Perfusion Window Chambers Enable Interventional Analyses of Tumor Microenvironments

Abstract Intravital microscopy (IVM) allows spatial and temporal imaging of different cell types in intact live tissue microenvironments. IVM has played a critical role in understanding cancer biology, invasion, metastases, and drug development. One considerable impediment to the field is the inability to interrogate the tumor microenvironment and its communication cascades during disease progression and therapeutic interventions. Here, a new implantable perfusion window chamber (PWC) is described that allows high‐fidelity in vivo microscopy, local administration of stains and drugs, and longitudinal sampling of tumor interstitial fluid. This study shows that the new PWC design allows cyclic multiplexed imaging in vivo, imaging of drug action, and sampling of tumor‐shed materials. The PWC will be broadly useful as a novel perturbable in vivo system for deciphering biology in complex microenvironments.

Polyjet matrix acrylic was most reliable, perfusable, and fully autoclavable.b, Microchannel appearance before and after wash protocol, shown when dry and when perfused with blue dye.c, Samples of manifold types tested.d, Percentage of channels perfusable pre-and post-cleaning.n=3 devices, two-way ANOVA with Fisher's LSD test, * p<0.05, ** p<0.01, *** p<0.001.e, Sample of connection types tested.f, pH of supernatant through cleaning protocol (n=3 parts, mean ±SD).g, Weight and cost (as quoted by suppliers at time of writing) of standard window chamber and of PWCs made of various materials.The chambers were sterilized in isopropanol and autoclaved before implantation.Scale bars: 5 mm.PS = polystyrene, SS = stainless steel, standard = standard titanium window chamber (nonperfusable).

Fig. S3. COMSOL modeling of different channel designs.
The overall objective of the PWC was to achieve uniform and robust filling through the chamber and absence of high/low pressure regions that could negatively affect tissue perfusion or chamber seal.Four microfluidic channel designs were simulated: a, 3-tier bifurcation, b, 2-tier bifurcation, c, single bifurcation with short channels, d, single bifurcation with long channels.The 3-tier design showed uniform velocity with minimal channeling of flow lines to support effective staining, and a gradual pressure drop between the inlet and chamber space that was contained within the device.These properties helped to maintain device stability during operation.Velocity and pressure legends are scaled to device in (a), with min and max velocities of 0.1-2 mm s -1 respectively at 10 mL hr -1 flow rate, and min and max pressures of 99-105 kPa.
Fig. S4: PWC leachate is non-toxic and non-chemotactic in vitro.Culture media was incubated for three days with PWCs to generate conditioned media.0%, 10%, 50%, and 100% conditioned media was used in cell culture with a, IMR90 fibroblast cells, and b, KPT tumor cells, to test for cytotoxicity of PWC leachate using the LDH assay.Cells were exposed to conditioned media for 3 days in vitro.Note that the leachate from PWC is not toxic to fibroblasts or tumor cells.n=3 replicates, with data normalized to average of 0% group, one-way ANOVA with Tukey multiple comparisons test, p<0.05, and mean ±SD shown.c, Boyden migration assay was used with freshly isolated mouse PBMCs to assess chemotactic effects when using 10% conditioned media as an approximation of chamber leachate levels.n=3 replicates repeated twice, with data normalized to average of 0% group, student's t-test, p<0.05, and mean ±SD shown.LDH: lactate dehydrogenase; PBMC: peripheral blood mononuclear cells.
Fig. S5: Re-usability of PWC.a, Three different PWC (chambers F, K, and L) were implanted into other mice, imaged, explanted, cleaned, re-sterilized, and then reimplanted.For example, chamber F was implanted into mouse 1, then mouse 4, then mouse 7, then mouse 10.MC38-H2B-apple cells were implanted in mice 2 and 3 to allow comparison between channels.Note that the PWC function properly over time, can be repeatedly re-sterilized, and image quality does not degrade.b-c, Parts were also serially washed in a 37˚C bath and autoclaved to monitor part integrity over time.b, Before and after images of PWC and backings after autoclaving.c, Number of PWC and backings that remain perfusable (PWC) and intact (backing) with full integrity after each autoclaving cycle.The PWC is connected to dual syringe pumps, one to push liquid into the chamber and one to remove it.a, Low-resolution serial images of PWC with MC38 TagBFP2 cells in blue.The perfusate (dextran-rhodamine) is shown over time at the fastest 20 mL hr -1 flow rate.Note the rapid kinetics of perfusion throughout the entire chamber.The solution was switched to PBS at 0.9 mL elapsed for each rate, corresponding to 2.7 minutes for the 20 mL hr -1 rate, 3.6 min at 15 mL hr -1 , 5.4 min at 10 mL hr -1 , and 10.8 min at 5 mL hr -1 .For additional detail, see Movie S and 2. b, Chamber fluorescence graphed as a function of time for different pump settings showing perfusion kinetics.With the highest pump setting of 20 mL hr -1 , the PWC can be thoroughly perfused and cleared within five minutes.At the lowest perfusion rate tested (5 mL hr -1 ), the entire cycle takes approx.20 minutes.To determine whether serial staining could be performed in situ through the perfusion chamber, we tested a recently developed bioorthogonal method of staining followed by cleaving off antibody labels. [36]a, Preperfusion image showed no background.b, Chamber perfusion using an anti-Ly6G-AF647 antibody stained superficial neutrophils in part of the chamber.c, Following administration of the Hk-Tz scissors, [36] S1: PWC engineering design factors that were explored to achieve the primary objective of stably transferring liquid in and out of the window chamber while keeping the system self-contained.• Stably supports ring placement

Backing thickness
• Minimize weight • Minimize overall construct thickness when implanted

Backing angle
• Minimizes pressure on body • Keeps chamber straight

Range of solutions tested Outcome
Coverglass opening size • Supports coverglass easy removal • Minimizes gap and space for leakage (i.e., must be snug)

Fig. S1 :
Fig. S1: 3D-printed microfluidic perfusable window chamber (PWC) enables fluid flow through mouse dorsal skin-fold window chamber for staining, drug testing, and TME sampling.a, Standard static titanium dorsal skin-fold window chamber.b, CAD model semitransparent and 2D outline views of the PWC and its flow path.c, Images of 8-channel PWC with blue dye perfused through microchannels and the assembled final version.d, Image of PWC installed in vivo with the corresponding multichannel confocal image obtained from it.Scale bar: 50 µm.

Fig. S2 :
Fig. S2: Design of PWC.a, Sample of materials tested.Polyjet matrix acrylic was most reliable, perfusable, and fully autoclavable.b, Microchannel appearance before and after wash protocol, shown when dry and when perfused with blue dye.c, Samples of manifold types tested.d, Percentage of channels perfusable pre-and post-cleaning.n=3 devices, two-way ANOVA with Fisher's LSD test, * p<0.05, ** p<0.01, *** p<0.001.e, Sample of connection types tested.f, pH of supernatant through cleaning protocol (n=3 parts, mean ±SD).g, Weight and cost (as quoted by suppliers at time of writing) of standard window chamber and of PWCs made of various materials.The chambers were sterilized in isopropanol and autoclaved before implantation.Scale bars: 5 mm.PS = polystyrene, SS = stainless steel, standard = standard titanium window chamber (nonperfusable).

Fig. S6 :
Fig. S6: The PWC allows both confocal and multiphoton imaging.B16-F10-H2B-mApple or MC38-TagBFP2 fluorescent tumor cells were implanted into PWCs on day zero.7 to 10 days after tumor injection, mice (C57BL/6J or mertK-GFP) were imaged using intravenous administration of fluorescent dextrans or GSL I (lectin) as a vascular label.Both a, confocal and b, multiphoton imaging were performed through the window chamber in live animals for both tumor models.Note the excellent correlation between the two imaging modalities.Scale bars: 100 µm.

Fig. S7 :
Fig. S7: Selective image acquisition through PWC without bleed-through.MC38-TagBFP2 fluorescent tumor cells were implanted into six C57BL/6J mice and imaged 7-10 days later.Mice received aliquots of anti-Ly6G antibody to label neutrophils.The antibody was labeled with either AF647, MB488, or AF555.Images are displayed at identical window-level settings.Images were acquired sequentially with four lasers, except one laser was turned off in each row to determine the bleed-through signal.a, No 405 nm laser results in tumor cells being non-excited.b, No 663 nm laser with lack of Ly6G-AF647 excitation.c, No 559 nm laser with lack of Ly6G-AF555 excitation.d, No 473 nm laser with lack of Ly6G-MB488 excitation.AF: AlexaFluor.Scale bars: 50 µm.

Fig. S8 :
Fig. S8: Perfusion kinetics.The PWC is connected to dual syringe pumps, one to push liquid into the chamber and one to remove it.a, Low-resolution serial images of PWC with MC38 TagBFP2 cells in blue.The perfusate (dextran-rhodamine) is shown over time at the fastest 20 mL hr -1 flow rate.Note the rapid kinetics of perfusion throughout the entire chamber.The solution was switched to PBS at 0.9 mL elapsed for each rate, corresponding to 2.7 minutes for the 20 mL hr -1 rate, 3.6 min at 15 mL hr -1 , 5.4 min at 10 mL hr -1 , and 10.8 min at 5 mL hr -1 .For additional detail, see Movie S and 2. b, Chamber fluorescence graphed as a function of time for different pump settings showing perfusion kinetics.With the highest pump setting of 20 mL hr -1 , the PWC can be thoroughly perfused and cleared within five minutes.At the lowest perfusion rate tested (5 mL hr -1 ), the entire cycle takes approx.20 minutes.

Fig. S9 :
Fig. S9: Systemic vs. PWC antibody labeling.To determine whether antibodies perfused through the PWC behaved differently from standard workflows, we compared systemic and PWC-administered labeling.a, anti-Ly6G-AF647 administered systemically 24 hours before imaging identified neutrophils in the TME.b-c, The TME was perfused with two different anti-Ly6G antibodies labeled with SAFE-MB488 (green) and SAFE-AF555 (red).Note the good co-localization between the two PWC antibodies.d, There was very good co-localization between the systemically administered antibody from (a) and antibodies given by the perfusion window chamber in (b) and (c).e, SNR of cells in three different channels (n=3 experiments, mean shown).f-h, Pearson's correlation plots for co-localization of antibody pairs.Scale bar: 50 µm.TME: tumor microenvironment.SNR: signal-to-noise ratio.

Fig. S10 :
Fig. S10: Confocal depth imaging following IV or PWC staining of tumoral neutrophils with anti-Ly6G.a, anti-Ly6G-AF647 was administered systemically 24 hours before imaging and revealed neutrophils in the TME.b-c, The TME was perfused with two different anti-Ly6G antibodies labeled with MB488 (green) and AF555 (red).d, Merge channel.Note the good colocalization between IV-and PWC-administered antibody down to 84 µm.Blue: MC38-TagBFP2 is shown only in the merge channel.IV: intravenous.TME: tumor microenvironment.AF: AlexaFluor.Scale bar: 50 µm.

Fig. S11 :
Fig.S11: In situ staining and cleaving.To determine whether serial staining could be performed in situ through the perfusion chamber, we tested a recently developed bioorthogonal method of staining followed by cleaving off antibody labels.[36]a, Preperfusion image showed no background.b, Chamber perfusion using an anti-Ly6G-AF647 antibody stained superficial neutrophils in part of the chamber.c, Following administration of the Hk-Tz scissors,[36]  the signal returns to near background.d, Restaining of the same section with a new anti-Ly6G antibody.e, Second round of cleaving.f, Quantification of fluorescent signal over time for the stain and cleave cycles for multiple cell ROI's.Since neutrophils are mobile, quantification was done on cohorts of cells rather than the same cell.n=3 experiments, one way ANOVA with Tukey multiple comparison test, p<0.05, with mean ±SD shown.ROI: region of interest.AU: arbitrary unit.Scale bar: 50 µm.
Fig.S11: In situ staining and cleaving.To determine whether serial staining could be performed in situ through the perfusion chamber, we tested a recently developed bioorthogonal method of staining followed by cleaving off antibody labels.[36]a, Preperfusion image showed no background.b, Chamber perfusion using an anti-Ly6G-AF647 antibody stained superficial neutrophils in part of the chamber.c, Following administration of the Hk-Tz scissors,[36]  the signal returns to near background.d, Restaining of the same section with a new anti-Ly6G antibody.e, Second round of cleaving.f, Quantification of fluorescent signal over time for the stain and cleave cycles for multiple cell ROI's.Since neutrophils are mobile, quantification was done on cohorts of cells rather than the same cell.n=3 experiments, one way ANOVA with Tukey multiple comparison test, p<0.05, with mean ±SD shown.ROI: region of interest.AU: arbitrary unit.Scale bar: 50 µm.

Fig. S12 :
Fig. S12: Extracellular vesicle sampling from TME.To determine whether the tumor microenvironment could be safely sampled for biomarker analysis, PWC implanted with MC38 tumors were perfused with PBS to collect effluent.Samples were processed to isolate tissue-derived EV, then stained for CD9 and CD63, typical tetraspanin markers present in EV. a, Experimental setup for TME sampling.b, Sample images of EV from PWC effluent labeled with the pan-EV TFP (left), CD9-FITC (middle), and CD63-AF647 (right).Insets show negative controls (no EV, antibody only).Circled objects indicate EVs identified in the TFP channel (left) which co-labeled with the respective marker (middle, right).Small numbers of vesicles were identified within the TME, as expected.EV: extracellular vesicles.TME: tumor microenvironment.Scale bar: 500 nm.

Table S3 :
Antibodies and stains.Antibodies were diluted to 2-4 µg/mL for imaging applications unless otherwise specified.