE) blocks and 10-µm-thick sections to the fluorescent signals. In the fluorescence images of all fresh
determine the tracer distribution on a microscopic level tissue slices that contained tumor tissue after confirmation with
(Fig. 1h–k), and (vi) fluorescence microscopy (Fig. 1l, m). histology, regions-of-interests of tumor tissue, as well as back-
ground tissue were manually segmented. The mean fluorescence
Quantitative macro-segmentation of the fresh tissue slices. In intensity (MFI) was measured per region-of-interest (ROI) and
all patients, qualitative assessment of fluorescence signals showed averaged per tissue type, resulting in an MFI of tumor tissue and
higher fluorescence signal intensities in tumor tissue compared to an MFI of background tissue per patient. TBR was calculated as
normal surrounding tissue at all ex vivo imaging modalities, the ratio of MFI of tumor compared to the MFI of normal tissue.
including the fresh tissue slices, FFPE blocks and 10-µm-thick Macro-segmentation analyses were performed on a total of 69
sections. Representative images per dose group and per imaging fresh tissue slices from 23 patients. In three patients, the light-

NATURE COMMUNICATIONS | (2018)9:3739 | DOI: 10.1038/s41467-018-05727-y | www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05727-y

4.5 mg 10 mg 25 mg 50 mg I Macro-segmentation
a g m s 40,000 * 5

Mean fluorescence intensity

ns

Tumor-to-background ratio
* 4
30,000 †
Fresh tissue slice

(MFI)
20,000 †

2
10,000
b h n t 1

0 0
4.5 10 25 50 4.5 1 2 50
mg mg mg mg mg 0 mg 5 mg mg

II MDSFR/SFF spectroscopy
0.0020 10
c i o u *

Tumor-to-background ratio
†

8
0.0015
6

Q. fa,x
0.0010
FFPE block

† 4
0.0005
2
d j p v
0.0000 0
4.5 10 25 50 4.5 1 2 50
mg mg mg mg mg 0 mg 5 mg mg

III Micro-segmentation
14,000 15
e k q w † ns
Mean fluorescence intensity 12,000

Tumor-to-background ratio
10,000
10
8000
(MFI)
10-µm-thick section

6000 † †
5
4000

f l r x 2000
0 0
4.5 10 25 50 4.5 1 2 5
mg mg mg mg mg 0 mg 5 mg 0 mg

Legend: Tumor tissue Normal tissue Individual data point Low High

Fig. 2 Representative images per dose group and per optical imaging method for ex vivo analyses, including MDSFR/SFF spectroscopy. Columns represent
the four dose groups: 4.5 mg (a–f), 10 mg (g–l), 25 mg (m–r), 50 mg (s–x). Rows represent the imaging modality, in the upper part a white light image and
in the lower part the representative fluorescence image. Tumor tissue is delineated with a dashed line. Scale bars represent 1 cm. I Mean fluorescence
intensity (MFI) of normal tissue (gray) and tumor tissue (black) are depicted per dose group on the left y-axis, the right y-axis shows the tumor-to-
background ratio per patient per dose group for macro-segmentation analyses, in II for MDSFR/SFF spectroscopy measurements and in III for micro-
segmentation analyses. Fluorescence images are scaled using the most optimal minimum and maximum displayed value. Boxplot centerline is at median,
the bounds of the box at 25th to 75th percentiles, the whiskers are depicting the min–max, tumor-to-background ratio data are depicted per patient; line
indicates median value per dose group. Asterisk denotes significant (P < 0.05, Kruskal–Wallis test) values. Obelisk denotes significant (P < 0.05,
Mann–Whitney U-test) values. FFPE formalin-fixed, paraffin embedded, MDSFR/SFF multi-diameter single-fiber reflectance/single-fiber fluorescence.
Q.μfa,x the product of the quantum efficiency across the emission spectrum, Q[-], where Q is the fluorescence quantum yield of IRDye-800CW and
μaf [mm−1] is the tracer absorption coefficient at the excitation wavelength

tight macroscopic fluorescence imaging device malfunctioned; MDSFR/SFF spectroscopy. MDSFR/SFF spectroscopy was per-
these patients were excluded. Quantitative macro-segmentation formed on the fresh tissue slices in order to quantify the intrinsic
analyses confirmed significantly higher fluorescence signals in fluorescence by correcting the fluorescence signal for the tissue
tumor tissue relative to normal background tissue in the 10 and optical properties scattering and absorption. Per patient three
25 mg dose groups (Fig. 2-I). The MFI of tumor tissue increased spots in the same fresh tissue slice were measured of both tumor
from a median of 5368 in the 4.5 mg group to a median of 18,472 tissue and normal tissue, per spot three measurements were done.
in the 50 mg group (Fig. 2-I). The 25 mg dose group showed a In the 13 patients with available MDSFR/SFF data, intrinsic
significantly higher MFI in tumor tissue compared to tumor fluorescence in tumor tissue was significantly higher compared to
tissue in the 10 mg dose group (median MFI 25 mg = 14,390, normal tissue in the 10 mg dose group (P = 0.0022,
median MFI 10 mg = 6014; P = 0.0297, Kruskal–Wallis test). No Mann–Whitney U-test) and the 25 mg dose group (P = 0.0159
increase in MFI of normal background tissue was observed Mann–Whitney U-test) (Fig. 2-II). Furthermore, a larger varia-
between the 10 and 25 mg dose groups (P = 0.0880, tion of fluorescence intensity between patients was observed in
Kruskal–Wallis test), resulting in a significantly higher TBR of the 25 and 50 mg groups. When comparing results of MDSFR/
3.07 in 25 mg group patients versus 1.79 in 10 mg group patients SFF spectroscopy with macro-segmentation of the fresh tissue
(P = 0.0097, Kruskal–Wallis test)(Fig. 2-III). slices, a similar trend of increasing fluorescence levels in tumor

4 NATURE COMMUNICATIONS | (2018)9:3739 | DOI: 10.1038/s41467-018-05727-y | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05727-y ARTICLE

HE / whole slide Tumor Parenchyma Fat Carcinoma in situ Combined

a c e g i k
Region of interest

b d f h j l
Fluorescence image

m
12,000
† 4.5 mg
10 mg
10,000
25 mg
Mean fluorescence intensity (MFI)

50 mg
* †
8000 Individual data point

*
6000

*
4000

ns
2000

*
0
Tumor Carcinoma Entire normal tissue Parenchyma Fat
in situ

Fig. 3 Microscopic biodistribution in breast cancer tissue of bevacizumab-800CW based on micro-segmentation analyses. The upper row shows a
representative example of the region-of-interest per tissue type based on H/E staining. The lower row shows the corresponding pseudo color fluorescence
intensity image of each tissue type. In a, b the whole section is depicted, and in c, d the tumor area, e, f parenchymal breast tissue including collagen,
g, h fat tissue, i, j carcinoma in situ tissue, and a combination of all tissue types in k, l. Mean fluorescence intensities of all patients per dose group, per
tissue type are shown in panel m. Asterisk denotes significant (P < 0.05, Kruskal–Wallis test) values. Obelisk denotes significant (P < 0.05, Mann–Whitney
U-test) values. Bars are representing the median, error bars are representing 95% confidence interval. Scale bars represent 5 mm

tissue with escalating tracer doses was observed, whereas no (median per dose group is indicated with a horizontal line). In
difference in fluorescence levels was measured in background five patients, the tumor-to-parenchyma ratio was below 1, what
normal breast tissue between the dose groups. means that the tumor MFI was lower than the MFI of the par-
enchyma tissue (Fig. 4j).

Quantitative micro-segmentation of 10-µm-thick FFPE sec-

tions. To assess the detailed microscopic biodistribution of Potential clinical value of fluorescence-guided surgery. Since
bevacizumab-800CW in human breast tissue, we performed macro-segmentation analyses and micro-segmentation analyses
micro-segmentation on a total of 200 10-µm-thick FFPE sections confirmed tumor-specific targeting of bevacizumab-800CW
(Fig. 3a–l). In all 26 patients, tumor tissue showed a higher MFI irrespective of the dosing, we evaluated the potential clinical
compared to the entire normal breast tissue. Normal tissue was value of molecular-fluorescence-guided surgery in breast cancer
defined as fat and parenchymal breast tissue including collagen patients. We qualitatively analyzed the intraoperative fluores-
(Fig. 2-III, Fig. 3m). When analyzing the MFI per tissue type per cence image and video data in combination with the fluorescence
dose group, we observed an increase in MFI for all tissue types. images of the freshly excised specimen. Intraoperative imaging
However, a higher tracer uptake in tumor tissue remains in took place at two time points during surgery; the tumor was
escalating doses, which indicates tumor-specific targeting of the imaged just before excision and the surgical cavity was imaged
tracer irrespective of dosing (Fig. 4e–i). To visualize the differ- directly after removal of the tumor. Since this clinical trial was not
ences in tumor tissue and normal parenchyma, we plotted the designed to alter the standard of care, surgeons were not allowed
tumor-to-parenchyma ratio, per patient and per dose group to excise additional tissue based on intraoperatively detected

NATURE COMMUNICATIONS | (2018)9:3739 | DOI: 10.1038/s41467-018-05727-y | www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05727-y

a b c d
4.5 mg 10 mg 25 mg 50 mg
15,000

Mean fluorescence intensity (MFI)

Mean fluorescence intensity (MFI)

4000 4000

Mean fluorescence intensity (MFI)

15,000
Mean fluorescence intensity (MFI)

3000 3000
10,000 10,000

2000 2000

5000 5000
1000 1000

0 0 0 0
Tu itu

H enc in
M rp m
or p ia

ry u in

in i m ch e
al oa in m
ll no tu
an a
m C es
tis st
e

ar P ad Tum in

oc ac a chy a
rie H tive in m
m er ss u
Bl etapla ue
l a
a
al C l
su t
e
t

fil tu
te
re um in

e
e a C A
et iss t
la e
a
t
or

ci re ltr r

Ap Re nomen om r

M hymr
m sse
t is a
e
Fa

tis ys
Fa

m l t ys

Fa
ar Pa infi mo

ci ar en o

nc o
yl F no n at

ch m

su
Fa

ne yp ti sit

oo p si
N dveasi

cl

ap u
si
rin rm FE
r Sk

N eta las

to T Sk

al y

Sk

Pa T Sk
e y

al si
su

dr br a y
ce de si

in si
tr a
m
s

us
yp h

m la
in

ic in
a

yt a
m

oc om
Pa

o
no

or

or
br

ph in

oc No
ci

N
at

Fi

m rc
ar

L y Ca
C

C
m

Ap
fla
In

C
e Normal tissue f Mamma parenchyma g Fat
20,000
Mean fluorescence intensity (MFI)

Mean fluorescence intensity (MFI)

20,000 20,000
Mean fluorescence intensity (MFI)

15,000 15,000 15,000

10,000 10,000 10,000

5000 5000 5000

0 0 0
4.5 1 2 50 4.5 10 25 50 4.5 10 25 50
mg 0 mg 5 mg mg mg m g m g mg
mg m g m g mg

h Tumor tissue i Carcinoma in situ j 4

20,000 20,000
Mean fluorescence intensity

Mean fluorescence intensity

Tumor-to-parynchema ratio
15,000 15,000 3

10,000 10,000 2

5000 5000 1

0 0 0
4.5 1 2 50 4.5 1 2 50 4.5 10 25 50
mg 0 mg 5 mg mg mg 0 mg 5 mg m g
m g mg mg mg

Fig. 4 Micro-segmentation per dose group, and per tissue type. Per dose group we plotted mean fluorescence intensity per tissue type (a–d); tumor and
carcinoma in situ components shown in red. The mean fluorescence intensity per tissue type was plotted in e-i. In j the tumor-to-parenchyma ratio per
dose group is plotted. Bars represent the mean and the error bars the standard deviation. Boxplot centerline is at median, the bounds of the box at 25th to
75th percentiles, the whiskers are depicting the min–max

fluorescence signals. Therefore, intraoperative findings could only imaging, suggesting a tumor-positive resection margin. In three of
be retrospectively correlated with histopathology. Representative those seven patients, the primary tumor was not completely
examples of fluorescence images from a patient with a tumor- resected, whereas in four other patients the surgical margin
involved surgical margin, and from a patient with a tumor-free contained additional carcinoma in situ components next to a
surgical margin are presented in Fig. 5. We observed in the completely resected primary tumor. In one patient, in which
fluorescence scan of the 10 µm slide also non-fatty is lit up by the histopathological analysis showed ink on the invasive primary
fluorescent tracer (Fig. 5t). We further investigated the possible tumor, no fluorescence signal was detected in the surgical cavity.
cause of this high uptake by sectioning the tissue FFPE block In contrast to the intraoperative imaging of this patient, ex vivo
several slides deeper, and strikingly, in these deeper sections we analyses of the fresh tissue slices showed clear uptake of the tracer
found tumor tissue present at the site where the high uptake is in the tumor tissue, and also a close surgical margin was
visible in the original slide (Fig. 5v). It is known that VEGF is suspected on the fluorescence images of the fresh tissue slices
present is in the microenvironment of the tumor10. Probably, the (Supplementary Fig. 2b).
VEGF expressed in the non-fatty tissue is a field-effect from In 18/26 patients (69%) a tumor-free surgical margin was
secretion from deeper seated underlying tumor cells which might reported after histopathological analyses (Table 2). In 16 of these
explain the high bevacizumab-800CW uptake. 18 patients (89%) no intraoperative signals were detected in the
In eight of the 26 patients (31%) a tumor-involved surgical surgical cavity, whereas in the two remaining patients with a
margin was reported after histopathological analyses; using tumor-free surgical margin (2/26, 7.6%), a positive fluorescence
current clinical surgical practice, none of these margins were cavity signal was detected. In these two patients, high fluorescence
detected intraoperatively (Table 2). When using fluorescence, in signals were observed in surrounding healthy tissue containing
seven of these eight patients (88%) a clear fluorescence signal was abundant collagen, normal parenchyma, accompanied by ade-
detected in the surgical cavity by intraoperative fluorescence nosis and a periductular plasma cell infiltrate as detected in

6 NATURE COMMUNICATIONS | (2018)9:3739 | DOI: 10.1038/s41467-018-05727-y | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05727-y ARTICLE

Intraoperative Fresh specimen Fresh tissue slice FFPE Block 10-µm-thick section
Tumor positive surgical margin a c e g i

*
*

b d f h j

10-µm-thick section
deeper sectioning
k m o q s u
Tumor negative surgial margin

l n p r t v

Fig. 5 Representative examples of intraoperatively detected tumor-involved surgical margin and a tumor negative surgical margin. Columns represent from
left to right intraoperative imaging, fresh specimen imaging, fresh tissue slice imaging, FFPE block imaging, and imaging of 10-µm-thick sections. The two
upper rows represent a patient with a tumor-positive surgical margin, a clear fluorescence signal was detected in the surgical cavity. Subsequently, the
corresponding resection plane of the excised specimen was marked with an extra suture (a, b). Fluorescence imaging of the fresh surgical specimen
showed high fluorescence signals at the area of the suture mark (c, d, asterisk). Corresponding fluorescence images of fresh tissue slices, FFPE blocks and
10-µm-thick sections showed high fluorescence signals at the margin (e–j, arrows). Histopathology confirmed the presence of tumor deposits in this area
(i). The lower rows represent a patient with a tumor-free surgical margin Fig. 5 k–t. Deeper sectioning of the FFPE block (q, r) was performed to investigate
the probable cause of the high fluorescent area within the green dashed line (t). u, v Arrow depicts the surgical positive margin. Dashed white/black circle
indicates the area with the highest fluorescence signal intensities. The asterisk represents the position of the extra suture mark. The gray box represents
the origin of the FFPE block in the fresh tissue slice. The dashed white/black line delineates tumor tissue. The dashed green line delineates collagen tissue
with normal parenchyma. Scale bars represent 1 cm

micro-segmentation analyses, which could explain these findings not influence the structural integrity and post translational
(Supplementary Fig. 2d). modifications of bevacizumab not leading to an affected mode of
action by the IRDye-800CW conjugation16. Data derived from
Discussion preclinical studies confirm that Bevacizumab-800CW has a
In the emerging field of molecular fluorescence imaging a robust comparable biodistribution as 89Zr-Bevacizumab17.
and broadly applicable analytical framework for clinical transla- By implementing our novel analytical framework for the first
tion of fluorescent tracers is lacking. Based on our experience in time in the current study, we confirmed the tumor-specific tar-
the first clinical trials investigating fluorescence-guided surgery in geting of bevacizumab-800CW in escalating doses by tracing
human, we propose a standard evaluation methodology for down bevacizumab-800CW on both a macroscopic and micro-
clinical translation of fluorescent tracers by combining com- scopic level within the individual components of the proposed
plementary qualitative and quantitative clinical optical imaging analytical framework. Because we demonstrated the tumor-
techniques5–8. specific targeting of bevacizumab-800CW irrespective of dosing,
Earlier, we demonstrated that a microdose of bevacizumab- we subsequently showed the potential clinical value of
800CW specifically targets vascular endothelial growth factor A fluorescence-guided surgery in breast cancer patients, indicating
(VEGF-A) in patients with primary breast cancer7. VEGF-A is the clinical feasibility and support of future studies to evaluate the
present in all breast cancer types11–14, as it is a generic target definitive clinical impact of fluorescence-guided surgery in pri-
upregulated in many solid tumors and regarded one of the hall- mary breast cancer patients. Using fluorescence-guided surgery in
marks of cancer15. Besides, we have demonstrated earlier that the primary breast cancer patients, we showed that the intraoperative
antibody bevacizumab still has intact affinity for the target after detection of tumor-involved margins is much better than stan-
conjugation with IRDye-800CW and the labeling procedure does dard surgical practice, this was confirmed by ex vivo analyses

NATURE COMMUNICATIONS | (2018)9:3739 | DOI: 10.1038/s41467-018-05727-y | www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05727-y

Table 2 Contingency table of molecular fluorescence-guided surgery in breast cancer patients

Surgical margin tumor positive Surgical margin tumor negative Total

Fluorescence signals in cavity positive 7 2 9
Fluorescence signals in cavity negative 1 16 17
Total 8 18 26

within the analytical workflow. In seven out of eight patients, Assuming that 15 patients with a tumor-involved surgical margin
tumor-positive resection margins were detected during is sufficient per dose group, a total of 45–75 patients per dose
fluorescence-guided surgery that were missed by intraoperative group might be needed, given the 20–30% tumor-involved sur-
assessment of surgical margins using standard visual inspection gical margin rate in breast cancer surgery known from this study
and palpation. Because the tumor-involved surgical margins and from literature.
could be detected intraoperatively in real time, these patients We observed a larger variation of fluorescence intensities in
might have avoided additional surgery or therapy. This indicates tumor tissue between patients in the 25 and 50 mg dose groups
the clinical value of intraoperative molecular fluorescence ima- compared to 4.5 and 10 mg dose groups. Factors that might
ging in breast cancer patients and supports a paradigm shift in the indicate protein saturation in tumors in doses from 25 mg on
future treatment of breast-conserving surgery, however the long- might be tumor size, tumor grade, and tumor type. Data derived
term impact of molecular fluorescence imaging on relevant from clinical studies evaluating cetuximab-800CW targeting
clinical endpoints needs to be confirmed in next phase clinical Endothelial Growth Factor Receptor (EGFR) in head- and neck
trials, for instance the reduction in positive-margin rates cancer, we learned protein saturation occurs in higher dose
(Table 2). groups as it has shown decreasing TBRs with higher doses18.
Besides guiding intraoperative decision making, fluorescence Based on literature data, it is known that higher VEGF mRNA
imaging could also have a significant impact on the workflow of expression values are associated with higher grade tumors, but
pathological analysis. In current clinical practice, histological also with negative ER/PR status, and positive HER2 status19.
analysis of the complete surgical specimen is not possible due to Most likely, the small sample size within our study limits defi-
practical and logistical constraints. Moreover, sampling tissue for nitive conclusions about correlation of tracer uptake with clin-
histological analyses is based only on gross examination by visual icopathological parameters. Therefore, the correlation of
inspection and palpation of the fresh serially sliced specimen by fluorescence intensity and clinicopathological parameters needs
the attending pathologist; therefore, tumor-involved margins may to be investigated in a next phase clinical trial.
not be included in total in the FFPE tissue blocks, thus causing a In five patients, the tumor-to-parenchyma ratio in the micro-
sampling error. Macroscopic fluorescence imaging of the fresh segmentation results was below 1, what means that the tumor
surgical specimen and the fresh tissue slices can provide the MFI was lower than the MFI of the parenchyma tissue. Especially
pathologist with a red-flag technique that precisely outlines tumor in the micro-segmentation results it becomes more apparent that
tissue (i.e., image-guided pathology). This could optimize current the tumor-to-parenchyma ratio is lower than the TBR, compared
tissue sampling procedures and prevent sampling errors. to for example the results of the macro-segmentation analyses of
Importantly, in our study we confirmed the cross-correlation of the fresh tissue slices. Although parenchymal tissue including
fluorescence-guided surgery with final histopathology, considered collagen showed high tracer uptake, it is to be expected that this
the gold standard. This is crucial for the further implementation tissue will only attribute relatively to background fluorescence
of fluorescence image-guided histopathology. intensity intraoperatively, which is supported by the fact that in
Additionally, the current intraoperative clinical workflow is only one out of these five patients this resulted in a false positive
constrained by a considerable time lag between the clinical cavity signal according to the intraoperative image analyses
decision making of the surgeon (intraoperative evaluation, min- (Supplementary Fig. 2d first row). Only large areas of par-
utes-hour) and the determination of presence or absence of a enchymal tissue including collagen may influence the TBR
tumor-positive margin by the pathologist (post-operative eva- in vivo, which might be challenging in patients with a tumor
luation, days-week). We propose that image-guided pathology directly behind the nipple and in premenopausal patients with
might bridge the gap between the surgical theatre and the dense breasts.
pathology laboratory for reliable margin assessment, as fluores- We described the analytical platform which is optimized for
cence images of the specimen can be provided in real-time and 800 nm optical agents in particular in terms of instrumentation
simultaneously to both disciplines, which will lead to a dynamic adapted to NIR fluorescence imaging (i.e., around the 800 nm
interaction between in vivo intraoperative imaging and ex vivo range). Moreover, the analytical workflow is generally applicable
macroscopic imaging of the surgical specimen. This could for analyses of optical agents with other wavelengths, considering
improve surgical outcome when it counts the most—during when a paired detection camera is used that is adapted to the
surgery—with a direct impact on clinical decision making. particular wavelength of interest of the fluorophore.
Although this study was designed as a dose escalation study, we In conclusion, by implementing a novel analytical workflow for
cannot draw definitive conclusions on the optimal tracer dose for molecular fluorescence imaging we have demonstrated the clin-
clinical decision making. This is due to the relatively small ical feasibility of molecular fluorescence-guided surgery using the
number of patients included in the lowest and highest dosing fluorescent tracer bevacizumab-800CW in breast-conserving
groups, leading to an unequal distribution of patients with tumor- surgery. A larger study including clinical endpoints is needed to
involved surgical margins in the four dose groups. While the confirm the optimal dose of bevacizumab-800CW to be used in a
current study already showed the value on detecting tumor- next phase randomized clinical trial. Furthermore, our analytical
involved surgical margins by an increased detection rate of 88%, platform could be used in future clinical studies on the clinical
sufficient data points are needed to determine the definitive translation and evaluation of other tumor-targeted fluorescent
diagnostic accuracy and to derive the optimal threshold of tracers for molecular fluorescence-guided surgery, and also in
fluorescence intensities for intraoperative decision making. different tumor types. Therefore, this analytical platform might

8 NATURE COMMUNICATIONS | (2018)9:3739 | DOI: 10.1038/s41467-018-05727-y | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05727-y ARTICLE

serve as a standard for data collection and fluorescence image QT correction for heart rate was done using the Bazett formula. In the first 12
analyses in trials investigating molecular fluorescence imaging patients of the current study, and in 17 patients in the clinical trial NCT02113202
no QTc prolonging was observed when patients received 4,5, 10, 25, and 50 mg
(Supplementary Fig. 5). bevacizumab-800CW, therefore the local investigational review board and the data
safety monitoring board agreed to terminate ECG measurements in Part II of this
Methods trial. Patients were asked for signs and symptoms before tracer injection, during
Bevacizumab-800CW synthesis. Clinical grade bevacizumab-800CW was pro- one-hour observation period after tracer injection, and prior to surgery. After
duced in the good manufacturing practice (GMP) facility of the UMCG by con- surgery, a post-surgery follow-up assessment was performed within two weeks. At
jugating bevacizumab (Roche AG) and IRDye-800CW-NHS (LI-COR Biosciences this visit wound healing and adverse events were monitored.
Inc) under regulated conditions16. The average conjugation molecule ratio of
bevacizumab (molecular weight: 149 KDa) to IRDye-800CW-NHS (molecular Standard surgical procedure. Patients underwent either a lumpectomy (n = 24)
weight: 1.166 KDa) was 1:2, generating the conjugate bevacizumab-800CW with a or a mastectomy (n = 2) with or without a sentinel lymph node biopsy or axillary
total molecular weight of 151.3 KDa. Vials containing 6.0 mg bevacizumab-800CW lymph node dissection, according to institutional standard of care procedures and
dissolved in 0.9% sodium chloride (NaCl) solution were used to prepare the guidelines. Tumor localization was done with either manual palpation, wire gui-
infusions in a concentration of 1 mg ml−1. After release of the final product by the dance, or using an iodine seed according to standard clinical care. Sentinel lymph
certified qualified person at the UMCG GMP facility, the tracer was intravenously node mapping was done using 99mTechnetium using a gamma-probe, 99mTech-
administered to the subjects. netium was injected intratumorally one day before surgery conform standard
clinical care.
Gel electrophoresis. Tumor lysates of a patient from the 10 mg group, and one Based on our previous experience in fluorescence imaging we adapted the
patient from the 25 mg group were analyzed by sodium dodecyl sulfate poly- standard of care minimally. We used blue non-fluorescent sterile covers in this
acrylamidegel electrophoresis (SDS-PAGE), to ensure the complete compound study and avoided blue dye injection for sentinel lymph node mapping, as green
bevacizumab-IRDye800CW was present in the primary breast tumor. Additional, a color sterile covers and patent blue interfere with fluorescence signals.
lysate of normal tissue was analyzed. Results were compared with labeled and
unlabeled clinically used bevacizumab. The gel was scanned with the Odyssey Intraoperative fluorescence imaging device. We used a fluorescence camera
flatbed scanner at the 800 nm channel. system dedicated to detect IRDye-800CW-NHS (SurgVision BV ‘t Harde, The
Netherlands). The system was configured with two LED lights for 800 nm illu-
Clinical trial design. The dose-finding study was performed in two centers in 26 mination and one LED light for white light illumination. Real-time color and NIR
patients with proven primary breast cancer scheduled for surgery. This study was fluorescence images and videos were acquired simultaneously with custom software
approved by the Institutional Review Board of the University Medical Center at video rate. Fluorescence was detected using a highly sensitive electron-
Groningen (UMCG, Groningen, the Netherlands) for conduction of the study in multiplying charge-coupled device (EMCCD) imaging sensor. In the color-NIR
both the UMCG and in the Martini Hospital (MZH; Groningen, the Netherlands), overlay images, 800 nm images were pseudo colored green. The working distance
a peripheral training hospital being representative for the general population of of the imaging system was 20 cm above the surgical field with a field of view of 15
breast cancer patient operated on in The Netherlands. The study was conducted cm × 15 cm, and a spatial resolution of ~2-line pairs per millimeter. For each
according to the principles of the Declaration of Helsinki and according to the experiment, settings were held constant on 50 ms exposure time and 300 gain; if
Dutch Act on Medical Research involving Human Subjects (WMO). Patients with fluorescence oversaturation occurred in higher dose groups we lowered the gain to
proven primary breast cancer scheduled for surgery were recruited during multi- 30 or 3 accordingly. Images and videos were recorded and stored in raw Flexible
disciplinary breast cancer meetings in either the UMCG or Martini Hospital. Eli- Image Transport System (FITS) format.
gible patients were given orally and written information about the study and the Before and after each surgical procedure the intraoperative camera system was
option to participate. All human participants gave written informed consent before calibrated using a calibration device (CalibrationDisk, SurgVision BV, The
the start of the study procedures. An independent data safety monitoring board Netherlands). The device consists of a disk with round windows that can hold 8
was appointed prior to the inclusion of the first patient to evaluate safety measures. clear polypropylene tubes of 0.65 ml (Catalog #15160, Sorenson, BioScience, Inc,
Serious adverse events, if present, were immediately reported to the investigational Murray, U.S.A.) (Supplementary Fig. 4). The tubes were filled with 2% intralipid
review board of the UMCG, the data safety monitoring board, and the Dutch and two-fold increasing concentrations of bevacizumab-800CW from 1:6400 till
central committee on research involving human subjects (CCMO). The trial was 1:100 including one tube without tracer. The CalibrationDisk was used to test the
registered at www.ClinicalTrials.gov (identifier: NCT02583568). system prior to and after surgery, whether low and high fluorescent signals could be
We designed an adapted 4 × 3 dose-finding study design, adhering to the FDA detected from dilutional series and whether the system was functioning
guidelines (Guidance for Industry, Developing Medical Imaging Drug and appropriately.
Biological Products, Part 2 Clinical Indications). This study consisted of two parts.
In part I, four ascending flat doses of 4.5 mg (4.5 mL), 10 mg (10 mL), 25 mg (25 Intraoperative imaging procedures. This clinical trial was not designed to alter
mL), and 50 mg (50 mL) bevacizumab-800CW were intravenously administered to the standard of care, and surgeons were not allowed to excise additional tissue
three patients each. The dosing scheme that was used in the trial is based on the based on fluorescence signals intraoperatively detected. Therefore, intraoperative
definition of microdosing. We wanted to be sure to stay more than three times fluorescence imaging took place at two predefined time points during the surgical
below the therapeutic dose in the highest dose group. For patients who are on procedure: (1) after skin incision the tumor area was imaged just before excision of
combination therapy with bevacizumab to treat their cancer, it is commonly the complete surgical specimen, and (2) after removal of the specimen the surgical
accepted that the patient can safely undergo surgery 6 weeks after termination of cavity was inspected for remaining fluorescence signals. During imaging, the sur-
the bevacizumab therapy: i.e., at this time the anti-angiogenetic effects have geon was looking at a stand-alone computer monitor connected to the intrao-
diminished sufficiently to assure there is no increased risk of bleeding or post- perative imaging system. During the imaging procedures the ambient light of the
operative complications related to bevacizumab. The plasma levels of bevacizumab surgical theater is switched off in order to prevent interaction of the ambient light
after a wash out period of 6 weeks equals the peak plasma levels after a 160 mg IV with the fluorescence signals and also to have the highest sensitivity for detection of
dose (as calculated by the Hospital Pharmacy and the department of Medical fluorescent signals during surgery, because the surgical field is also illuminated by
Oncology at the UMCG). Since the Bevacizumab-800CW will be used in surgery, the white light of the camera system, the surgeon can still see in real life what
the dose should stay below 160 mg total injected dose, for which the maximum flat occurs in the surgical field. This set up did not influence the standard of care.
dose of 50 mg in this clinical trial stays significantly below. We administered a flat
dose per cohort, the dose was not adjusted for body weight or body surface area.
In part II, the most optimal performing dose group and one de-escalating dose Specimen handling. After excision of the surgical specimen orientation marks
were chosen on the basis of TBR to be expanded to a total of 10 subjects in each were placed according to standard clinical care. A short-short suture marked the
group in order to obtain a sufficient number of data points to decide on the optimal posterior side of the specimen and a long-long suture marked the nipple side of the
dose for a future phase III clinical study (Supplementary Fig. 1). Patients received a specimen.
single dose of one of the 4 dosages bevacizumab-800CW three days prior to
surgery. The lower doses of 4.5 and 10 mg were injected by slow bolus injection,
and for 25 and 50 mg an infusion pump was used (infusion speed: 150 mL per Fluorescence imaging systems for ex vivo imaging. The light-tight macroscopic
hour). After injection, the infusion line was flushed with 5 mL 0.9% NaCl. fluorescence imaging device (SurgVision BV, The Netherlands) is designed for
ex vivo fluorescence imaging and consists of an object table and a Complementary
Metal Oxide Semiconductor (CMOS) camera which are fully shielded by a light-
Safety measurements. Vital signs were measured prior to tracer injection, shielded box in order to create a dark imaging environment. The distance between
immediately after tracer injection and one-hour post-injection. Before tracer the object table and the CMOS camera is fixed with a field of view of 10 cm by 10
administration blood levels of potassium, magnesium, and calcium was measured. cm. For each experiment, settings were held constant with a fluorescence exposure
A pregnancy test was performed if patients were premenopausal. A standard 12- time of eight seconds. In two cases the light-tight macroscopic fluorescence ima-
lead electrocardiogram (ECG) was made before tracer injection and one-hour post- ging device did malfunction and the intraoperative imaging system was used for
injection. The following parameters were reported: heart rate, QT- and QTc time. imaging of the fresh surgical specimen and fresh tissue slices in a dark

NATURE COMMUNICATIONS | (2018)9:3739 | DOI: 10.1038/s41467-018-05727-y | www.nature.com/naturecommunications 9

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05727-y

environment. Before each experiment started, the ex vivo imaging device was Fluorescence microscopy. We made additional 4-µm-thick sections for micro-
calibrated with the same calibration device as previously described. scopic assessment of the NIR signal derived from bevacizumab-800CW in order to
The multi-diameter single-fiber reflectance/single-fiber fluorescence (MDSFR/ evaluate the tracer distribution at a cellular level. The cell nuclei were counter-
SFF) spectroscopy system calibrates scattering signals in the reflectance spectra and stained with Hoechst (33258, Invitrogen, Waltham, MA, USA) The sections were
provides a quantitative measurement of the NIR signal emitting from the mounted under a cover glass in modified Kaiser’s glycerin.
bevacizumab-800CW tracer20. The MDSFR/SFF spectroscopy device was calibrated
internally using a 6.6% intralipid phantom.
We used the Odyssey® CLX fluorescence flatbed scanning system (LI-COR Macro-segmentation of the fresh tissue slices. We used images of the fresh
Biosciences Inc. Lincoln, Nebraska) for detecting fluorescence in FFPE blocks and tissue slices of all 26 patients to determine the TBR per patient. TBR was defined as
10-µm-thick sections. the MFI measured in breast cancer tissue divided by the MFI in surrounding
An inverted microscope (DMI6000B, Leica Biosystems GmbH, Wetzlar, healthy tissue at macroscopic level. We used images of the fresh tissue slices as a
Germany) was used for fluorescence microscopy with a pixel size 6.45 µm, a field of representative model for the in vivo situation for the macro-segmentation analyses
view: 120 × 120 mm. To optimize NIR visualization, the microscope was equipped for calculating the TBR. Fresh tissue represents the in human situation best because
with additional accessories, including a NIR LED light source ranging up to 900 nm this tissue is not yet fixed with formalin or embedded in paraffin and the conditions
(X-Cite 200DC, Excelitas Technologies, Waltham, MA, USA), an NIR filter set of imaging are the most optimally standardized. The tumors within the slices are all
(microscope two band- pass filters 850–890 m–2p and a long-pass emission filter on the surface without overlaying tissue, the distance from stage to camera is equal
HQ800795LP; Chroma Technology Corp, Bellows Falls, VT, USA), a monochrome in all patients, and no ambient light is influencing the fluorescent signals. All raw
DFC365 FX fluorescence camera (1·4 M Pixel CCD, Leica Biosystems GmbH), and (FITS-format) fluorescence images of fresh tissue slices that contained tumor tissue
LAS-X software (Leica Biosystems GmbH). We used an acquisition time of 10 s for were imported in ImageJ (Fiji, version 1.0). ROIs were defined using the analytical
images of the 800 nm channel. workflow, because we could exactly correlate the origin of all FFPE blocks from the
fresh tissue slice. As we know this origin we used the corresponding histological
slice to confirm tumor areas and background areas of normal tissue in the fresh
tissue slices. ROIs of the total tumor tissue area, as well as the total background
Imaging procedures of the fresh surgical specimen. All the procedures took tissue per fresh tissue slice are defined by MK and drawn manually. Mean fluor-
place in a dark environment as much as possible, to prevent photobleaching of the escence intensities (MFI, arbitrary units) of all fresh tissue slices containing tumor
tracer. The fresh surgical specimen is handled conforming current clinical practice tissue were measured per ROI and averaged per tissue type per patient, resulting in
(see also page 18). Upon arrival at the pathology department, the fresh surgical a MFI of tumor tissue and MFI of background tissue per patient. The TBR was
specimen was imaged in the light-tight macroscopic fluorescence imaging device calculated for each patient by dividing the MFI of tumor tissue by the MFI of
on every six sides corresponding to the in vivo situation, which are anterior, surrounding healthy tissue. After each dose group was finished, MFI of tumor
posterior, medial, lateral, cranial, and caudal sides. The specimen was imaged on tissue, MFI of healthy surrounding tissue and TBR for each patient were plotted in
average of 60 min after removal of the tissue, image duration was 6 min per spe- graphs (GraphPad Prism, version 7.0b). Derived from previous studies executed
cimen. After freezing the whole fresh specimen in a −20 °C freezer for 15 min, the with 800CW labeled cetuximab18, it was anticipated that a plateauing of TBRs level
whole specimen was marked with black and blue ink, because these are non- might occur with increasing doses and therefore further increasing the dose is of no
fluorescent in the NIR range and do not interfere with the bevacizumab-800CW further clinical need in terms of imaging TBRs. Once the TBR reached a plateau, it
tracer signal. The limitation on the use of ink color did not affect the standard of was considered as an indicator that the optimal dose was reached.
care pathology practices in both institutions participating in the study. Subse-
quently, the fresh surgical specimen was serially sliced into 0.5-cm-thick fresh
MDSFR/SFF spectroscopy. The MDSFR/SFF spectroscopy gains two reflectance
tissue slices. A photograph of all fresh tissue slices was made. Before formalin
spectra via two different optical fibers and one raw fluorescence spectrum. The
fixation, all fresh tissue slices were imaged on both sides in the light-tight mac-
scattering and absorption coefficients were determined from the reflectance spec-
roscopic fluorescence imaging system. The fresh tissue slices were imaged on
tra, which were used to determine the intrinsic fluorescence (Q.μfa,x) of
average of 180 min after removal of the tissue, image duration was 15 min for both
bevacizumab-800CW by correcting the fluorescence spectrum for the calculated
sides of all fresh tissue slices of a patient. Furthermore, one fresh tissue slice per
tissue optical properties20. The intrinsic fluorescence Q.μfa,x is defined as the
patient which clearly contained tumor based on gross examination, was used for
product of the quantum efficiency across the emission spectrum, Q[-], where Q is
MDSFR/SFF spectroscopy analysis. We placed the MDSFR/SFF spectroscopy probe
the fluorescence quantum yield of IRDye-800CW and μaf [mm−1] is the tracer
on top of tumor tissue and normal tissue for quantitative measurements of NIR
absorption coefficient at the excitation wavelength. MDSFR/SFF spectroscopy was
fluorescence. Per patient three spots were measured of both tissue types, per spot
performed in UMCG patients only; since the system was not available in the MZH
three measurements were done. Thereafter, the fresh tissue slices were fixed in
center. In one patient, the measurements failed because the device malfunctioned.
formalin overnight. The next day, the pathologist macroscopically examined the
specimen and selected tissue samples that were embedded in paraffin blocks and
processed further for histological analyses. Tissue was embedded conforming Micro-segmentation for assessment of the biodistribution. We performed
standard clinical practice; in our institution the pathologist decides, based on visual micro-segmentation of 10-µm-thick FFPE sections to determine biodistribution of
inspection and palpation and gross examination, which tissue areas need to be bevacizumab-800CW in human breast tissue. In summary, after fluorescence
embedded in FFPE blocks. This study was performed without altering the standard scanning and HE staining of 10-µm-thick tissue sections, an experienced breast
of care and therefore we did not influence the pathologist on selection of which cancer pathologist (BVDV) reviewed the histology of all slides. Different tissue
tissue to be embedded in FFPE blocks. After the pathologist was finished with components (e.g., invasive carcinoma, carcinoma in situ, benign proliferative
macroscopic selection, additional tissue samples were embedded if high fluores- lesions, reactive lesions, and healthy parenchymal tissue including collagen and fat)
cence signals were detected in images of the fresh tissue slices in regions that were identified and delineated manually on the digitalized HE slides. Delineated
would not have been embedded for standard clinical care. The tissue cassette tissue components were exported as ROIs using Photoshop (Adobe Creative Cloud
numbers were marked on a printed photograph of all fresh tissue slices, to enable 2017). Fluorescence images of the 10-µm-thick sections as well as the ROIs con-
direct correlation between fluorescence signals in fresh tissue slice images and taining different tissue components were imported in ImageJ (Fiji, version 1.0). Per
histology. ROI a fluorescence measurement was performed resulting in an MFI per tissue
component per slide. Per patient a mean MFI was calculated per tissue component
and plotted in a graph (Graphpad Prism, version 7.0b).
Imaging procedures of formalin-fixed tissue. All FFPE blocks of all patients were
requested from the pathological department and were scanned with the Odyssey® The potential clinical value of fluorescence-guided surgery. Clinicopathological
CLX fluorescence flatbed scanning system. All FFPE blocks were scanned with the analyses of specimens were reported conforming standard clinical care, which
same imaging settings (wavelength: 800 nm, resolution 21 µm, quality: highest, contained at least macroscopical description, microscopical description including
intensity: 5). tumor type, modified Bloom-Richardson grade, surgical margins, and receptor
We made 10-µm-thick tissue sections of all FFPE blocks of all patients. The 10- status. If present, microscopic description of carcinoma in situ was reported
µm-thick sections were deparaffinized in xylene for two times five minutes each. It accordingly.
has been shown in an earlier clinical study executed by our group that dehydration All intraoperative fluorescence images and videos of the surgical cavity of all 26
or deparaffination in xylene steps has no effect on the presence of the compound, patients were reviewed by MK, a trained and experienced technical team member
and no effect on the measurements of the fluorescent signals (unpublished data and blinded for histopathology. The analyses of all the images took place after the
from clinical trial: Lamberts et al.)7. Thereafter, we left the slides to dry in the air in study was finished and all data were collected. Patients were divided on having
a dark environment. When dry, we imaged the slides using the Odyssey® CLX presence or absence of fluorescence signals in the surgical cavity. Presence of
fluorescence flatbed scanning system (LI-COR Biosciences Inc.) with the same fluorescence signals was defined as clear fluorescence signals that have higher
imaging settings to all slides (wavelength: 800 nm; resolution: 21 µm, quality: fluorescence intensities compared to lower fluorescence intensities from
highest, intensity: 8). After scanning the tissue slides, we directly performed background tissue, what means that high fluorescence signals could be easily
hematoxylin/eosin (H/E) staining to enable direct correlation between fluorescence delineated from lower background signals. To correlate fluorescence signals with
signal and histology on the same slide. H/E slides were digitalized using a digital having a tumor-involved surgical margin, a contingency table was analyzed. The
slide scanner (Hamamatsu, Japan). surgical margin was considered to be positive if ink was present on invasive cancer

10 NATURE COMMUNICATIONS | (2018)9:3739 | DOI: 10.1038/s41467-018-05727-y | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05727-y ARTICLE

or carcinoma in situ, according to the most recent SSO-ASCO guidelines on breast 17. Terwisscha van Scheltinga, A. G. T. et al. Intraoperative near-infrared
cancer9,21. fluorescence tumor imaging with vascular endothelial growth factor and
human epidermal growth factor receptor 2 targeting antibodies. J. Nucl. Med.
Statistical analysis. MFI was defined as total counts per ROI pixel area in tumor 52, 1778–1785 (2011).
and background normal tissues. Data were tested for Gaussian distribution; none of 18. Rosenthal, E. L. et al. Safety and tumor specificity of cetuximab-IRDye800 for
the data was normally distributed. Fluorescence signal intensities between four surgical navigation in head and neck cancer. Clin. Cancer Res. 21, 3658–3666
different dose groups were compared using Kruskal–Wallis test with a Dunn’s (2015).
multiple comparison test as post hoc analyses. Fluorescence signal intensities 19. Linardou, H. et al. The prognostic and predictive value of mRNA expression
between different tissue types within one dose group was analyzed with the of vascular endothelial growth factor family members in breast cancer: a study
Mann–Whitney U-test. Data are presented as boxplots with bars depicting mini- in primary tumors of high-risk early breast cancer patients participating in a
mum and maximum values, or median values were indicated with lines. Correla- randomized Hellenic Cooperative Oncology Group trial. Breast Cancer Res.
tion between fluorescence intensity and clinicopathological parameters was tested 14, R145 (2012).
with Spearman’s correlation test. A two-sided P value of less than 0.05 was con- 20. van Leeuwen van Zaane, F. et al. In vivo quantification of the scattering
sidered significant. We used GraphPad Prism, version 7.0b Software for statistical properties of tissue using multi-diameter single fiber reflectance spectroscopy.
analysis. Biomed. Opt. Express 4, 696–708 (2013).
21. Rosenberger, L. H. et al. Early adoption of the SSO-ASTRO Consensus
Data availability. The datasets generated during and/or analysed during the guidelines on margins for breast-conserving surgery with whole-breast
current study are available from the corresponding author on reasonable request. irradiation in Stage I and II invasive breast cancer: initial experience from
Memorial Sloan Kettering Cancer Center. Ann. Surg. Oncol. 23, 3239–3246
(2016).
Received: 14 February 2018 Accepted: 20 July 2018

Acknowledgements
Wytske Boersma-van Ek for her technical assistance. Emma Smeijers, Sharon Compeer,
and Oumaima Boutsa for their help in processing the FFPE tissue. Our physician
assistants Clara Lemstra and Arieke Prozee for helping recruiting the patients in the
References UMCG and medical doctor Tessa de Vries in the Martini Hospital. Pathological assistant
1. Zhang, R. R. et al. Beyond the margins: real-time detection of cancer using Lisette Jansen for implementing the workflow at the pathology department in the Martini
targeted fluorophores. Nat. Publ. Group 124, 171 (2017). hospital. The research leading to the results was supported by an unrestricted research
2. Tipirneni, K. E. et al. Oncologic procedures amenable to fluorescence-guided grant from SurgVision BV. M.K. and G.M.V.D. reports grants from the FP-7 Framework
surgery. Ann. Surg. 266, 36–47 (2017). Programme BetaCure grant no. 602812, during the study
3. Houssami, N., Macaskill, P., Marinovich, M. L. & Morrow, M. The association
of surgical margins and local recurrence in women with early-stage invasive Author contribution
breast cancer treated with breast-conserving therapy: a meta-analysis. Ann. M.K. and S.-Q.Q. designed the study, performed data acquisition, analyzed, and inter-
Surg. Oncol. 21, 717–730 (2014). preted data and drafted the manuscript. M.K. and S.-Q.Q. contributed equally to this
4. St John, E. R. et al. Diagnostic accuracy of intraoperative techniques for work. M.D.L. developed the GMP production of bevacizumab-800CW and critically
margin assessment in breast cancer surgery: a meta-analysis. Ann. Surg. 265, revised the manuscript. L.J., W.K., and J.D.V. designed the study, performed breast
300–310 (2017). cancer surgeries, and supervised the study. I.K. analyzed the pathology data and critically
5. Nagengast, W. B. et al. Near-infrared fluorescence molecular endoscopy detects revised the manuscript. G-J.Z. interpreted the data. D.J.R. assisted in spectroscopy data
dysplastic oesophageal lesions using topical and systemic tracer of vascular acquisition, performed the spectroscopy data analysis, and critically revised the manu-
endothelial growth factor A. https://doi.org/10.1136/gutjnl-2017-314953 (2017) script. W.B.N. obtained funding, designed the study, and interpreted the data. A.J-S.
6. Harlaar, N. J. et al. Molecular fluorescence-guided surgery of peritoneal supervised the development of GMP production of bevacizumab-800CW. B.V.D.V. was
carcinomatosis of colorectal origin: a single-centre feasibility study. Lancet involved in histopathological analyses and micro-segmentation analyses, critically revised
Gastroenterol. & Hepatol. 1, 283–290 (2016). the manuscript. G.M.V.D. obtained funding, designed and supervised the study, inter-
7. Lamberts, L. E. et al. Tumor-specific uptake of fluorescent bevacizumab- preted data, and drafted the manuscript. All authors reviewed the final manuscript.
IRDye800CW microdosing in patients with primary breast cancer: a phase I
feasibility study. Clin. Cancer Res. 0437, 2016 (2016).
8. van Dam, G. M. et al. Intraoperative tumor-specific fluorescence imaging in Additional information
ovarian cancer by folate receptor-α targeting: first in-human results. Nat. Med. Supplementary Information accompanies this paper at https://doi.org/10.1038/s41467-
17, 1315–1319 (2011). 018-05727-y.
9. Buchholz, T. A. et al. Margins for breast-conserving surgery with whole-breast
irradiation in stage I and II invasive breast cancer: American Society of Clinical Competing interests: G.M.V.D. is member of the scientific advisory board of SurgVision
Oncology endorsement of the Society of Surgical Oncology/American Society for BV. The remaining authors declare no competing interests.
Radiation Oncology consensus guideline. J. Clin. Oncol. 32, 1502–1506 (2014).
10. Ellis, L. M. & Hicklin, D. J. VEGF-targeted therapy: mechanisms of anti- Reprints and permission information is available online at http://npg.nature.com/
tumour activity. Nat. Rev. Cancer 8, 579–591 (2008). reprintsandpermissions/
11. Chhieng, D. C., Tabbara, S. O., Marley, E. F., Talley, L. I. & Frost, A. R.
Microvessel density and vascular endothelial growth factor expression in Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in
infiltrating lobular mammary carcinoma. Breast J. 9, 200–207 (2003). published maps and institutional affiliations.
12. Liu, Y. et al. The association between vascular endothelial growth factor
expression in invasive breast cancer and survival varies with intrinsic subtypes
and use of adjuvant systemic therapy: results from the Nurses’ Health Study.
Breast Cancer Res. Treat. 129, 175–184 (2011). Open Access This article is licensed under a Creative Commons
13. Rakha, E. A., Tun, M., Junainah, E., Ellis, I. O. & Green, A. Encapsulated Attribution 4.0 International License, which permits use, sharing,
papillary carcinoma of the breast: a study of invasion associated markers. J. adaptation, distribution and reproduction in any medium or format, as long as you give
Clin. Pathol. 65, 710–714 (2012). appropriate credit to the original author(s) and the source, provide a link to the Creative
14. Wang, Q. et al. Expression of vascular endothelial growth factor and caspase-3 Commons license, and indicate if changes were made. The images or other third party
in mucinous breast carcinoma and infiltrating ductal carcinoma-not otherwise material in this article are included in the article’s Creative Commons license, unless
specified, and the correlation with disease-free survival. Oncol. Lett. 14, indicated otherwise in a credit line to the material. If material is not included in the
4890–4896 (2017). article’s Creative Commons license and your intended use is not permitted by statutory
15. Hanahan, D. & Weinberg, R. A. Hallmarks of cancer: the next generation. Cell regulation or exceeds the permitted use, you will need to obtain permission directly from
144, 646–674 (2011). the copyright holder. To view a copy of this license, visit http://creativecommons.org/
16. Ter Weele, E. J. et al. Development, preclinical safety, formulation, and licenses/by/4.0/.
stability of clinical grade bevacizumab-800CW, a new near infrared
fluorescent imaging agent for first in human use. Eur. J. Pharm. Biopharm.
104, 226–234 (2016). © The Author(s) 2018

NATURE COMMUNICATIONS | (2018)9:3739 | DOI: 10.1038/s41467-018-05727-y | www.nature.com/naturecommunications 11