A Cautionary Tail: Changes in Integrin Behavior with Labeling

Genetic expression of fluorescently labeled proteins is essential to visualizing dynamic behavior within live cells. Recent advances in microscopy have increased resolution to the level where it is now possible to capture individual molecules interacting. However, the criteria for determining whether a fluorescent label perturbs protein function have not undergone a corresponding increase in resolution. The effects of protein labeling on cell function are still judged by whether populations of protein localize and interact with known binding partners. Here we use integrins, bidirectional signal adhesion molecules that regulate interactions between the extracellular matrix and the cytoskeleton through a well-defined series of conformational changes to show that not all labeling strategies are the same. We found that labeling the beta subunit decreased the mobility of individual integrin molecules and the protrusive activity of the entire cell. While integrins with labeled alpha subunits behaved similarly to unlabeled integrins, labeling the beta subunit increased the size of adhesions by elevating integrin affinity and exposing the ligand induced binding domain to change the molecule conformation. Thus, our single molecule and cellular data indicate that the ability of labeled proteins to localize and interact with known binding partners does not guarantee it does not alter protein function. We propose that the behaviors of individual molecules rather than the ensemble behavior of populations need to be considered as criteria to determine if a probe is non-perturbative.

Genetic expression of fluorescently labeled proteins is essential to visualizing dynamic behavior within live cells. Recent advances in microscopy have increased resolution to the level where it is now possible to capture individual molecules interacting. However, the criteria for determining whether a fluorescent label perturbs protein function have not undergone a corresponding increase in resolution. The effects of protein labeling on cell function are still judged by whether populations of protein localize and interact with known binding partners.
Here we use integrins, bidirectional signal adhesion molecules that regulate interactions between the extracellular matrix and the cytoskeleton through a well-defined series of conformational changes to show that not all labeling strategies are the same. We found that labeling the beta subunit decreased the mobility of individual integrin molecules and the protrusive activity of the entire cell. While integrins with labeled alpha subunits behaved similarly to unlabeled integrins, labeling the beta subunit increased the size of adhesions by elevating integrin affinity and exposing the ligand induced binding domain to change the molecule conformation. Thus, our single molecule and cellular data indicate that the ability of labeled proteins to localize and interact with known binding partners does not guarantee it does not alter protein function. We propose that the behaviors of individual molecules rather than the ensemble behavior of populations need to be considered as criteria to determine if a probe is non-perturbative.

INTRODUCTION
Genetic expression of fluorescently labeled proteins is essential for visualizing dynamic behavior within live cells. However, it is difficult to know whether labeling has an effect on cell function unless the label induces changes that are significantly larger than inherent cell-to-cell variability. Indeed, labeling is typically assumed to be non-perturbative if the proteins properly localize, respond to known stimulus, and interact with other proteins. Here we present data that challenges that assumption --molecular labeling strategies that do not inhibit proper localization or interactions, but do induce conformational changes that alter protein affinity for ligand.
We chose integrins as our model because integrins are bidirectional signaling molecules that mediate interactions with the matrix outside of the cell and with the cytoskeleton and signaling molecules inside of the cell by a well-defined series of conformational changes.
Integrins are alpha-beta heterodimers; both subunits are type I transmembrane glycoproteins with large extracellular domains, single spanning transmembrane domains, and short cytoplasmic domains. The extracellular domain is a large globular N-terminal binding head domain, while the transmembrane and cytoplasmic domains are legs or stalks that are severely bent at the knee in the inactive conformation and fully extended in the active conformation when bound to ligand 1,2 . As integrins transition from the inactive to active conformation, they become primed, which is a elevated affinity state where the receptor is not yet bound to ligand 3 . The transition to an elevated affinity state begins with separation of the cytoplasmic and transmembrane domains of the two subunits. As a direct consequence, the interface between the subunits in the tailpiece destabilizes, facilitating straightening of the legs 4 . Mn 2+ treatment induces extension of the integrin, but results in a mixture of open and closed headpieces, suggesting elevated affinity but incomplete activation. In contrast, ligand binding exclusively produces extended legs with open headpieces, suggesting complete activation 1,2 . These well-defined connections between conformational transitions and molecular behavior provide a welldefined system for studying the effects of labeling on molecular behaviors.
Extensive live-cell studies of GFP labeled integrins have shown that irrespective of whether the label is placed on the alpha or the beta subunit, integrins express on the cell surface and interact with their extracellular and intracellular binding partners. While these studies suggest that integrins are functional 5 , they have been limited to ensemble measures of populations, or measures of individual molecules sequestered into adhesion complexes 6 . We have recently shown that populations of individual unsequestered molecules reveal reproducible molecular behaviors that are obscured by ensemble measures. Therefore, we applied this approach to test whether labeling strategies altered molecular behaviors 7 . To our surprise, we found that molecular mobility as well as the affinity of integrins for matrix changed with labeling strategy. Our results suggest that localization and the ability to interact with binding partners are not adequate metrics to confirm that a label does not perturb protein functionunderstanding the effects of molecular labeling requires measuring behaviors of individual molecules.

Single-Molecule Mobility of Integrins Depends on which Subunit is Fluorescently Labeled
To determine whether labeling one subunit versus the other changes single molecule behavior we expressed both subunits of alpha V beta 3 integrins in CHO-K1 cells, which do not endogenously express either subunit 8 . We left one subunit unlabeled and labeled the other subunit with mEos2, a photoactivatible fluorophore. We then stochastically photo-converted the labeled subunits and used super-resolution microscopy to localize the position of the integrins in live cells from images collected at a frame rate of 40Hz (Supplemental movies S1 and S2).
Photo-converted integrins were tracked so that we could calculate the diffusion coefficient and classify molecular mobility behavior as confined, free, or directed diffusion 9,10 . We discovered that for all cells analyzed, the mobility of freely diffusing integrins was statistically lower when the expressed integrins had labeled beta subunits rather than labeled alpha subunits (Fig 1a, b).
Integrins confined within adhesion complexes had similar mobilities regardless of which subunit was labeled (Fig 1b, inset). These data indicate that labeling the beta subunit decreases the molecular mobility of the integrins that are not already interacting with either ligand or other proteins.

Protrusive Activity Changes When the Beta Subunit is Labeled
We then sought to determine if these decreases in integrin mobility that occurred when the beta subunit was labeled also changed dynamic cell behavior when the integrins were labeled with a non-photoactivatible, conventional fluorophore. We replaced the mEos2 labels with mEmerald and 24 hrs post-transfection collected time-lapse images of cells plated on fibronectin. We found that the leading edges of cells transfected with the labeled alpha subunit were consistently more dynamic than the leading edges of cells transfected with the labeled beta subunit (Fig 2a). The quiescence of the leading edge of the cells transfected with the beta subunit labeled is consistent with the notion that labeling the beta subunit increases the affinity of integrins for ligand, and it is also consistent with the more organized adhesions present in the cells transfected with labeled beta subunits (Fig 2b).

Adhesion Size Increases When the Beta Subunit is Labeled
We next investigated whether labeling the beta subunit would alter the size of adhesion complexes, macromolecular scaffolds that connect the cell to matrix and the cytoskeleton via linkages constructed by activated integrins 11 . In addition to transfecting cells with either labeled alpha or beta subunits, we also transfected cells with both subunits unlabeled. We then treated a separate group of the unlabeled integrins with 0.5 mM Mn 2+ to activate the integrins 11 . Cells transfected with unlabeled integrin were labeled with the human alpha V beta 3 specific LM609 antibody and Alexa 488. All cells were plated on fibronectin-coated coverslips for 3-5 hrs then fixed. We discovered that adhesion size was not statistically different if cells were transfected with unlabeled subunits or with labeled alpha V (Fig 3a). However, adhesions were consistently larger when cells were transfected with labeled beta 3 or when cells expressing unlabeled integrins were treated with Mn 2+ . This data, as well as the similarity in size between the beta labeled integrins and the unlabeled integrins treated with Mn 2+ , further support the interpretation that labeling the beta subunit is activating the integrin.

Whole Cell Response to Matrix Ligand Increases when Beta Subunit is Labeled
We also analyzed the effect of subunit labeling on the interaction of the whole cell with matrix by measuring the ability of cells to spread on matrix-coated surfaces. Cells were again transfected with unlabeled subunits, labeled alpha V, or labeled beta 3 integrin. Approximately 24 hrs post transfection, cells were trypsinized and separated into two groups. One group was treated with Mn 2+ , and both groups were allowed to spread on fibronectin for 30, 60, or 90 min prior to fixation. Treatment of the cells expressing unlabeled integrins or labeled alpha V with Mn 2+ increased cell spreading, indicating, as expected, that Mn 2+ activated these integrins. In contrast, treating cells expressing labeled beta 3 integrin with Mn 2+ did not produce any additional increase cell spreading (Fig 3b). Together these results indicate that labeling the alpha subunit does not activate the integrins, but labeling the beta integrin activates the integrin to a level that is similar to Mn 2+ treatment.

Labeling the Beta Subunit Exposes the Ligand Induced Binding Site
To directly test whether labeling increases integrin affinity for ECM, we expressed either unlabeled, alpha subunit labeled, or beta subunit labeled of another integrin, α 5 β 1 , in CHO-B2 cells, which lack endogenous alpha 5 integrin 12 . We then measured integrin affinity for ligand by quantifying the intensity of 9EG7, a commercially available β1 integrin antibody that detects the ligand-induced binding site (LIBS), which is exposed by Mn 2 treatment. Integrin affinity was not significantly different when both subunits were unlabeled or when only the alpha subunit was labeled. However, integrins with labeled beta subunits and unlabeled integrins treated with Mn 2+ both had elevated affinity levels and were not statistically different from each other (Fig   4a). These data demonstrate that labeling the beta subunit elevates the integrin affinity state and induces the same conformational changes as Mn 2+ .
Integrin conformational state is indicative of its ability to bind ligand and organize into adhesions. We found that integrins assembled from unlabeled subunits, labeled alpha subunits, or labeled beta subunits all properly localized to adhesion complexes, indicating that they are all functional. However, measuring the mobility of individual integrin molecules revealed that integrins with labeled beta subunits had a significantly lower diffusion coefficient than integrins with labeled alpha subunits outside of adhesion complexes. These changes at the single molecule level were reflected at the cellular level as less dynamic leading edges, larger adhesions, and larger surface areas in spreading assays compared to cells expressing labeled alpha subunits. These changes are functionally and statistically significant, but were not so morphologically abnormal that they would be identified as outside of normal cell variability.
However, by comparing these cells to cells expressing unlabeled integrins treated with Mn 2+ , our data suggested that labeling the beta subunit was activating the integrin (Fig 4b).
An earlier study noted that CHO-K1 cells expressing GFP-labeled alpha IIb (beta 3) or alpha IIb (GFP-labeled beta 3) were more likely to spontaneously aggregate in the presence of soluble fibrinogen, the ligand for alpha II (beta 3) integrin 5 . However, that study also noted that both labeling strategies did not inhibit the formation of adhesions and cell spreading. Therefore, they concluded that labeling either subunit was non-perturbative 5 . Our single molecule results, as well as our comparison of focal adhesion size and cell spreading in untreated cells and in cells with integrins activated by exposure to Mn 2+ all suggest that labeling the beta subunit changes the function of the expressed integrin by increasing its affinity for ligand.
Further evidence for the interpretation that labeling the beta subunit increases integrin affinity for ligand comes from quantification of the exposure of the LIBS domain in untreated cells and in cells exposed to Mn 2+ . Here our data suggests that labeling the beta subunit induced separation of the alpha and beta cytoplasmic tails to expose the LIBS domain, producing a response similar to Mn 2+ treatment. These results point to a cautionary tale -localization and the ability to interact with binding partners are not adequate to confirm that a label is truly non-perturbative. Our data suggest that functionality of molecules needs to be evaluated by measuring molecular behaviors and that localization and reorganization on the cellular level are insufficient metrics for guaranteeing that labeling does not alter molecular functionality.

CHO-K1 cells, which do not express endogenous alpha V or beta 3 integrins (ATCC)
and CHO-B2 cells, which do not express endogenous alpha 5 were grown in DMEM-F12 supplemented with 10% FBS. Cells were transfected with either human alpha V and beta 3 (CHO-K1) or alpha 5 and beta 1 (CHO-B2) using a Nucleofector II (Lonza) and Ingenio (Mirus) transfection reagents following manufacturer's protocols. Unlabeled integrins were in either pcDNA3.1 vectors (alpha V, beta 3, and alpha 5) or pRK5 (beta 1), and labeled vectors (mEos2 or Emerald) were constructed as previously described 7

Cell Spreading and Integrin Activation Assays
Approximately 24 hrs after transfection, cells were trypsinized and plated for 30, 60, or 90 min for adhesion assays or 4 hrs for spreading assays. Cells were then fixed with 2% paraformaldehyde in PHEM 13 . Cells that were transfected with unlabeled integrins were cotransfected with an empty Emerald vector for visualization of the cell perimeter. For activation, treatment with 0.05mM Mn 2+ was initiated 5 min prior to plating and was maintained throughout the spreading assay 14 . To detect adhesions in cells transfected with unlabeled integrins, cells were labeled with LM609 (Millipore) prior to secondary labeling with Alexa 488.
To quantify the amount of integrin activation transfected cells were plated overnight prior to fixation and labeling with 9EG7 (BD Pharmigen) prior to secondary labeling with Alexa 647.
9EG7 interacts with high affinity mouse and human beta 1 integrin 3 , binding the ligand-induced binding epitope exposed by treatment with Mn 2+15 , 9EG7 does not interact with endogenous hamster integrin. Activation with 0.05mM Mn 2+ was initiated 1 hr prior to fixation.

Microscopy
All imaging experiments were performed on an Olympus IX71 with a 60X 1.

Image processing, single molecule analysis, and statistics
The cell edge and adhesions were detected by thresholding images in Fiji 16 after smoothing with a 1 pixel Gaussian kernel sigma to reduce noise. Canny edge detection was used on the whole cell images after thresholding to obtain cell contours.
Single molecule analysis was performed using uTrack software 17 to localize and track individual mEos2 integrin molecules. Only molecules localized to better than 25nm precision were used for mobility analysis. Diffusion coefficients for tracks greater than 20 frames were analyzed as previously described and classified as either confined, freely diffusing or undergoing directed movement (i.e. drift) 7,10 . An average of between 6000 and 7000 molecules per cell was analyzed with a minimum of 6 cells per experimental group.
One-way ANOVA analysis was performed on all experimental groups. Sheffer post-hoc multiple comparison test was used to identify which treatments significantly differed from each other.
The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request.    Cells expressing labeled beta subunits did not increase their area in response to Mn 2+ treatment (p>0.5). N=20, 18, 27, 21, 24, 24 cells for unlabeled, unlabeled Mn 2+ , labeled-alpha, labeledalpha Mn 2+, labeled beta, and labeled-beta Mn 2+ , respectively. with 9EG7, an antibody that detects the conformational change that occurs when integrins are primed or activated and their affinity for ligand increases. Quantitative immunofluorescence detects a similar conformational change in CHO-B2 cells expressing labeled beta subunits or unlabeled subunits treated with Mn 2+ , but not in untreated cells expressing unlabeled or labeled alpha subunits. N=18, 17, 17, and 13 cells for unlabeled, alpha labeled, beta labeled, and unlabeled and treated with Mn 2+ , respectively. b) Cartoon illustrating the four-different integrin states: i) unlabeled integrin, ii) labeled alpha subunit, iii) labeled beta subunit, and iv) unlabeled integrins exposed to Mn 2+ .