Limits...
Image artifacts in single molecule localization microscopy: why optimization of sample preparation protocols matters.

Whelan DR, Bell TD - Sci Rep (2015)

Bottom Line: As a result of the up to an order-of-magnitude improvement in spatial resolution, substantially more detail is observed, including changes in distribution and ultrastructure caused by the many steps required to fix, permeabilize, and stain a sample.We present three well-optimized fixation protocols for staining microtubules, mitochondria and actin in a mammalian cell line and then discuss various artifacts in relation to images obtained from samples prepared using the protocols.The potential for such errors to go undetected in SMLM images and the complications in defining a 'good' image using previous parameters applied to confocal microscopy are also discussed.

View Article: PubMed Central - PubMed

Affiliation: School of Chemistry, Monash University, Clayton, Victoria 3800, Australia.

ABSTRACT
Single molecule localization microscopy (SMLM) techniques allow for sub-diffraction imaging with spatial resolutions better than 10 nm reported. Much has been discussed relating to different variations of SMLM and all-inclusive microscopes can now be purchased, removing the need for in-house software or hardware development. However, little discussion has occurred examining the reliability and quality of the images being produced, as well as the potential for overlooked preparative artifacts. As a result of the up to an order-of-magnitude improvement in spatial resolution, substantially more detail is observed, including changes in distribution and ultrastructure caused by the many steps required to fix, permeabilize, and stain a sample. Here we systematically investigate many of these steps including different fixatives, fixative concentration, permeabilization concentration and timing, antibody concentration, and buffering. We present three well-optimized fixation protocols for staining microtubules, mitochondria and actin in a mammalian cell line and then discuss various artifacts in relation to images obtained from samples prepared using the protocols. The potential for such errors to go undetected in SMLM images and the complications in defining a 'good' image using previous parameters applied to confocal microscopy are also discussed.

No MeSH data available.


Related in: MedlinePlus

Varying fixative and antibody concentration affects the apparent clustering distribution of import receptors on the mitochondrial membrane.(A–I) COS-7 cells stained using Alexa Fluor 647 primary/secondary antibodies against the Tom20 protein subunit of the import receptor on the outer mitochondria membrane. Cells were stained with a high (A–C, 1:50), mid-range (D–F, 1:500) or low (G–I, 1:2000) concentration of primary and secondary antibodies. Cells were fixed using the optimized paraformaldehyde protocol (A, D, G), a mixed 3% paraformaldehyde, 0.5% glutaraldehyde protocol (B, E, H) or the optimized glutaraldehyde protocol (C, F, I). Scale bar: 1 μm.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4300460&req=5

f6: Varying fixative and antibody concentration affects the apparent clustering distribution of import receptors on the mitochondrial membrane.(A–I) COS-7 cells stained using Alexa Fluor 647 primary/secondary antibodies against the Tom20 protein subunit of the import receptor on the outer mitochondria membrane. Cells were stained with a high (A–C, 1:50), mid-range (D–F, 1:500) or low (G–I, 1:2000) concentration of primary and secondary antibodies. Cells were fixed using the optimized paraformaldehyde protocol (A, D, G), a mixed 3% paraformaldehyde, 0.5% glutaraldehyde protocol (B, E, H) or the optimized glutaraldehyde protocol (C, F, I). Scale bar: 1 μm.

Mentions: Our observations of significant changes in fluorophore distribution and apparent structure in response to permeabilization and antibody dilution prompted us to investigate the impact that similar changes in preparation would have on MC structure. As in Figure 1, the mitochondrial membrane import receptor was imaged by staining the Tom20 subunit after fixation with various protocols. In Figure 6 the first column shows cells fixed using the optimized PFA protocol, the second column shows cells fixed using a mixed PFA/GA (3%/0.5%) (SI Methods 6) protocol, and the right column shows cells fixed according to the optimized GA protocol. The three rows show cells stained with a relatively high concentration of primary and secondary antibody (1:50, Top, Fig 6a–c), a mid-range concentration of antibody (1:500, Fig 6d–f), and a low concentration of antibody (1:2000, Fig 6g–i). Cells in the top and middle rows have not been quenched or blocked whereas those depicted in the bottom row (6g–i) were quenched using 1% NaBH4 and blocked in 3% BSA. Differences in clustering size and distribution are immediately apparent across antibody concentrations as well as fixatives; indeed, upon visual examination differences in all nine protocols can be seen. This is further demonstrated with basic cluster analysis which was performed by scanning the dSTORM images for clustered fluorophores forming features larger than 2 × 2 10 nm pixels (400 nm2) and smaller than 30 × 30 10 nm pixels (90,000 nm2) and then measuring the size of these clusters. Extensive preprocessing of the data was not necessary, nor was the definition of regions of interest, in order to acquire reasonable estimates of average cluster size. This was found to range from 2020–4330 nm2 corresponding to a range of average diameters (assuming circular arrangements) from 51–74 nm across all fixation and immunostaining conditions tested. The PFA/GA fixed preparation with 1:500 dilution of antibodies showing the smallest clusters (Figure 6e) and the 3.7% PFA fixation with 1:50 dilution of antibodies showing the largest clusters (Figure 6a) (Results summarized SI Table 1). In some cases, this analysis returned clusters larger than previous work which estimated them to be 30–40 nm in diameter. This is a consequence of both the enlargement of the Tom20 clusters by antibodies and fluorophores and the simplicity of the analysis performed. It should not be taken as truly quantitative due to the minimal preprocessing, however the large overall variation in mean size is in good agreement with visual examination of the images and confirms the role of fixation artifacts in the final images. In Figure 6 it can be seen that PFA fixation gives clear clustering at all three antibody concentrations but cluster size decreases as antibody concentration decreases. The mixed PFA/GA fixation yields what appears to be a lower degree of clustering with a more homogeneous distribution of localized dyes irrespective of antibody concentration. GA fixation shows a different pattern of clustering once more as well as some indication of shrinkage of the overall MC structure. This demonstration that slight changes in sample preparation—all of which yielded essentially visually identical epifluorescence images—can alter the sub-diffraction distributions and structures in relation to the known native state23 is extremely important because in many cases the sub-diffraction structure is unknown prior to SMLM. In the case of the Tom20 subunit of the import receptor, previous work using a PFA-based protocol similar to that depicted in Fig. 6d) has detected and quantified this clustering23. Without the reference point of this previously conducted work it would be more difficult to substantiate any claim that one protocol was more biologically accurate compared to the others. The fact that nine different levels of clustering were observed from nine variations of the protocol once again stresses the importance of determining which resembles the true biological native state most closely and to identify the causes of any artifacts.


Image artifacts in single molecule localization microscopy: why optimization of sample preparation protocols matters.

Whelan DR, Bell TD - Sci Rep (2015)

Varying fixative and antibody concentration affects the apparent clustering distribution of import receptors on the mitochondrial membrane.(A–I) COS-7 cells stained using Alexa Fluor 647 primary/secondary antibodies against the Tom20 protein subunit of the import receptor on the outer mitochondria membrane. Cells were stained with a high (A–C, 1:50), mid-range (D–F, 1:500) or low (G–I, 1:2000) concentration of primary and secondary antibodies. Cells were fixed using the optimized paraformaldehyde protocol (A, D, G), a mixed 3% paraformaldehyde, 0.5% glutaraldehyde protocol (B, E, H) or the optimized glutaraldehyde protocol (C, F, I). Scale bar: 1 μm.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4300460&req=5

f6: Varying fixative and antibody concentration affects the apparent clustering distribution of import receptors on the mitochondrial membrane.(A–I) COS-7 cells stained using Alexa Fluor 647 primary/secondary antibodies against the Tom20 protein subunit of the import receptor on the outer mitochondria membrane. Cells were stained with a high (A–C, 1:50), mid-range (D–F, 1:500) or low (G–I, 1:2000) concentration of primary and secondary antibodies. Cells were fixed using the optimized paraformaldehyde protocol (A, D, G), a mixed 3% paraformaldehyde, 0.5% glutaraldehyde protocol (B, E, H) or the optimized glutaraldehyde protocol (C, F, I). Scale bar: 1 μm.
Mentions: Our observations of significant changes in fluorophore distribution and apparent structure in response to permeabilization and antibody dilution prompted us to investigate the impact that similar changes in preparation would have on MC structure. As in Figure 1, the mitochondrial membrane import receptor was imaged by staining the Tom20 subunit after fixation with various protocols. In Figure 6 the first column shows cells fixed using the optimized PFA protocol, the second column shows cells fixed using a mixed PFA/GA (3%/0.5%) (SI Methods 6) protocol, and the right column shows cells fixed according to the optimized GA protocol. The three rows show cells stained with a relatively high concentration of primary and secondary antibody (1:50, Top, Fig 6a–c), a mid-range concentration of antibody (1:500, Fig 6d–f), and a low concentration of antibody (1:2000, Fig 6g–i). Cells in the top and middle rows have not been quenched or blocked whereas those depicted in the bottom row (6g–i) were quenched using 1% NaBH4 and blocked in 3% BSA. Differences in clustering size and distribution are immediately apparent across antibody concentrations as well as fixatives; indeed, upon visual examination differences in all nine protocols can be seen. This is further demonstrated with basic cluster analysis which was performed by scanning the dSTORM images for clustered fluorophores forming features larger than 2 × 2 10 nm pixels (400 nm2) and smaller than 30 × 30 10 nm pixels (90,000 nm2) and then measuring the size of these clusters. Extensive preprocessing of the data was not necessary, nor was the definition of regions of interest, in order to acquire reasonable estimates of average cluster size. This was found to range from 2020–4330 nm2 corresponding to a range of average diameters (assuming circular arrangements) from 51–74 nm across all fixation and immunostaining conditions tested. The PFA/GA fixed preparation with 1:500 dilution of antibodies showing the smallest clusters (Figure 6e) and the 3.7% PFA fixation with 1:50 dilution of antibodies showing the largest clusters (Figure 6a) (Results summarized SI Table 1). In some cases, this analysis returned clusters larger than previous work which estimated them to be 30–40 nm in diameter. This is a consequence of both the enlargement of the Tom20 clusters by antibodies and fluorophores and the simplicity of the analysis performed. It should not be taken as truly quantitative due to the minimal preprocessing, however the large overall variation in mean size is in good agreement with visual examination of the images and confirms the role of fixation artifacts in the final images. In Figure 6 it can be seen that PFA fixation gives clear clustering at all three antibody concentrations but cluster size decreases as antibody concentration decreases. The mixed PFA/GA fixation yields what appears to be a lower degree of clustering with a more homogeneous distribution of localized dyes irrespective of antibody concentration. GA fixation shows a different pattern of clustering once more as well as some indication of shrinkage of the overall MC structure. This demonstration that slight changes in sample preparation—all of which yielded essentially visually identical epifluorescence images—can alter the sub-diffraction distributions and structures in relation to the known native state23 is extremely important because in many cases the sub-diffraction structure is unknown prior to SMLM. In the case of the Tom20 subunit of the import receptor, previous work using a PFA-based protocol similar to that depicted in Fig. 6d) has detected and quantified this clustering23. Without the reference point of this previously conducted work it would be more difficult to substantiate any claim that one protocol was more biologically accurate compared to the others. The fact that nine different levels of clustering were observed from nine variations of the protocol once again stresses the importance of determining which resembles the true biological native state most closely and to identify the causes of any artifacts.

Bottom Line: As a result of the up to an order-of-magnitude improvement in spatial resolution, substantially more detail is observed, including changes in distribution and ultrastructure caused by the many steps required to fix, permeabilize, and stain a sample.We present three well-optimized fixation protocols for staining microtubules, mitochondria and actin in a mammalian cell line and then discuss various artifacts in relation to images obtained from samples prepared using the protocols.The potential for such errors to go undetected in SMLM images and the complications in defining a 'good' image using previous parameters applied to confocal microscopy are also discussed.

View Article: PubMed Central - PubMed

Affiliation: School of Chemistry, Monash University, Clayton, Victoria 3800, Australia.

ABSTRACT
Single molecule localization microscopy (SMLM) techniques allow for sub-diffraction imaging with spatial resolutions better than 10 nm reported. Much has been discussed relating to different variations of SMLM and all-inclusive microscopes can now be purchased, removing the need for in-house software or hardware development. However, little discussion has occurred examining the reliability and quality of the images being produced, as well as the potential for overlooked preparative artifacts. As a result of the up to an order-of-magnitude improvement in spatial resolution, substantially more detail is observed, including changes in distribution and ultrastructure caused by the many steps required to fix, permeabilize, and stain a sample. Here we systematically investigate many of these steps including different fixatives, fixative concentration, permeabilization concentration and timing, antibody concentration, and buffering. We present three well-optimized fixation protocols for staining microtubules, mitochondria and actin in a mammalian cell line and then discuss various artifacts in relation to images obtained from samples prepared using the protocols. The potential for such errors to go undetected in SMLM images and the complications in defining a 'good' image using previous parameters applied to confocal microscopy are also discussed.

No MeSH data available.


Related in: MedlinePlus