Early detection and staging of chronic liver diseases with a protein MRI contrast agent

Early diagnosis and noninvasive detection of liver fibrosis and its heterogeneity remain as major unmet medical needs for stopping further disease progression toward severe clinical consequences. Here we report a collagen type I targeting protein-based contrast agent (ProCA32.collagen1) with strong collagen I affinity. ProCA32.collagen1 possesses high relaxivities per particle (r1 and r2) at both 1.4 and 7.0 T, which enables the robust detection of early-stage (Ishak stage 3 of 6) liver fibrosis and nonalcoholic steatohepatitis (Ishak stage 1 of 6 or 1 A Mild) in animal models via dual contrast modes. ProCA32.collagen1 also demonstrates vasculature changes associated with intrahepatic angiogenesis and portal hypertension during late-stage fibrosis, and heterogeneity via serial molecular imaging. ProCA32.collagen1 mitigates metal toxicity due to lower dosage and strong resistance to transmetallation and unprecedented metal selectivity for Gd3+ over physiological metal ions with strong translational potential in facilitating effective treatment to halt further chronic liver disease progression.


100-101
See above, 90-91 101 114-117 check "raveled" The measured hydration number of 0.5 for the Tb-containing system may not be transferable to Gdcontaining protein

116-117
A unique and consistent terminology should be defined and used throughout the paper 125 This sentence is not correct, since ProCA32.collagen1 is not among clinical GBCAs. Rewrite as "…ProCA32.collagen1 has a greater metal selectivity for Gd3+ than clinical gadolinium-based contrast agents."

128-129
Actually from Fig 1f one can see that the main protein band is decreasing and possibly some lower MW bands appear. The image should reproduce the whole migration lanes, in order to allow the reader to assess the stability of the main product.

130-134
These sentences cannot be evaluated without Supplementary materials and related study methods.

135-136
The comment of possible reduced dosage does not pertain to the table of Fig. 1e. Possibly to the Relaxivity table (1d), but it is not clear on which data the ratio is calculated 141 Was the dose of Eovist also 0.02 mmol/kg? The standard dose, after correction for body surface, is 0.3 mmol/kg. This dose would compensate for the lower relaxivity of Eovist and yield a more meaningful comparison.

330-332
From previous description (see [105][106], it seemed that ProCA32-P40 contains PEG, differing only in the lack of the collagen binding peptide. A clearer compound identification should be provided. The contribution of PEG is expected to influence tR, more than second sphere water molecules [380][381] See above 130-134. Comparison with other agents should imply the same administration protocol and analytical methods

382
"molecular biomarkers" too generic, better "collagen" How was the antibody prepared? Is a rabbit antiserum?
502-3 The COL-1 antibody is from mouse. The corresponding secondary Ab is not reported Reviewer #2 (Remarks to the Author): This work describes a novel imaging approach with an innovative MRI contrast agent to visualize fibrosis in mouse models.
1. I have a series of concerns regarding the animal models of NASH/ASH/fibrosis used in this manuscript. Throughout the manuscript, it remains quite blurry, which models are presented ("early fibrosis", "late fibrosis"). The figure legends need to be clear on these aspects (e.g., figure 2). Moreover, it is simply not true that a single injection of DEN in c57bl/6 mice would induce liver cirrhosis after 10 months. This is a non-fibrotic / non-cirrhotic model of HCC. If the authors observe cirrhosis, we are likely looking at artifacts related to tumor development (e.g., pathogenic angiogenesis, leading to Sirius Red stained areas). Same is true for "NASH diet induced cirrhosis"this will simply not happen after 6 months of Western diet. changes in a quantitative manner. A main conclusion is that the change in MR relaxation rates (pre to post injection) at 24hrs allows the ability to distinguish normal, early stage, and late stage fibrosis, while an earlier time point (3hrs post-injection) is able reveal facets of angiogenesis and portal hypertension, indicative of more advanced disease.
Results and Claims: In my opinion, there is an excess of data and claims presented in this manuscript, which distracts from the overall objective and my understanding of the essential findings. I believe it is longer than it needs to be, and can be written more concisely. Several claims are given in the manuscript that point to novel methods for detecting and staging fibrosis, but it is never clear which finding is preferred or advantageous in a clinical setting. I feel there needs to be a more specific and in-depth analysis of 1-2 primary claims the authors feel most strongly about. Moreover, from the manuscript, it is difficult to piece together a singular procedure for using ProCA32collagen1 and an MR method in clinical application. Ultimately, my question is still, "Once injected, what single MR method should I perform, and when? And what primary data analysis method should I perform to best elucidate fibrosis stage?" If this still needs investigation, you should state the need for further work.
Pharmacokinetics of ProCA32collagen1: You should consider adding more pharmacokinetic data into the main document, and not in supplemental data. For a new agent, I think it's crucial to discuss distribution kinetics, elimination half-life, and clearance, in addition to relaxivity, stability, and safety. I'm curious to know the short time-scale wash-in/wash-out characteristics of this agent in both normal and fibrotic tissue environments. Given a high R1 for late stage fibrosis and low R1 for normals at 3hrs, does this imply faster wash-in/wash-out for normals? How long will ProCA32collagen1 bind to collagen markers in fibrotic livers before being eliminated?
Pathology and Histology: The pathology work in this study is extensive and well-done. I am curious whether you analyzed any gross liver specimens from the mouse models. I think a valuable qualitative and quantitative comparison would be fibrosis seen on gross specimens versus slicematched post-injection (24hrs) MRI R1 and R2 maps. This would truly validate contrast agent uptake, distribution, and wash-out in fibrotic and remote liver regions. I feel biopsy and histology proves the presence of fibrosis degree, but it is limited in emphasizing the geographic match between pathology and MR imaging data.
Time points: MR imaging of ProCA32collagen1 distribution was performed at 3, 24, and 48hrs, along with baseline measurements. Please state the logic for this procedure, especially in light of the agent's half-life (~9hrs). Even though the use of animal models may preclude imaging at earlier time points, almost all current dynamic MR imaging occurs < 1hr post-injection. The use of the word "dynamic" (e.g DMI and DCE) should be tempered when considering 3, 24, and 48hrs post-injection time-points; this word implies imaging over a short time scale, usually over minutes. The authors should at least address the possibility of missing the peak signal or max R1 time points, which may affect AUC and other results presented in Fig 3 and 4.
Methods: Some descriptions of experimental methods contain explicit data results. For example, the "statistical analysis" section describes the AUC results. Other method descriptions contain results shown in supplemental data figures. This may be ok, but I leave it to the Editors to determine whether these should be referenced in the text. In fact, the supplemental data contain lots of methods descriptions not referred to in main document.
From the histogram analysis of R1, R2, delta R1, and delta R2, there seems to be large variance in the images (Fig 3h, Fig 4e), yet very small error bars on plots. It is a bit unclear whether mean and standard deviation at each time point represents the entire imaging slice or volume. Further, there is no mention of a segmentation algorithm.
P2.L44 -consider improving the word usage of the statement beginning with "which" P2.L48 -not all methods are based on liver stiffness.
P2.L50 -PDFF is not used for fibrosis detection. It's a fat quantification technique P2.L51 -many of these techniques do provide information about disease heterogeneity.
P2.L52 -you do not need to re-state "non-invasive" as a pre-requisite for a non-invasive tool. The pre-requisite is accurate and precise staging, to allow detection and monitoring of disease and therapy response (which implies frequent and safe studies) P2.L54 -Not sure what is meant by "deep tissue penetration". While high resolution can be achieved with MRI, it is lower than CT, so not really a clear advantage. Same goes with "full liver coverage".
P2.L55 -improved "detection" is not a direct consequence of MRI being non-ionizing.
P2.L57 -not only because of "limitation of currently available contrast agents". There are other reasons too.
P3.L63 -That Gd only has positive (T1) contrast is not entirely true. The same contrast agents are used for dynamic susceptibility weighted (T2*) contrast (negative contrast). It merely depends on the pulse sequence used.
P3.L65 -Needs more clarity. Most Gd agents distribute to the liver, albeit extracellularly. Moreover, relaxivity is static for a given field strength and tissue, so an increase in this parameter is not the cause of bright T1 signal. Also, "short arterial and short hep uptake" is confusing.
P3.L71 -are the authors proposing a dual imaging protocol for fibrosis detection? I think it is sufficient to propose the exploration of both imaging mechanisms of this protein contrast agent. But how do you know which is better?

Results
P4.L86 -Judging by the references in this section, much of the methodological description of design seems based on prior publications, not an original result.
P4.L87 -stability and blood retention time is not referenced in Fig 1d. P5.L105 -once again, this result seems based on a prior publication (ref 27,28), not original to this paper.
P5.L108 -what is the difference between the expressions "per Gd" vs "per particle"?
P6.L139 -Is detection based on heightened relaxivity or collagen binding affinity? I would think the latter.
P6.L144 -was mouse fibrosis uniform throughout liver? If not, there may be an internal control of fibrosis degree.
P10.L212 -was distribution measures and pathology results taken at 24 or 48hrs post-injection? Please state.

P11
.L222 -what does AUC of signal enhancement signify?
P11.L227 -ROC analysis result: shouldn't this be placed in the previous section, "Robust detection of early and late stages liver fibrosis and NASH with dual contrast property"? P11.L237 -why does early fibrosis R1 increase at 24hr relative to 3hr (while late fibrosis decreases)? What's the rationale or mechanism? Are you possibly missing the peak R1 change of early fibrosis?
P11.L240 -are these 3hr findings specific to this time point only, or even earlier time points as well? Please describe the mechanism of these results (possibly in the discussion section) in terms of the pharmacokinetics of the agent. This "suggestion" from these results is not clear to me. Figure 3f -the existence of portal hypertension in cirrhotic livers is common, and not related to the "injection of ProCA32.collagen1". I can not piece together how the presence of a contrast agent bound to collagen in liver is indicative to the pressure of blood flow through the portal vein. P14.L269 -I don't follow how the "targeting ability" of the contrast agent reflects the existence of portal hypertension. Is it purely based on slow wash out? Then it's not really due to the targeting ability.
P14.L270 -there are several claims in this sentence. You may want to be more concise, otherwise it gets confusing. Is it the time-dependent R1 contrast profiles that distinguish fibrosis stage, or delta R1 (and R2) at 24hrs?
P14.L272 -Suggesting the contrast agent reveals patterns in cirrhotic liver may be true, but it's a qualitative assessment, which may require a separate investigation. It is valid to speculate on the distribution differences in each disease model in the discussion section.
P14.L278 -Do all mice in the DEN group exhibit this difference in CPA (12% vs. 6.7%)? Is this considered mild, moderate, or severe fibrosis?
P14.L279 -Why was T1, T2, and T1 IR performed here, and not R1 and R2 mapping (consistent with earlier analysis)?  P16.L311 -please revise this sentence; it is unclear. There are many uses of the word "or" that makes the claims confusing.
P16.L313 -why was this analysis of enhancement over time only performed with the DEN model, and not other mouse models. Figure 4fg -what do the two color bars on one image signify? Are you showing both 3 and 24hrs in one image? Also, the colorbar associated with "maintained" and "washout" is unclear… dark red indicates >100% relative enhancement, yet it's defined as "maintained"? The color matrix is somewhat unclear to me.
P17.L316 -from Fig4c, it seems T2w has higher sensitivity for heterogeneity (difference b/w leftright segments?) than IR. Please define how you are determining sensitivity. How are you quantifying which method (T1 or T2 or IR) is more "sensitive"? Discussion P17.L323 -What were the data results for correlation with histology, CPA, and Gd3+?
P17.L325 -Is it due to 5x increase in relaxivity, or collagen binding affinity? I would think the latter, since the mechanism seems to be whether the CA is present in liver tissue or not, right?
P17.L326 -IR + long TE may need to be elaborated. Depending on inversion time, IR will enhance T1 differences, while long TE enhances T2 differences.
P17.L327 -It is not clear how the use of enhanced r2 properties, coupled with various imaging techniques (IR, T1 and T2-weighted) overcome small changes in "liver morphology" P17.L328 -explain how "precision" is "doubled". Precision of what? Does this mean some coefficient of variation metric is two-times lower by using this agent?
P17.L333 -Are these statements "likely due to…" an effect? Are they not definite?
P17.L336 -What is meant by "tissue penetration"? I don't see correlation values mentioned between CPA and delta R1 and R2. Is correlation based of visual inspection only?
P17.P339 -It is important to precisely define "dynamic molecular imaging (DMI)". It's not an intuitive concept. You mention high "spatial and temporal resolution", but what is considered high for DMI? 0, 3, 24, 48hrs seem very spaced out time points. It will be important to fully discuss how "early" (i.e. +3hrs) time points point to angiogenesis. Where is contrast being "retained" in this scenario vs. normal liver, and why is that indicative to new vessel formation.
P18.L350 -is your claim that a "dramatic" increase in R2 and R2 at 3hrs indicative of portal hypertension? Why? What increase is considered "dramatic"? Presumably, ProCA32collagen1 concentration in liver is high at 3hrs (vs. normal and early-stage). Doesn't this mean that contrast agent wash-in and retention (collagen affinity) is greater in cirrhotics vs. early-stage? Do you suspect a more gradual CA uptake in early stage disease? Also, please explain why R1 and R2 appear to continue increasing in early-stage from 3 to 24 hrs.
P18.L352 -This is a one sentence paragraph and seems out of place.
P18.L358 -It sounds like you're distinguishing DCE from DMI. Please explicitly state the difference, since it's not obvious.
P18.L362 -The connection between improved "penetration capability" and the formation of "new vascular structures" is unclear. The agent binds to collagen, so how does this point to vascular information?
P18.L367 -Please state/disclose these precise pattern differences in the Results, and discuss them here. Are they consistent? Is there a metric to quantitatively determine pattern differences?
P19.L370 -This is a run-on sentence. Also, are you claiming that AUC_0-48 of both R1 and R2 is a robust metric to stage fibrosis? What are the threshold values these need to be to distinguish normal, early and late stage? Are these better than delta R1 and R2 at 24hrs for staging fibrosis?
P19.L376 -Some institutions have migrated from MultiHance to Prohance or Dotarem. Eovist is usually used in specialized cases (low overall usage), and is fairly stable.

Methods
P24L509 -Is CPA calculated over the entire liver, individual slices, or just one segment? In other words, how big was the CPA sample? What was the CPA for the various mouse models?
P25.L523 -Please comment on the choice of inversion times. Do you feel there is enough T1 sampling resolution? Your results show that your R1 range is ~1.5 to ~4s-1 (or T1~250ms to 666ms), which means the ideal null point is around 170 to 460ms. Also, were the inversion times acquired in separate acquisitions? Spin echo or gradient echo? What was the scan time? Similar questions for T2 mapping.
P25.L529 -What model was used for curve fitting T1 and T2? Was it performed pixel by pixel?
P25.L534 -What is meant by "intensity MRI images"? Are these the R1 and R2 maps? If so, these are somewhat quantitative, not qualitative.
P25.L536 -Not sure I follow. Are you saying each image voxel has a min and max R1 value?
P26.L542 -I assume T1, T2, and IR are different acquisitions than R1 and R2 mapping. What were the parameters? So, if I understand, at each time point (0, 3, 24, 48 hrs), T1, T2, IR, R1-mapping, and R2-mapping was acquired? Please state how many scans were performed at each time point, and how long each scan was?
P26.L544 -awkward sentence ("To make sure the enhanced area remained enhanced or the contrast agent is washed out…") P26.L547 -Not sure if you can conclude this; it depends on imaging technique. For T2-weighting, more negative contrast may indicate more T2 effect, hence greater agent concentration. A negative T1 contrast may also indicate increase concentration, since a T2 effect dominates. I can presume you may not have to worry about the latter case, since delta R2 were not extremely high, but you may want to acknowledge that this is possible.
P26.L549 -It's ok to formulate a unique analysis like this. But since it's challenging for readers to grasp initially, please summarize the overall results concisely (in the Results). What significance does it have?
P26.L563 -How were Eovist and ProCA32collagen dilutions prepared for ex vivo measurements? In saline, plasma, serum? How was it different than the 1.4T preparation?
P27.L583 -this stats section only deals with AUC analysis. Please re-state section heading.
P27.L584 -PSE was already defined and stated earlier in Methods (line 543).
P28.L591 -Most of the statements in this section are results, not methods.
P28.L598 -Were the values from an ROI of the entire liver? Please state ROI size and any segmentation routines (automatic or manual).
P28.L600 -some analysis was paired, correct? Such as R1 time course within the same animal model.
P28.L603 -why weren't the pharmacokinetic results included in the main document?
P28.L605 -how many sampling points? What is meant by n=3-6? Figure S4 - Figure S1. The ratio is in fact 2:1 for Gd-ProCA32.collagen1 complex and this mistake was corrected in the manuscript (line 92). Metals in ProCA32.collagen1 were removed by chelex-100 and metal content in ProCA32.collagen1 was analyzed by ICP-OES. The protein concentration was analyzed by UV-Vis spectrometry by monitoring Trp signal and obtaining the extinction coefficient from the protein sequence (line 445-447).  Fig. 1d-1e, where PEG-ProCA32.collagen1 appears to be different from ProCA32.collagen1. Neither ProCA32-P40 is clearly described.

Supplemental Figures
Response: As suggested, a consistent terminology was used throughout the paper. Wherever ProCA32 and ProCA32.collagen1 names are mentioned, they are referring to PEGylated contrast agents on Lysine residues. If "ProCA32.collagen1" is not PEGylated on Lysine residues it is referred to as "Non-PEGylated ProCA32.collagen1". These names are edited in figures legends and text for further clarifications.
Specific Comment 9: 96 Check "Ominiscan" instead of Omniscan (also in Fig.1c Response: All experiments in this manuscript was performed with 2:1 Gd-ProCA32.collagen1 complex except the transmetallation studies which were performed with a 1:1 Gd-ProCA32.collagen1 complex. This ratio was used to demonstrate a more meaningful comparison with clinical contrast agents since in all of them Gd 3+ is loaded in a 1:1 ratio with chelators. However, as suggested we performed the transmetallation experiment with 2:1 Gd-ProCA32.collagen1 complex as well and the results are shown in Fig. 1c. Due to cooperative binding, there is no difference in affinity between two binding sites.

Specific Comment 12:
The measured hydration number of 0.5 for the Tb-containing system may not be transferable to Gd-containing protein

Response:
We have determined water coordination number in ProCA32.collagen1 by Terbium Lifetime Luminescence. The number of coordination water molecule in the inner sphere of Gd 3+ -ProCA32.collagen1 was determined by the difference in Tb 3+ luminescence decay between H2O and D2O. This system was used since both Gd 3+ and Tb 3+ have very similar coordination chemistry, and the Tb 3+ system is commonly used for determining the water hydration number in clinical contrast agents such as Gd-DTPA and Gd 3+ chelators. Below are the relevant, associated references including our publication used to measure the water number:  Fig. 3a-d where the response to ProCA32.collagen1 is significantly higher. Therefore a comparison is not possible

Response:
The purpose of testing Eovist in animals with Ishak =1 or =3, was to demonstrate the inability of Eovist to detect early stage liver fibrosis and NASH and these data were compared with animal with Ishak =1 or =3 that were injected with ProCA32.collagen1.

Specific Comment 23: 328 "precision is doubled" where is it shown?
Response: To avoid any confusion, the sentence was removed from the manuscript. It was meant to show that using multiple imaging sequences provide multiple ways to eliminate MRI artifacts associated with heterogeneous liver background (i.e. false positives) and increase accuracy (line 359).
Specific Comment 24: 330-332 From previous description (see [105][106], it seemed that ProCA32-P40 contains PEG, differing only in the lack of the collagen binding peptide. A clearer compound identification should be provided. The contribution of PEG is expected to influence tR, more than second sphere water molecules Response: In order to avoid any confusion and for clarify, we used "ProCA32" instead of "ProCA32-P40" in the manuscript. As suggested, we modified the manuscript to reflect the contribution of PEG in influencing τR, rather than second sphere water molecules (line 332).

Response:
We indeed applied the same administration protocol of tail vein injection and analytical methods such as organ digestion and ICP-OES to measure the Gd 3+ concentration, biodistribution and organ deposition for both ProCA32.collagen1 and Eovist. Different metal binding assays were required to determine metal binding affinity due to special nature of proteins and small chelators since classic pH potentiometric method is not applicable to protein, ProCA32.collagen1 Also, because pH titration is not suitable for determining Kd between proteins and metals, we were not able to compare them at the same conditions. Also, due to the lack of fluorescence signal, the Tbbuffer system cannot be applied to small chelators to determine binding constants. Gd-Tb competition can also not be applied to get the affinity of Gd 3+ for small molecules.
However, we applied the same transmetallation method to directly compare their kinetic stability against endogenous zinc replacement or "inertness" (See Fig 1c). Clearly, ProCA32.collagen1 has superior metal stability and selectivity for Gd 3+ over Zn 2+ that is consistent with our metal binding assay.

Specific Comment 27: 384 See above: the dosage in humans is speculative
Response: As suggested, the data regarding the projected dosage of ProCA32.collagen1 in humans were removed from the manuscript.

Specific Comment 29: 404 The TAA dose (not concentration) is not well defined and not consistent with above 200 mg/kg
Response: The TAA dose used was 200 mg/kg and this was corrected in the manuscript (line 453).

Specific Comment 30: 500 How was the antibody prepared? Is a rabbit antiserum?
Response: The self-generated antibody was purified from rabbit antiserum using PEGylated-ProCA32 as the immunogen.

Specific Comment 31: 502-3 The COL-1 antibody is from mouse. The corresponding secondary Ab is not reported
Response: Goat anti-Mouse IgG1 Cross-Adsorbed Alexa Fluor 488 was used as secondary antibody. The information was added to the "Methods" section, "Immunofluorescence Imaging" (line 560-562).

Reviewer #2
This work describes a novel imaging approach with an innovative MRI contrast agent to visualize fibrosis in mouse models. 7

Specific Comment 1: I have a series of concerns regarding the animal models of NASH/ASH/fibrosis used in this manuscript. Throughout the manuscript, it remains quite blurry, which models are presented ("early fibrosis", "late fibrosis"). The figure legends need to be clear on these aspects (e.g., figure 2). Moreover, it is simply not true that a single injection of DEN in c57bl/6 mice would induce liver cirrhosis after 10 months. This is a non-fibrotic / non-cirrhotic model of HCC.
If the authors observe cirrhosis, we are likely looking at artifacts related to tumor development (e.g., pathogenic angiogenesis, leading to Sirius Red stained areas). Same is true for "NASH diet induced cirrhosis"this will simply not happen after 6 months of Western diet.
Response: This comment has several sub comments which are addressed point-by point below: I have a series of concerns regarding the animal models of NASH/ASH/fibrosis used in this manuscript.
Reponses: No details are mentioned by the Reviewer to justify concern for the ASH model.
Throughout the manuscript, it remains quite blurry, which models are presented ("early fibrosis", "late fibrosis"). The figure legends need to be clear on these aspects (e.g., figure 2).

Response:
We have carefully rearranged figures and the legends in each figure were rewritten and shortened to clarify the data presented in each figure.

Moreover, it is simply not true that a single injection of DEN in c57bl/6 mice would induce liver cirrhosis after 10 months. This is a non-fibrotic / non-cirrhotic model of HCC. If the authors observe cirrhosis, we are likely looking at artifacts related to tumor development (e.g., pathogenic angiogenesis, leading to Sirius Red stained areas).
Response: We defined "Cirrhosis" as "a diffuse process characterized by fibrosis and the conversion of normal liver architecture into structurally abnormal nodules that affect the whole organ based on the original published definitions. Fibrosis is defined as the presence of excess collagen due to new fiber formation that causes only minor clinical symptoms or disturbance of liver cell function". Thus, throughout the manuscript, we include histological analysis of collagen content to verify the degree of fibrosis in every animal model we used (Fig. 2f, 3f, 3g, 5b). References below used for these definitions: The main purpose here is to use a model to mimic cirrhosis conditions and associated collagen heterogeneity. To provide further clarity, we have replaced the term "cirrhosis" by "collagen heterogeneity" in the related section) in the revised manuscript for the DEN-induced model (line 249. For the DEN mice model, our histological results (Fig. 5b, Fig. I) show that there is extensive accumulation and deposition of collagen in the tumor region of the tissue and the entire heterogeneous distribution of collagen in the liver can be detected by our developed contrast agent, which is a key application in field of hepatology. We have also provided the entire scanned slide in Fig I which highlights the intra tumor collagen accumulation shown in red by Sirius red stain.
In addition, the Sirius Red stains are typical collagen stains (typical crossing-networking pattern), not artifact of pathological angiogenesis that has no crossing-networking stain pattern.

Same is true for "NASH diet induced cirrhosis"this will simply not happen after 6 months of Western diet.
Response: It is not true that NASH diet was used to induce cirrhosis in normal, wild type mice. Instead, we have clearly stated in the "Methods" section (line 472), that liver-specific Comparative Gene Identification-58 (CGI-58) knocked out (LivKO) mice were used which develop severe hepatic cirrhosis even on chow diet (reference 1) and the NASH diet facilitates the development of hepatic fibrosis to develop cirrhosis conditions. In previous publications (Feng Guo et al, Journal of Lipid Research, Volume 54, 2013;J. Mark Brown et al, Journal of Lipid Research Volume 51, 2010), it has been clearly demonstrated that CGI-58 deficiency in the liver directly causes not only hepatic steatosis but also steatohepatitis and fibrosis in mice. Our histological results (Fig. 3g)

Response:
The whole purpose of creating this MRI contrast agent is to develop future translational applications in humans, given the high translational potential of MRI. In this manuscript, we have provided extensive results to support potential human applicability of our developed contrast agents especially for reduced metal toxicity (Fig.1e, 1c, 1f, Supplemental 7d), reduced dosage, and biodistribution (Supplemental 5a). To address one of the major causes of nephrogenic systemic fibrosis (NSF) and brain deposition associated with free Gd 3+ toxicity, we determined that the Gd 3+ binding affinity of ProCA32.collagen1 is comparable to the approved clinical contrast agents (Fig. 1e). ProCA32.collagen1 also exhibited 10 14 -to 10 16 -fold increases in metal selectivity (kinetic stability) for Gd 3+ over Ca 2+ and Zn 2+ compared with all clinically approved contrast agents. To the best of our knowledge, ProCA32.collagen1 has the greatest metal selectivity among all other clinically approved Gd 3+ -based contrast agents. In addition, the required injection dose is much lower than any clinical contrast agents.
We have cited and discussed references  in our manuscript to support human translation potential of collagen targeting. Collagen expression has been observed in several chronic human patient diseases including NASH and HCC ( As suggested, human HCC tissue samples were used (Fig II) and stained both with our contrast agent and Sirius red to demonstrate high collagen expression and ProCA32.collagen1 binding to collagen in human HCC samples. These results demonstrated the specificity of our contrast agent to bind to collagen in human HCC tissue arrays and the applicability of our contrast agent in humans. This figure has been added to our manuscript as Fig. 7c.

Specific Comment 3:
The true challenge in clinical trials is monitoring treatment efficacy noninvasively. There is a large series of effective antifibrotic treatments in NAFLD under development that we effective in mouse models (e.g., elafibranor, selonsertib, cenicriviroc). The authors need to demonstrate that the imaging can be performed longitudinally and is capable of monitoring fibrosis regression (or reduced fibrosis development) in mouse models. Figure II. Demonstration of three Hepatocellular carcinoma (HCC) and normal human liver tissue microarray stained with Sirius red (red) and ProCA32.collagen1 (brown) exhibiting the strong binding of the contrast agent to human collagen. In this procedure, the contrast agent was incubated with the ProCA32.collagen1 antibody (1:2 ratio) and then the preincubated mixture was added into the slides. Patient #1 and #2 are showing adjacent slides and Patient #3 is the control.

Red
ProCA32.collagen1 Response: Our results in Fig. III in this cover letter show that our developed contrast agent is in fact capable of monitoring liver fibrosis regression and treatment. In order to test the efficacy response, cirrhotic mice generated from TAA/Alcohol model were treated with pirfenidone which has been tested previously (Oleksii Seniutkin et al, Toxicol Appl Pharmacol. 2018 Jan 15;339:1-9), and the mice were scanned before and after injection with the contrast agent, and the results were compared with normal mice.
Our data suggest that ProCA32.collagen1 can distinguish pirfenidone-treated liver from normal liver. We did not include these results in the manuscript since our goal here was to provide extensive results to demonstrate the novel capability of developed contrast agents for the "early detection and staging of liver diseases". This is the essential step prerequisite to human clinical trials, to monitor treatment efficacy. In addition, we have shown that ProCA32.collagen1 can detect different stages of the disease in several models as you discussed here.

Response:
We have provided extensive assessment of fibrosis stage using Sirius Red, IHC staining and αSMA for each model (Fig 2f, 2j,

Reviewer #3
Comment 1: Please be advised that the manuscript ("Article") format does not abide by Nature recommendations: -Abstract >150 words -Overall Text >3000 words -Methods >3000 words -References > 50 -Large number of sub-panels per figure (as many as 10 sub-panels occur in figure 3). Legends are very long >500 words

Response:
We have completely revised and rewritten some parts of the manuscript, in order to emphasize on the key findings and improve readability. Heterogeneous MR signals coming from fibrosis liver and tissue background are the major challenge and cause of low specificity of MRI with current clinical approved contrast agents in morphology detection. In addition to lack of disease molecular biomarker detection, all approved Gd 3+ contrast agents have very small r1 and r2 relaxation values. Only r1 contrast is able to provide some image enhancement in vivo. Thus, all clinical imaging method is built only on one single imaging technique, which leads to low 14 specificity in differentiation of fibrosis vs tissue background and inability to detect early-stage fibrosis.
Our developed contrast agent has several unique properties including dual relaxation and collagen binding. One of the major advantages of using our developed contrast agent, is that you can use multiple imaging pulse sequences and imaging methodology with advantages in early detection and staging fibrosis in vivo instead of one single method. By using only one single imaging technique, there is a high chance of imaging artifacts due to heterogeneous MR signal coming from liver background, the use of multiple imaging techniques provide complementary ways to validate and confirm the stage of the disease with significantly improved sensitivity and specificity. We have explicitly mentioned in our manuscript that both the T1 and T2 mapping 24 hrs time point post-injection of the contrast agent is used to stage the disease (normal vs early vs late) while the 3 h time point provides additional information such as portal hypertension and angiogenesis associated with late stage liver cirrhosis. Furthermore, we have shown a novel dynamic molecular imaging (DMI) methodology, taking into consideration the entire imaging response curve of T1 and T2 mapping, which provides the highest sensitivity and specificity for early detection and staging of fibrosis. Response: Based on your suggestion, we have added the pharmacokinetic data in the main text (see Fig. 1f and 1g). The distribution of the contrast agent, elimination half-life, and clearance have all been reported in the manuscript in the supplemental information section: "Clearance of ProCA32.collagen1 was low at 0.36 mL/min/kg but more than 3 times higher than ProCA32-P40, and with very high exposure (>100000 ng.h/mL blood). ProCA32.collagen1 exhibited a terminal elimination half-life of 9.9 h, which was slightly higher than ProCA32-P40 (8.09 h), with a mean residence time of 14.5 h that was higher than 13.9 h of ProCA32-P40. In addition, volume of distribution (Vc) and volume of distribution at steady state (Vdss) were 1.53 and 1.77 L/kg, respectively, which were more than 2-7 times higher than ProCA32".
We have performed more statistical analysis to demonstrate the uptake and washout rate of our contrast agent in both normal and different stages of fibrotic liver based on MR mapping data, and since R1 and R2 values will reflect the concentration of Gd 3+ , they can be a good indication of uptake and washout of the contrast agent. Fig. IV (top) shows the overall uptake characteristics of this agent (0-48 hrs) in both normal and fibrotic tissue environments using our MRI data and percentage of signal enhancement in R1 map. Fig. IV (bottom) demonstrates the washout rate of the contrast agent between 3-48 hrs post-injection. The washout rate is: normal liver > early stage fibrosis > late stage cirrhosis. Uptake is as follows: late stage cirrhosis > early stage fibrosis > normal liver. We have added these data to our manuscript (see Fig 6c, and 6d). Based on AUC analysis in Fig. 6, after 48 hrs the majority of the contrast agent is washed out. Response: It is important to note that ProCA32.collagen1 has much greater size than clinically approved contrast agents of small chelators with much longer half-life in blood (~10hrs). Thus, we performed the MRI scans at longer time points (longer than 3 hrs) since the half-life in liver tissues is likely to be longer than in the blood.
Dynamic Contrast Enhanced (DCE) imaging measures MRI signal changes such as T1 changes in tissues over time after bolus administration of gadolinium contrast agent. Dynamic refers to the changes as a function of time. It does not imply "short time". All clinical contrast agents that are based on small molecules (<1 kD in size) have much shorter half-lives (<10 min), and that is why the MR imaging occurs < 1 hrs post-injection for other MR imaging contrast agents. The short half-life of clinically approved contrast agents results in limitations in missing the peak signal and large errors in AUC calculation. Since our contrast agent has a half-life of ~9 hrs and has collagen binding capability, thus we tailored the acquisition time points to capture the entire curve without missing the peak signal or max R1 time points. We purposely use the new term "dynamic molecular imaging" (DMI) to avoid the confusion with DCE and highlight the targeting ability and collagen binding of our contrast agent with a new mechanism. Once the contrast agent is injected, liver will be enhanced along with the blood vessels due to the biodistribution of ProCA32.collagen1 in the blood vessel and sinusoid space in the liver. ProCA32.collagen1 gradually binds to collagen in the liver of diseased mice, which in turn prolong the half-time of contrast agent in the liver. Non-specifically distributed ProCA32.collagen1 was eliminated and ProCA32.collagen1 remained in the liver regions with overexpression of collagen.
This new dynamic process enhanced by our targeted contrast agent enables us to visualize differential washout rates based on the stage of the disease (see Fig. 6). This is consistent with heterogeneous MRI signal enhancement within 24 h post injection in Fig. 5d and 5e.

Response:
We have modified this sentence to avoid any confusion. There might be several reasons for this high collagen affinity. Addition of a flexible hinge (GGG linker) gives enough freedom to the targeting moiety to bind to collagen (increase binding capacity).

Specific Comment 66: P17.L336 -What is meant by "tissue penetration"? I don't see correlation values mentioned between CPA and delta R1 and R2. Is correlation based of visual inspection only?
Response: Collagen proportional area (CPA) values described in Fig. S9 were used to generate Fig. 6e and 6f along with ΔR1 and ΔR2 values. Fig. 2c, Fig. 3c, Fig. S8b Response: Please see our Response to Comment 5. We define this new MRI methodology as dynamic molecular imaging (DMI) to emphasize our novel features. We used the term "dynamic molecular imaging", simply to convey the point that the contrast agent property is different from other contrast agents, however the word "dynamic" was used because the contrast agent enhancement is different at different time points and has different washout rates based on the stage of the disease (see Fig. 6c, 6d). Furthermore, Fig. 5d  Response: R1 and R2 increase at 3 hrs time point were higher compared to 24 hrs time point mainly due to slow washout originating from decrease in number of fenestrae in liver for late stage of fibrosis as well as collagen binding (Fig 4e, 4f, 6c, 6d). Please also see our Response for Comment 43). The accumulation of contrast agent is predominant for the late stage of fibrosis. However, due to lack of structural changes in fenestrae numbers and size in early stage of fibrosis, we do not have accumulation of protein contrast agent. The slow and gradual increase of R1 and R2 is largely due to the contribution of collagen binding process. The observed enhancement at 24 hours represents the expression level of collagen associated with a disease stage.

Specific Comment 70:
P18.L352 -This is a one sentence paragraph and seems out of place.

Response:
We have modified and moved this sentence to "Introduction" (line 49-52). Response: Dynamic Contrast Enhanced (DCE) imaging measures MRI signal changes such as T1 changes in tissues over time after bolus administration of gadolinium contrast agent. Dynamic refers to the changes as a function of time. It does not imply "short time". All clinical contrast agents that are based on small molecules (<1 kD in size) have much shorter half-livers (<10 min), and that is why the MR imaging occurs < 1hr post-injection for other MR imaging contrast agents. Their short half-life of clinically approved contrast agents results in limitations in missing the peak signal and large errors in AUC calculation. We did not use the word "DCE" for our contrast agent, and DCE is usually used for clinical contrast agents and shorter time points, but our contrast agent mechanism is clearly different from clinical contrast agents. Since our contrast agent has a halflife of ~9 hours and has collagen binding capability, thus we tailored the acquisition time points to capture the entire curve without missing the peak signal or max R1 time points. ProCA32.collagen1 binds to collagen in the liver, which extends its retention in the liver with much longer half-life in the liver compared with that of blood vessel. Once the contrast agent is injected, liver will be enhanced along with the blood vessels due to the biodistribution of ProCA32.collagen1 in the blood vessel and sinusoid space in the liver. ProCA32.collagen1 gradually binds to collagen in the liver of diseased mice, which extends/elongates the half-time of contrast agent in the liver. Non-specifically distributed ProCA32.collagen1 is eliminated and ProCA32.collagen1 remains in the liver regions with overexpression of collagen.

Specific
We defined this new property of the contrast agent as dynamic molecular imaging (DMI) to emphasize our novel features. We used the term "dynamic molecular imaging", simply to convey the point that the contrast agent property is different from other contrast agents, however the word "dynamic" still can be used because the contrast agent enhancement is different at different time points and has different washout rates based on the stage of the disease (see Fig. 4, 6). Furthermore, Fig. 5dand  Response: Intrahepatic angiogenesis involves the formation of new vessels and vascular structures with and without complete connection (see Fig. 4g). Such new vascular structural change is packed by overexpress collagen. The significant increase of R1 and R2 at 3-hour time point is a result of accumulation of high concentration of contrast agent due to slow wash out and altered vascular structure as well as collagen binding.