Review Topical Sections

Silk nanoparticles—an emerging anticancer nanomedicine

  • Silk is a sustainable and ecologically friendly biopolymer with a robust clinical track record in humans for load bearing applications, in part due to its excellent mechanical properties and biocompatibility. Our ability to take bottom-up and top-down approaches for the generation of silk (inspired) biopolymers has been critical in supporting the evolution of silk materials and formats, including silk nanoparticles for drug delivery. Silk nanoparticles are emerging as interesting contenders for drug delivery and are well placed to advance the nanomedicine field. This review covers the use of Bombyx mori and recombinant silks as an anticancer nanomedicine, highlighting the emerging trends and developments as well as critically assessing the current opportunities and challenges by providing a context specific assessment of this multidisciplinary field.

    Citation: F. Philipp Seib. Silk nanoparticles—an emerging anticancer nanomedicine[J]. AIMS Bioengineering, 2017, 4(2): 239-258. doi: 10.3934/bioeng.2017.2.239

    Related Papers:

    [1] Haiyan Wang, Nao Yamamoto . Using a partial differential equation with Google Mobility data to predict COVID-19 in Arizona. Mathematical Biosciences and Engineering, 2020, 17(5): 4891-4904. doi: 10.3934/mbe.2020266
    [2] Adil Yousif, Awad Ali . The impact of intervention strategies and prevention measurements for controlling COVID-19 outbreak in Saudi Arabia. Mathematical Biosciences and Engineering, 2020, 17(6): 8123-8137. doi: 10.3934/mbe.2020412
    [3] Lei Nie, Xin Guo, Chengqi Yi, Ruojia Wang . Analyzing the effects of public interventions on reducing public gatherings in China during the COVID-19 epidemic via mobile terminals positioning data. Mathematical Biosciences and Engineering, 2020, 17(5): 4875-4890. doi: 10.3934/mbe.2020265
    [4] Enahoro A. Iboi, Oluwaseun Sharomi, Calistus N. Ngonghala, Abba B. Gumel . Mathematical modeling and analysis of COVID-19 pandemic in Nigeria. Mathematical Biosciences and Engineering, 2020, 17(6): 7192-7220. doi: 10.3934/mbe.2020369
    [5] Fernando Saldaña, Hugo Flores-Arguedas, José Ariel Camacho-Gutiérrez, Ignacio Barradas . Modeling the transmission dynamics and the impact of the control interventions for the COVID-19 epidemic outbreak. Mathematical Biosciences and Engineering, 2020, 17(4): 4165-4183. doi: 10.3934/mbe.2020231
    [6] Fang Wang, Lianying Cao, Xiaoji Song . Mathematical modeling of mutated COVID-19 transmission with quarantine, isolation and vaccination. Mathematical Biosciences and Engineering, 2022, 19(8): 8035-8056. doi: 10.3934/mbe.2022376
    [7] M Kumaresan, M Senthil Kumar, Nehal Muthukumar . Analysis of mobility based COVID-19 epidemic model using Federated Multitask Learning. Mathematical Biosciences and Engineering, 2022, 19(10): 9983-10005. doi: 10.3934/mbe.2022466
    [8] Sarah R. Al-Dawsari, Khalaf S. Sultan . Modeling of daily confirmed Saudi COVID-19 cases using inverted exponential regression. Mathematical Biosciences and Engineering, 2021, 18(3): 2303-2330. doi: 10.3934/mbe.2021117
    [9] Wentao Hu, Yufeng Shi, Cuixia Chen, Ze Chen . Optimal strategic pandemic control: human mobility and travel restriction. Mathematical Biosciences and Engineering, 2021, 18(6): 9525-9562. doi: 10.3934/mbe.2021468
    [10] Avinash Shankaranarayanan, Hsiu-Chuan Wei . Mathematical modeling of SARS-nCoV-2 virus in Tamil Nadu, South India. Mathematical Biosciences and Engineering, 2022, 19(11): 11324-11344. doi: 10.3934/mbe.2022527
  • Silk is a sustainable and ecologically friendly biopolymer with a robust clinical track record in humans for load bearing applications, in part due to its excellent mechanical properties and biocompatibility. Our ability to take bottom-up and top-down approaches for the generation of silk (inspired) biopolymers has been critical in supporting the evolution of silk materials and formats, including silk nanoparticles for drug delivery. Silk nanoparticles are emerging as interesting contenders for drug delivery and are well placed to advance the nanomedicine field. This review covers the use of Bombyx mori and recombinant silks as an anticancer nanomedicine, highlighting the emerging trends and developments as well as critically assessing the current opportunities and challenges by providing a context specific assessment of this multidisciplinary field.


    The current pandemic brought several uncertainties about social and economic prospects for many people due to public intervention policies [1]. Social distancing, lockdowns and quarantines became part of our daily base routines because they are assumed to be the most efficient non-pharmaceutical interventions (NPIs) to reduce the exponential tendency for any epidemic growth when vaccines are a scarce resource. Nevertheless, many disagreements on the matter of the mobility potential impact are still ignited by civil authorities over the world. While limiting people's mobility is a legit concern, especially justified by the economic downturns resulting from lockdowns, the lack of a solid argument and evidence counteract presidential speeches. However, they cannot prevent a biased public opinion on the subject.

    Preventing jeopardizing health systems is undoubtedly the primary goal for any public authority resorting to quarantines and other measures reducing people's mobility. The current pandemic has ignited a response from the scientific community in basically all streams of research. Investigating the potential impact of different mobility dynamics is essential because public authorities' actions and studies on non-pharmaceutical interventions often include mobility sanctions as the primary strategy. Such measures include active case detection and isolation of infected persons, maintaining social distance, closure of schools and universities and most businesses, working from home when possible, quarantining and monitoring close contacts, and disclosure of multimedia epidemic and protection information by governments [2].

    NPIs, including border restrictions, quarantine, social distancing, and changes in population behavior were associated with reduced transmission of COVID-19 in Hong Kong, and are also likely to have substantially reduced influenza transmission in early February 2020 [3]. Leung et al. [4] showed that the reproduction number (Rt) decreased substantially when control measures were implemented in all analyzed cities and provinces of China. In Brazil, Nepomuceno et al. [5,6] applied data envelopment analysis for healthcare performance assessment from two different perspectives considering hospital technical and human resources. The authors reported a positive association of the most inefficient Brazilian states in preventing infections (São Paulo, Bahia, Rio Grande do Sul and Paraná) with low rates of social isolation.

    Brazil experienced peaks of 4211 and 4190 daily deaths on April 6 and April 8, 2021, and peaks of 115,041 daily cases (infections) on July 23, 2021 and 287,149 on February 3, 2022 (this last one resulted from the Omicron variant). Since June 2020, Brazil has been the second in the world in the number of COVID-19 deaths, above some of the most populated countries. One of the many reasons for the country's drastic situation is the recurrent position by the Federal Government downgrading the benefits of state-level mobility interventions. Brazil is perhaps the only big nation, in terms of geography, population or economy, that still has a leader maintaining a strong position against vaccines and the most commonly non-pharmaceutical measures. According to authors of Reference [7], who investigated geographic location data from 60 million mobile phones combined with voting information of 2018 national elections, social distancing and quarantines are significantly reduced in pro-government localities after Brazil's president dismissing the risks associated with the COVID-19 Pandemic, emphatically advising people to rebel against state mobility interventions.

    Despite contrary and contradictory opinions by the federal government, most mobility sanctions and social distancing measures in Brazil started early, enforced by states and municipal authorities. From the third week of March 2020, many public policies were applied. Measures such as closing businesses, industries and services (considered non-essential), closing schools, universities and public places, gyms, theaters, cinemas, stadiums, limiting access to transport stations, and reinventing how typical workplaces operated. Nevertheless, except for very few moments in particular regions, the social isolation index has been beneath the expected. Brazil was declared the pandemic's emerging epicenter twice in May 2020 and March 2021, leading scholars to worry about a premature easing of restrictions effects on the number of COVID-19-related deaths [8].

    The different interpretations of the current trade-off between economic development and the stability of the health systems can be attributed to a lack of feasible evaluations. From all the applied methodologies, generalized linear models (GLM) outstands among the approaches for determining associations of the COVID-19 and environmental, meteorological and ecological factors, including population density, temperature, solar radiation, air quality and meteorological variables [9,10,11,12,13]. Zhang et al. [14] employed a segmented Poisson model to analyze the available daily new cases data of the COVID-19 outbreaks in six countries, considering the governments' interventions (stay-at-home advises/orders, lockdowns, quarantines and social distancing). Another interesting instance is found in References [15,16], which investigated the contagion dynamics of the COVID-19 using a Poisson model autoregressed on daily new observed cases.

    This work aims at evaluating the mobility impact on the pandemic over different mobility and socioeconomic categories. Section 2 is dedicated to details concerning the used dataset and the applied GLM and geostatistics methodologies. Sections 3–5 are reserved for the analysis, reporting how the most appropriate model was selected (based on a multiple criteria approach) and applied, discussing how the marginal contribution of each mobility category is coherent for the aggregate and first growth moments in Brazil. The spatial incidence in the most critical areas was investigated through optimized hot spots analysis in the fourth section, questioning some of the crisis's potential determinants and government responses. The conclusion summarizes the main findings and contributions of the model development and analysis.

    During the first wave of COVID-19 in Brazil, especially in the first three months, the testing and data curation limitations lead the country to an average time lag of three weeks from the initial contamination until health authorities disclose information. This lag was because it took from one to two weeks to manifest the most severe symptoms that would lead a person to get tested for COVID-19, and from one to two weeks for the viral testing results. This empirical evidence had significant statistical support in our assessment when we performed a series of modeling considering time lags of 1, 2 and 3 weeks for new cases and total cases provided by Brazil (2020) ministry of health epidemiological report. We evaluated the percentage mobility in the six place categories offered by the google community mobility reports [17]: retail and recreation, grocery and pharmacy, parks, transit stations, workplaces, and residences. Georeferenced data at the federative level were obtained from Cota [18] COVID-19 repository.

    The google mobility reports offer information on country and regions' movement trends over different place categories. The data reports how visits or time spent (in the case of residences) varies compared to a baseline proxy for what would represent a typical value for that day of the week. The baseline day is the median value from the five weeks from January 3 to February 6, 2020, which is appropriate for our study (first COVID-19 case in Brazil reported on February 26 and most mobility interventions took place from mid-March). Some locations composing the Parks category are gardens, castles, forests, campsites and observation decks. Some areas composing Transit Stations are seaports, taxi rank, metropolitan stations, car hire agencies and motorway services. Residences are measured in terms of duration percent changes instead of visit percent changes because this is where people spend much of the day. Box 1 presents all the variables considered in the analysis.

    After the first crucial pandemic months, most regions observed more efficient testing and data curation. For this reason, the analysis for the aggregate mobility impact considering the entire 2020 was made on two weeks' time lag instead of three, which provided a more appropriate fitting for all potential model candidates with additive and multiplicative interactions, different probability distribution families (Gaussian, Gamma, Poisson) and link functions (identity, inverse and log).

    Table Box 1.  Variable definitions.
    Infection Variables:
    Y1: New COVID-19 infections two weeks later;
    Y2: Total COVID-19 infections two weeks later;
    Y1.3: New COVID-19 infections three weeks later;
    Y2.3: Total COVID-19 infections three weeks later;
    y1: New COVID-19 infections in the same week of the mobility (two weeks before);
    y2: Total COVID-19 infections in the same week of the mobility (two weeks before);
    Explanatory variables:
    x1: Mobility percent change in retail and recreation (from the baseline of February 2020);
    x2: Mobility percent change in grocery and pharmacy (from the baseline of February 2020);
    x3: Mobility percent change in parks (from the baseline of February 2020);
    x4: Mobility percent change in transit stations (from the baseline of February 2020);
    x5: Mobility percent change in workplaces (from the baseline of February 2020);
    x6: Mobility percent change in residential areas (from the baseline of February 2020);
    x7: Number of patients recovered
    Interactions:
    Y1x1; Y1x2; Y1x3; Y1x4; Y1x5; Y1x6: Interactions of new infections (2 weeks later) with the mobility;
    Y2x1; Y2x2; Y2x3; Y2x4; Y2x5; Y2x6: Interactions of total infections (2 weeks later) with the mobility;
    y1x1; y1x2; y1x3; y1x4; y1x5; y1x6: Interactions of new infections (2 weeks before) with the mobility;
    y2x1; y2x2; y2x3; y2x4; y2x5; y2x6: Interactions of total infections (2 weeks before) with the mobility.

     | Show Table
    DownLoad: CSV

    We have conducted two analyses for measuring the mobility impact on the number of infections: the first regarding the entire year, and the second, and more specific, for the beginning of the pandemic. The choice for the most appropriate models was based on statistical support, parsimony and quality of data/parameters to the empirical knowledge. Models of the Poisson family with log links presenting three weeks' time lag had a higher number of statistically significant variables and interactions (measured by the p-values) for the second analysis. Gaussian identity models with two weeks' time lag had the higher parsimony for the first analysis. Parsimony, measured by the Akaike information criterion (AIC) [19], refers to the trade-off between representativeness and simplicity of a model. The third criterion is an assumption based on specialized knowledge to avoid poor quality inputs. We have combined these perspectives into a multicriteria selection approach. This is explained in detail in Subsection 3.1 (model selection). General linear models with interactions sustained the best adherence to these criteria.

    All daily-based data were aggregated per week by summing (the number of infections and recoveries) or taking the arithmetic average for the mobilities percent change. The boxplot visualizations of Figure 1 and the information in Table 1 summarize the main descriptive statistics. The x-axis assigns one box for each place category (retail and recreation, grocery and pharmacy, parks, transit stations, workplaces, and residences). The y-axis measures the minimum, first quartile, median, third quartile, maximum values, and outliers. Notches at the left and right sides of the median line are used to investigate potential differences in distributions' medians. When boxplots' notches overlap, there is some evidence that the medians are equal.

    Figure 1.  Boxplot visualizations for the mobility categories.
    Table 1.  Descriptive statistics.
    Variable Min. 1st Qu. Median Mean 3rd Qu. Max. Std. Dev.
    Y1: New infections 0.000 1336 4122 6794 8328 79,086 9068.515
    Y2: Total infections 0.000 9334 67,261 116,660 158,156 466,191 173,289.1
    x1: Retail and recreation −81.29 −46.57 −26.86 −30.11 −15.00 13.71 20.742
    x2: Grocery and pharmacy −47.86 −6.286 6.857 5.895 17.85 51.714 17.024
    x3: Parks −82.43 −45.93 −27.50 −27.94 −11.18 36.50 23.191
    x4: Transit stations −80.43 −49.29 −29.79 −28.85 −12.61 131.43 26.223
    x5: Workplaces −58.85 −21.43 −95.000 −11.71 −0.428 208.571 15.473
    x6: Residential areas −8.857 6.857 9.714 10.23 13.86 25.857 4.871
    x7: Recovery cases 0.000 2930 49,837 99,108 137,026 287,986 153,204.4

     | Show Table
    DownLoad: CSV

    Autoregressive models with exogenous determinants are common for public policy, laws and sanctions evaluations [20]. Following a similar approach, and based on the decision criteria for choosing the most appropriate model, the number of new coronavirus infections is autoregressed to the number of new cases preceding 3 weeks and regressed to the interaction of COVID-19 base (the total number of cases up to the three weeks lag) with the place mobility variables for the analysis on the first growth. Formally:

    lnλi=Xβyi=eXβ|yi Poisson(λi)i=1,2,3,,n (1)

    where, Xβ is the linear predictor (b0+7j=1bjxi)i=1,2,3,,n, with 120 multiplicative interactions, such that:

    b0 = coefficient for the effect of COVID-19 new cases 3 reported weeks before (the R0+)

    b1 = coefficient for the effect of COVID-19 cases 3 reported weeks before (the R0)

    b2 = coefficient for the effect of retail and recreation mobility reported 3 weeks before

    b3 = coefficient for the effect of grocery and pharmacy mobility reported 3 weeks before

    b4 = coefficient for the effect of parks mobility reported 3 weeks before

    b5 = coefficient for the effect of transit stations mobility reported 3 weeks before

    b6 = coefficient for the effect of workplaces mobility reported 3 weeks before

    b7 = coefficient for the effect of residences mobility reported 3 weeks before

    Low mobility levels are expected to have a positive effect on reducing virus propagation considering the place categories. The log link function (lnλi) in the Poisson GLM is appropriate in this modeling for keeping non-negative the projection λ when the mobility regressors X or coefficients β is expected to have negative values. Likewise, higher mobility is likely to contribute to increasing transmission.

    The week from March 14 was considered the reference for counting the initial seed of cases. This is the week that 15 of the 27 federations disclosed the first cases, and most of\the state-level mobility sanctions were first implemented. The initial seed of cases works as a baseline to control the empirical estimation of the autoregressive covariate in this dynamic relationship. In addition, it works to identify the first week of significant macro data to model the phenomenon. Some federations have cases prior to the designed initial week. The northern states of Pernambuco (2 cases on March 12), Rio Grande do Norte (1 case on March 13), Alagoas (1 case on March 08) and Bahia (1 case on March 06) are some instances. The bigger seed of initial cases is for São Paulo, which had their first case on February 26, before any other federation, counting 56 new cases along the weeks until March 14, and 340 new cases in the first week from March 14, resulting in a seed of 396 initial cases.

    This section reports the results and discussion regarding the predictive modeling and marginal contributions of location mobility changes on different increasing or decreasing pandemic scenarios, considering an overall estimation for the country and additional regional assessments. First, we select the most appropriate econometric models by developing a multicriteria analysis considering the number of significant covariates using ROC weights for p-value intervals. Among the options, the choice is made over AIC [19] and empirical reasoning and vetoes imposed by multicollinearity testing. The mobility marginal contribution of each sector is then estimated in three pandemic scenarios for 2020 for each socioeconomic location at the state and national level.

    One of the primary goals in measuring the impact of some potential determinants on a dependent occurrence is selecting the most appropriate model based on identifying relevant independent (predictor) variables [21]. For this purpose, we set 40 generalized linear models with multiplicative interactions as potential candidates among many family distributions and link functions to explain the relationship between mobility in different locations and the pandemic outbreak. According to [22], the advantage of this approach is removing an unnecessarily restrictive assumption on a normally distributed error and the capacity to model different sorts of relations between the dependent variable and predictors. Among the models, we selected the 32 candidates that required less cognitive effort to explain the measure of marginal contribution for the mobility covariates (the impact of mobility on the COVID-19 spread).

    On this set of 32 potential model candidates, we employed the multicriteria PROMETHEE (preference ranking organization method for enrichment evaluation) outranking [23] considering four significance intervals as criteria for the pairwise comparison: Criteria A: number of significant covariates with p-value varying in the interval from 0 to 0.001; Criteria B: number of significant covariates with p-value ranging in the interval from 0.001 to 0.01; Criteria C: number of significant covariates with p-value ranging in the interval from 0.01 to 0.05; and Criteria D: number of covariates with p-value ranging in the interval from 0.05 to 0.1. The PROMETHEE explores outranking relations by pairwise comparisons among models, and it provides a full rank of the candidates based on positive flows (score based on comparing models outranking other models) negative flows (score based on comparing models outranked by other models) and net flows (difference between positive and negative flows for a complete order).

    One of the main issues in this process is how to define appropriate weights for the evaluation criteria (significant p-value intervals). The surrogate weighting procedure is the simplest method, not requiring additional cognitive efforts from a decision-maker. They come from predefined rankings converted into numerical weights by surrogate functions [24,25]. According to Almeida et al. [26], the surrogate weighting procedure more appropriate for the PROMETHEE procedure is the ROC (rank order centroid) method of Barron [27]. Compared to other methods, ROC weights attribute a more significant emphasis on the criteria ranked higher in the ranking construction, which is appropriate for our model selection since we consider more important the models with higher significant variables in explaining changes in the pandemic outbreak. Consider Rw={w1w2wkwL} a vector of weights for "k" models. The values for each rank position can be determined by:

    wk=1LLω=k1ω,k={1,2,.,L} (2)

    From the 32 candidates, 16 have a positive net flow, which are ranked according to the AIC in Table 2. The BP p-value refers to the Breusch-Pagan test [28] for heteroskedasticity, considering the log transformation for the dependent variable. Because this multicriteria approach favors complex multivariate models, ranking the candidates based on AIC prevents a choice for an overfitting model, favoring the simplest yet most robust goodness of fit alternative. The first nine models are not significantly different in terms of AIC mean deviation. Despite a sizeable explanatory power, many models reported spurious correlations, multicollinearity, and aggregate data compensation bias (as defined by Nepomuceno and Costa [29]). For instance, models that considered the number of recovery cases reported a positive relation of this data regarding the spread of COVID-19 (models 17, 20, 21, 23 and 24), or some models reporting a negative association of the number of new cases (models 11 and 15) and total cases (model 4) with COVID-19 cases two weeks later (which would mean that more cases this week implies less propagation).

    Table 2.  Model candidates.
    Criteria A B C D Net Flow AIC BP P-value
    ROC Weights 0.5208 0.2708 0.1458 0.0625
    Model 2 7 0 0 0 0.1015 19,011.66 0.004306
    Model 21 7 0 0 1 0.164 19,020.83 0.006528
    Model 17 7 0 0 0 0.1015 19,021.72 0.004239
    Model 24 8 0 0 0 0.3031 19,061.39 0.216777
    Model 8 7 0 0 0 0.1015 19,068.6 0.137666
    Model 20 7 0 0 0 0.1015 19,083.3 0.199999
    Model 7 6 2 0 0 0.1653 19,188.64 0.053166
    Model 1 7 0 0 1 0.164 19,220.5 0.006641
    Model 23 6 1 1 1 0.2554 20,319.41 0.091411
    Model 4 8 0 0 0 0.3031 22,260.87 0.000582
    Model 18 7 1 0 0 0.3374 22,306.72 0.001029
    Model 12 8 0 0 0 0.3031 23,548.04 0.0000152
    Model 11 7 0 0 0 0.1015 25,899.05 0.009428
    Model 13 7 1 0 0 0.3374 25,924.29 0.001113
    Model 15 6 1 0 0 0.0517 26,996.51 0.002051
    Model 9 6 1 0 0 0.0517 28,975.32 0.000189

     | Show Table
    DownLoad: CSV

    The presence of multicollinearity means that two or more covariates are near perfect linear combinations of one another. It may imply in unstable or biased regression estimators, high standard errors or wide confidence intervals that may not reject the null hypothesis for the significance of variables, i.e., model parameters become indeterminate or problematic to estimate precisely. For testing multicollinearity in our models, we used variance inflation factor (VIF) diagnosis [30]. VIF measures how much variance is inflated from the expected true value due to correlations among the predictors. When VIF equals 1, no correlations among the predictors are observed. VIFs exceeding 10 indicate evidence of multicollinearity requiring correction [30].

    The VIF is a veto for choosing the most appropriate candidate in the model selection. The most appropriate model was model 13, reporting minimum correlation among predictors. Table 3 reports the VIFs and tolerance (variance percentage that cannot be accounted for by other predictors when the evaluated variable is regressed on the rest of the predictors in the model). Figure 2 illustrates the correlation matrix with a set of scattergrams between pairs of variables, excluding the variable Y2x1 reporting high multicollinearity.

    Table 3.  Multicollinearity diagnostic.
    Variables Description Tolerance VIF
    y1 New COVID-19 infections in the same week of the mobility 0.14639 6.9870
    Y2x2 Interactions of infections (2 weeks later) with percent change mobilities on Groceries and Pharmacies 0.63853 1.5661
    Y2x3 Interactions of infections (2 weeks later) with percent change mobilities on Parks 0.30537 3.2747
    Y2x4 Interactions of infections (2 weeks later) with percent change mobilities on Transit Stations 0.12747 7.8446
    Y2x5 Interactions of infections (2 weeks later) with percent change mobilities on Workplaces 0.26626 3.7557
    Y1x6 Interactions of infections (2 weeks later) with percent change mobilities on Residences 0.24295 4.1161

     | Show Table
    DownLoad: CSV
    Figure 2.  Correlation matrix and dispersions.

    Each model candidate was checked considering the empirical reasoning reported in the previous subsection. The most appropriate model (Model 13) has seven very high significant variables (criteria A), the highest positive net flow of 0.3374 and AIC scoring 26,452.66. Retail and recreation reports a significant multicollinearity (VIF = 20.593403, tolerance = 0.04855924). For informational purposes, it is highlighted in red in the summary of Table 4. The model reports residual standard error (RSE) equal to 28,020 on 1127 degrees of freedom and Adjusted R-squared of 0.9739. Table 4 summarizes the model, residuals, covariates and other relevant statistics.

    Table 4.  Summary of results (aggregate impact).
    Parameter Marginal Contribution Std. Error p-value Signf. Criteria Mean (X) Description
    β0 1059.2 1086.3 0.33 Not significant Intercept
    R0 1.7355 0.246110 < 3.2e−12 Criteria A 86862.47 Marginal impact from the total COVID-19 Cases two weeks before (y2)—the R0.
    β1 −0.0178 0.000068 < 2e−16 Criteria A −30.11235 Marginal impact from the interaction between cases and mobility in retail and recreation (Y2×x1)
    β2 0.03444 0.000328 < 2e−16 Criteria A 5.895083 Marginal impact from the interaction between cases and mobility in groceries and pharmacies (Y2×x2)
    β3 −0.00619 0.000300 < 2e−16 Criteria A −27.93886 Marginal impact from the interaction between the cases and the mobility in parks two weeks before (Y2×x3)
    β4 −0.01366 0.000610 < 2e−16 Criteria A −28.84866 Marginal impact from the interaction between cases and mobility in transit stations (Y2×x4)
    β5 −0.00929 0.000928 < 2e−16 Criteria A −11.71315 Marginal impact from the interaction between cases and the mobility in workplaces (Y2×x5)
    β6 −0.06828 0.016185 2.63e−05 Criteria A 10.23406 Marginal impact from the interaction between cases and mobility in residences (Y2×x6)
    Note: Model 13: glm(formula = Y2~y1 + Y2x2 + Y2x3 + Y2x4 + Y2x5 + Y1x6); Residuals: Min = −187800, 1st Quartile = −9118, Median = −1165, 3rdQuartile = 6739, Max = 210505; Residual standard error: 28020 on 1127 degrees of freedom; Multiple R-squared: 0.974; Adjusted R-squared: 0.9739; F-statistic: 7037 on 6 and 1127 DF; p-value: < 2.2e16; AIC = 26452.66; PROMETHEE Net Flow = 0.3374; Min VIF = 1.566095 (β2); Max VIF = 7.844674 (β4).

     | Show Table
    DownLoad: CSV

    The estimated model has good explanatory power with the intercept as the only not significant parameter. The autoregressive parameter (R0) reporting 1.7355 means that, keeping all the mobility changes constant, 100 COVID-19 infections in the country two weeks before would produce about 173.55 new cases on average, with 99% confidence. Of course, this generalization does not consider individual regional characteristics, but it helps understand the complete panorama. According to the marginal contributions, the information about the mean of covariates reported in the sixth column helps interpret the relationship between the mobilities and COVID-19 cases. The mobility percentage change for retail and recreation, parks, transit stations and workplaces are negative for most of the year due to state-level mobility interventions and quarantine campaigns, which highly downward the overall mean to a negative value. For the mobility occurring in retail and recreation, one percent decrease in the negative mobility (one percent increase in mobility) provokes, on average, a positive response in the number of new COVID-19 cases in the amounts of 0.0178 × Y2. For instance, during the last week of September in the state of Ceará, this number would be 0.0178*4633 = 82.46 new cases, keeping everything else constant.

    For the mobility occurring in parks, one percent increase in mobility (one percent decrease in the negative mobility) provokes, on average, a positive response of 0.00619 × Y2 new COVID-19. For instance, during the second week of August in the state of Pernambuco, this estimate would be 0.00619 × 7537 = 46.65 new cases, keeping everything else constant. For the mobility occurring in transit stations, one percent decrease in the negative mobility (one percent increase in mobility) provokes, on average, a positive response in the number of new COVID-19 cases in the amounts of 0.01366 * Y2. For instance, during the last week of November in the state of Rio de Janeiro, this estimate would be 0.01366 × 20293 = 277.20 new cases, keeping everything else constant.

    A slight (positive) change in mobility is observed for the groceries and pharmacies, which on average increased about 5.9% during the pandemic compared to one month before the first mobility restrictions. This can be attributed to the nature of those businesses (essential activities) which were not affected by the mobility sanctions and lockdowns. For the specific case of this mobility category, one percent increase in the mobility in groceries or pharmacies provokes, on average, a positive response in the number of COVID-19 new cases in the amounts of 0.000328 × Y2. For instance, during the first week of May in the state of São Paulo, this number would be 0.000328×11456 = 3.75 new cases, keeping everything else constant. During the last week of December in the state of Minas Gerais (one of the states with the best response to the pandemic in terms of mortality rate), this estimate would be 0.000328 × 26666 = 8.74 new cases, keeping everything else constant.

    Increasing mobilities in workplaces provoke, on average, 0.00929 × Y2 new cases. For the last example in Minas Gerais, this represent 0.00929 × 26666 = 247.72 new cases. Residential areas are the only mobility category reporting a negative relation with the number of COVID-19 new infections. The same relation was verified for the vast majority of models during the selection phase. The mobility change in residential areas is negative related to the number of new cases in the amount of -0.06828 × Y2. This estimate means that, in the examples of Ceará, São Paulo, Rio de Janeiro, Pernambuco and Minas Gerais, for every one percent increase in the mobility to residences, the number of COVID-19 cases would decrease in the amounts of 0.06828 × 4633 = 316.34, 0.06828×11456 = 782.21, 0.06828×20293 = 1385.60, 0.06828×7537 = 514.62 and 0.06828×26666 = 1820.75 cases, respectively.

    Table 5 reports the main results for the mobility impact during the first pandemic growth. The coefficient sign represents the overall data relation of that specific base or place category with COVID-19 propagation (new cases). Two places categories exhibit a negative relation with COVID-19 (retail, recreation and workplaces). This can be interpreted as a spurious correlation problem: the same weeks an aggressive mobility reduction was observed in business, civic and cultural activities due to sanctions aiming at closing malls, shops, commercial centers, theaters, stadiums, outdoor events, service units and public administrations coincide with the COVID-19 exponential increase in Brazil. We cannot attribute, however, this increase to the mobility reduction in these places. This leads to believe correlation does not imply causation for these categories considering Brazil as a whole. This prospect may not be true for smaller disaggregated data on regions, states or municipalities.

    Table 5.  Estimates for Brazil on the first growth.
    Covariates Marginal contribution1, 2 Exp. (Coeff.) Coefficients P-values Signf. criteria
    COVID-19 (R0) 0.9026342 2.1831049 7.807e−01 < 2e−16 Criteria A
    COVID-19 (R0+) 7.49274 1.0020168 2.015e−03 < 2e−16 Criteria A
    Retail & recreation 5.543848 0.1355574 −1.998e+00 < 2e−16 Criteria A
    Grocery & pharmacy −1.600595 19.7734385 2.984e + 00 < 2e−16 Criteria A
    Parks 0.1696293 1.1263827 1.190e−01 0.01271 Criteria C
    Transit stations 0.7903152 2.2323520 8.031e−01 4.03e−12 Criteria A
    Workplaces −3.774877 0.7883294 −2.378e−01 0.000103 Criteria A
    Residences 3.760099 2.4628843 9.013e−01 0.000437 Criteria A
    Note: glm(formula = Y1~b0+7j=1bjxi; Poisson (link = "log"); AIC: 14313; null dv.: 3229781 (173 df); res. dv.: 12680 (45 df); Fisher SI: 6; mult. interact. 120; min. dvrs.:22.392; max. dvrs.: 34.825; 1q. dvrs.:3.702; 3q. dvrs: 2.399; med. dvrs.: 0.018. 1 Marginal contribution (Impact) in the cases from May 9 to May 16, 2020 (third week of May). 2 Estimates for a marginal increment of one additional COVID-19 case and 0.001% additional mobility.

     | Show Table
    DownLoad: CSV

    Another evidence is the exponential coefficient for these places categories. The exponential coefficient (Exp. Coeff.) is a proxy for how much new cases respond, on average, to cases and to mobility changes three weeks before. The estimates for retail and recreation and workplaces are small (0.135 and 0.788, respectively) compared to other estimates such as transit stations (2.232) and residences (2.463), despite the very significant p-values. The very high exponential coefficient for Grocery and Pharmacy is another instance of spurious data correlation in this case. The same weeks supermarkets, grocery stores and pharmacies have experienced a considerable demand increase coinciding with the COVID-19 exponential increase in Brazil. Nevertheless, we have no additional evidence to attribute COVID-19 exponential increase to the mobility increase in these places, leading to believe correlation does not imply causation for this category either, considering Brazil as a whole.

    Because of the multiple interaction modeling structures, the exponential coefficients cannot be interpreted as additive predictors individually. Nevertheless, they provide a notion of the overall importance each base or place mobility has on new cases of COVID-19. We provide the marginal contribution estimates for a more empirical quantitative measure of the mobility impact. In this dynamic modeling, estimates for the R0, R0+ and for the mobility impact of each place category change along the time and along the regions and states. For each week and for each region, we have a different marginal contribution. The marginal contributions reported in Table 4 are estimations for the individual effect of each variable on the number of new coronavirus cases from May 9 to May 16, 2020. These are obtained by adding one unit change to the base of cases (new cases or total) or by adding 0.001% mobility change to the place category (retail and recreation, grocery and pharmacy, parks, transit stations, workplaces or residences), keeping everything else constant, them investigating the potential increase or decrease in the prediction for COVID-19. Statistical support is measured by the p-values.

    Estimates for the R0 and R0+ vary according to the week and region. For the third week of May we can assume each new case three weeks before was responsible, on average, for 7.49 new cases. The mobility increase in grocery and pharmacy and workplaces had a negative effect on new cases, whereas the mobility increase in retail & recreation, parks, transit stations and residences had a positive effect. The positive effect of mobility changes in residential areas can be attributed to the vulnerability in favelas and poor communities. Despite the pandemic, Baile funk parties, peladas (pick-up soccer games) and many other forms of social organization are reported in many peripheries and informal urban settlements across the country.

    The flexible scatterplots illustrate the mobility impact of each place category (retail and recreation, grocery and pharmacy, parks, transit stations, workplaces, residences) in the COVID-19 dissemination (new cases) in Brazil. They have different associations along the time depending on the number of confirmed cases (total). Taking the impact of retail and recreation as an example, the Poisson coefficient estimates an overall negative impact for this place category (see Table 5). However, the marginal contribution in the third week of May is positive because the number of cases 3 weeks before is 2167, and the retail and recreation mobility is between 60% and −40% (−57.66138 to be precise). This locates our estimation for the marginal contribution of retail and recreation mobility to the third frame of Figure 3 Panel (a), a positive effect of 5.5438 new cases for each 0.001% increase in mobility.

    Figure 3.  Dynamic mobility impact by place category in Brazil (first growth).

    According to the different regions (north, northeast, center west, southeast and south), some estimates are reported in Tables 610. Figures 48 illustrate the spatial concentration. When disaggregating the country data into small datasets, there is a lack of significant samples for each region on the first growth. Therefore, most estimation for the mobility impact and contagion does not have statistical support. For this reason, this information is omitted in the Tables, only reporting the significant mobility impacts. The most significant impacts are the mobility in workplaces for the southeast region, parks for the center-west, residential areas in the northeast and retail and recreation in the north region. The estimates refer to a 1% marginal change in all categories.

    Table 6.  Estimates for the north region.
    Covariates Marginal contribution P-value
    COVID-19 (R0) 3.981363 < 2e−16
    Retail & recreation 635.1778 < 2e−16
    Note: AIC: 2411; Poisson (link = "log"); null dev.: 74461.6 on 40 df.; res. Dev.: 2049.2 on 18 df; fisher SI: 5.

     | Show Table
    DownLoad: CSV
    Table 7.  Estimates for the northeast region.
    Covariates Marginal Contribution P-value
    COVID-19 (R0) 0.7959314 < 2e−16
    Retail & recreation 27.84628 < 2e−16
    Parks 11.64888 < 2e−16
    Workplaces 183.5822 < 2e−16
    Residences 301.5848 < 2e−16
    Note: AIC: 585.69; Poisson (link = "log"); null dev.: 1.054e+06 on 58 df.; res. dev.: 2.921e12 on 0 df; fisher SI: 3.

     | Show Table
    DownLoad: CSV
    Table 8.  Estimates for the central-west region.
    Covariates Marginal Contribution P-value
    Parks 40.19183 0.0079
    Note: AIC: 243.31; Poisson (link = "log"); null dev.: 7514.92 on 25 df.; res. dev.: 16.98 on 2 df; fisher SI: 4.

     | Show Table
    DownLoad: CSV
    Table 9.  Estimates for the southeast region.
    Covariates Marginal Contribution P-value
    COVID-19 (R0)
    COVID-19 (R0+) 1.291627 < 2e−16 ***
    Parks 9.183066 < 2e−16 ***
    Workplaces 80.36371 < 2e−16 ***
    Note: AIC: 655.84; Poisson (link = "log"); null dev.: 113107.42 on 27 df.; res. dev.: 351.29 on 4 df; fisher SI: 4.

     | Show Table
    DownLoad: CSV
    Table 10.  Estimates for the south region.
    Covariates Marginal contribution P-value
    No significant covariate

     | Show Table
    DownLoad: CSV
    Figure 4.  North.
    Figure 5.  Northeast.
    Figure 6.  Central west.
    Figure 7.  Southeast.
    Figure 8.  South.

    The most critical cities are evaluated using Optimized Hot Spots Analysis. This methodology offers statistics for each location indicating where cities with high or low COVID-19 cases cluster spatially [31]. The objective is to assess the COVID-19 intensity through cities. We question whether potential mobility contributions to eventual hot spots could indicate a spatial concentration of municipalities composing the input features aggregating the COVID-19 occurrences. A similar analysis can be accessed in References [29,32], investigating the spatial concentration of crime in Brazil's urban spaces. In this case, the occurrences for each geographic location are the analyzed field. Figure 9 presents the visualization.

    Figure 9.  COVID-19 Optimized hot spots in Brazil (first growth).

    The hot spot cluster for 99% confidence is composed of the metropolitan areas of the capitals Manaus in the north, Ceará in the northeast, São Paulo and Rio de Janeiro in the southeast. In addition to significant COVID-19 incidence, these cities are surrounded by other cities with high COVID-19 incidence. These regions are the most probable responsible for the pandemic in the country. The impact on the health system is summed to a higher probability of contagion in these areas, which has made this significant spatial association critical and required urgent attention by the public authorities. Table 11 reports regional and aggregate descriptive typologies for additional insights in the spatial analysis. The feature input type is Brazil's municipalities (cities) containing COVID-19 infections per city as the analysis field. The number of valid inputs is 4965 with a maximum of 41451 infections in one single city in the southeast region, mean of 67.736, std. dev. equal to 781.06 and 66 outlier locations. The 99% confidence hot spots represent 6.76% of input features, highlighting the spatial concentration.

    Table 11.  Estimates and statistics for the spatial analysis.
    Descriptive Statistics North Northeast South Southeast Central West Brazil
    Cities 584 1898 787 1412 266 4965
    Hotspots features* 41 79 75 175 11 336
    Min. cases 1 1 1 1 1 1
    Max. cases 12317 18644 711 41451 5542 41451
    Total cases 65333 107939 14259 129531 9858 336311
    Mean 111, 87 559.26 18.118 91.736 37.060 67.736
    Std. dev. 651, 71 2385.2 62.635 225.67 345.75 781.062
    Mean Z-score 0, 1227 0.0169 0.3124 0.2413 0.0684 0.0987
    Std. dev. Z-score 1, 2170 1.1324 1.3373 1.4642 0.8758 1.38345
    Mean P-value 0, 5200 0.6476 0.3767 0.6406 0.7292 0.62183
    Std. dev. P-value 0, 2459 0.2574 0.3023 0.2592 0.2639 0.24533
    Higher spatial concentration Amazonas Pernambuco; Ceará Santa Catarina São Paulo; Rio de Janeiro Goiás; Distrito Federal São Paulo; Rio de Janeiro
    Note: *At 99% confidence.

     | Show Table
    DownLoad: CSV

    The distances for geographic coordinates are analyzed using chordal distance in meters. Outliers were used to compute the optimal fixed distance band. There are 336 output features statistically significant based on false discovery rate correction for multiple testing and spatial dependence [33]. Figure 10 illustrates the distributions of optimized hot spots p-value frequencies, indicating the strength of spatial concentration per region. The p-values represent the statistical significance for whether or not to reject the null hypothesis that the observed high or low spatial clustering is different from a random distribution for the same features. High z-scores and small p-values for a feature indicate a spatial clustering of high COVID-19 infections. Low negative z-scores and small p-values indicate a spatial clustering of low COVID-19 cases. Higher p-values indicate no significant spatial clustering. The region with smaller p-value distributions (below 0.05) is the South, reporting an average 0.3767 p-value.

    Figure 10.  P-value frequency distributions.

    We can see the regions reporting the most significant hot spots (north, northeast and southeast) undergo unregular frequency distributions, basically concentrating the pandemic in singular features. Some potential explanations are discussed ahead. Rio de Janeiro and São Paulo states have the second and third highest population density, and they have the fifth and second capitals with the highest population density. Ceará has the capital with the highest population density in Brazil. This information, combined with the airline hubs for international flights in these capitals, can explain, at least partially, the high initial incidence and fast spread of COVID-19.

    Another factor that may explain the critical scenario for those regions is related to the most famous Brazilian festival: the carnival. Table 12 reports the average percentage mobility 3 weeks prior to the first confirmed cases of community Transmission on each state in the hot spot cluster. Community transmission is characterized when authorities are unable to trace the source of the infection. The states of São Paulo (SP) and Rio de Janeiro (RJ) have the first community transmission officially reported on March 13. According to our specifications, the first case can be traced back to the last week of February during the carnival holidays. During the carnival week, all place categories in São Paulo and Rio de Janeiro had a negative mobility change, with an exemption on parks and residential areas which are the places concentrating street carnival and parades.

    Table 12.  Average mobility (3 weeks prior to the first to community transmission).
    UF Week n-3 Retail & Recreation Grocery & Pharmacy Parks Transit stations Workplaces Residences
    SP 4th Feb. (carnival) −9571 −1857 6286 −11,286 −12,571 4429
    RJ 4th Feb. (carnival) −13,857 −5714 32,286 −11,000 −27,143 5429
    CE 1st March −9571 1714 −29,571 −2,857 8143 1571
    AM 2nd March 2286 6857 1000 15,000 15,143 −1429

     | Show Table
    DownLoad: CSV

    Brazil's COVID-19 response has been one of the most controversial in the world. If, on one hand, state governments and municipalities' efforts follow international protocols and recommendations, on the other, there is a big question about the federal administration's institutional preparedness, which constantly jeopardizes the prospects of mobility sanctions, social distancing, and quarantine campaigns. The results reported in this assessment aim at contributing to this discussion. In particular, reducing mobility in residential areas and retail places is a challenge local authorities should prioritize due to their high potential. Non-parametric governance models for prioritizing and allocating resources and benchmarking knowledge from local administrations with efficient mobility strategies can be an interesting avenue [34,35,36]. The absence of the state in favelas can make longer the trade-off between the economy and health. Unquarantine planning based on the individual impact of these place categories and the interaction with external localities can anticipate optimal responses for Brazilian communities to prevent drastic outcomes from the pandemic. For a more profound knowledge on this topic, please refer to this reference list with recent applications in Brazil [37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55].

    The authors acknowledge the support from the Federal University of Pernambuco through the called "Edital Propesq nº 06/2020 – Edital Emergencial de Credenciamento e Fomento de Projetos, Visando Ações para o Diagnóstico e Prevenção da Covid-19".

    Authors declare no conflict of interest.

    [1] Vollrath F, Porter D (2009) Silks as ancient models for modern polymers. Polymer 50: 5623–5632. doi: 10.1016/j.polymer.2009.09.068
    [2] Lubec G, Holaubek J, Feldl C, et al. (1993) Use of silk in ancient Egypt. Nature 362: 25.
    [3] Altman GH, Diaz F, Jakuba C, et al. (2003) Silk-based biomaterials. Biomaterials 24: 401–416. doi: 10.1016/S0142-9612(02)00353-8
    [4] Omenetto FG, Kaplan DL (2010) New opportunities for an ancient material. Science 329: 528–531. doi: 10.1126/science.1188936
    [5] Elices M, Plaza GR, Perez RJ, et al. (2011) The hidden link between supercontraction and mechanical behavior of spider silks. J Mech Behav Biomed Mater 4: 658–669. doi: 10.1016/j.jmbbm.2010.09.008
    [6] Cranford SW, Tarakanova A, Pugno NM, et al. (2012) Nonlinear material behaviour of spider silk yields robust webs. Nature 482: 72–76. doi: 10.1038/nature10739
    [7] Vollrath F, Porter D, Holland C (2013) The science of silks. MRS Bull 38: 73–80. doi: 10.1557/mrs.2012.314
    [8] Kluge JA, Rabotyagova O, Leisk GG, et al. (2008) Spider silks and their applications. Trends Biotechnol 26: 244–251. doi: 10.1016/j.tibtech.2008.02.006
    [9] Gatesy J, Hayashi C, Motriuk D, et al. (2001) Extreme diversity, conservation, and convergence of spider silk fibroin sequences. Science 291: 2603–2605. doi: 10.1126/science.1057561
    [10] Hardy JG, Scheibel TR (2009) Silk-inspired polymers and proteins. Biochem Soc T 37: 677–681. doi: 10.1042/BST0370677
    [11] Rising A, Johansson J (2015) Toward spinning artificial spider silk. Nat Chem Biol 11: 309–315. doi: 10.1038/nchembio.1789
    [12] Kim S, Marelli B, Brenckle MA, et al. (2014) All-water-based electron-beam lithography using silk as a resist. Nat Nanotechnol 9: 306–310. doi: 10.1038/nnano.2014.47
    [13] Omenetto FG, Kaplan DL (2008) A new route for silk. Nature Photonic 2: 641–643. doi: 10.1038/nphoton.2008.207
    [14] Zhu B, Wang H, Leow WR, et al. (2016) Silk fibroin for flexible electronic devices. Adv Mater 22: 4250–4265.
    [15] Doblhofer E, Schmid J, Riess M, et al. (2016) Structural insights into water-based spider silk protein-nanoclay composites with excellent gas and water vapor barrier properties. ACS Appl Mater Interface 8: 25535–25543. doi: 10.1021/acsami.6b08287
    [16] Marelli B, Brenckle MA, Kaplan DL, et al. (2016) Silk fibroin as edible coating for perishable food preservation. Sci Rep 6: 25263–25273. doi: 10.1038/srep25263
    [17] Abbott RD, Kimmerling EP, Cairns DM, et al. (2016) Silk as a biomaterial to support long-term three-dimensional tissue cultures. ACS Appl Mater Interface 8: 21861–21868. doi: 10.1021/acsami.5b12114
    [18] Jao D, Mou X, Hu X (2016) Tissue regeneration: a silk road. J Funct Biomater 7: 22–39. doi: 10.3390/jfb7030022
    [19] Kasoju N, Bora U (2012) Silk fibroin in tissue engineering. Adv Healthc Mater 1: 393–412. doi: 10.1002/adhm.201200097
    [20] Werner V, Meinel L (2015) From silk spinning in insects and spiders to advanced silk fibroin drug delivery systems. Eur J Pharm Biopharm 97: 392–399. doi: 10.1016/j.ejpb.2015.03.016
    [21] Yucel T, Lovett ML, Kaplan DL (2014) Silk-based biomaterials for sustained drug delivery. J Control Release 190: 381–397. doi: 10.1016/j.jconrel.2014.05.059
    [22] Seib FP, Kaplan DL (2013) Silk for drug delivery applications: opportunities and challenges. Israel J Chem 53: 756–766.
    [23] Thurber AE, Omenetto FG, Kaplan DL (2015) In vivo bioresponses to silk proteins. Biomaterials 71: 145–157. doi: 10.1016/j.biomaterials.2015.08.039
    [24] Pritchard EM, Dennis PB, Omenetto FG, et al. (2012) Physical and chemical aspects of stabilization of compounds in silk. Biopolymers 97: 479–498. doi: 10.1002/bip.22026
    [25] Chiu B, Coburn J, Pilichowska M, et al. (2014) Surgery combined with controlled-release doxorubicin silk films as a treatment strategy in an orthotopic neuroblastoma mouse model. Brit J Cancer 111: 708–715. doi: 10.1038/bjc.2014.324
    [26] Seib FP, Coburn J, Konrad I, et al. (2015) Focal therapy of neuroblastoma using silk films to deliver kinase and chemotherapeutic agents in vivo. Acta Biomater 20: 32–38. doi: 10.1016/j.actbio.2015.04.003
    [27] Coburn J, Harris J, Zakharov AD, et al. (2017) Implantable chemotherapy-loaded silk protein materials for neuroblastoma treatment. Int J Cancer 140: 726–735. doi: 10.1002/ijc.30479
    [28] Seib FP, Pritchard EM, Kaplan DL (2013) Self-assembling doxorubicin silk hydrogels for the focal treatment of primary breast cancer. Adv Funct Mater 23: 58–65. doi: 10.1002/adfm.201201238
    [29] Jastrzebska K, Kucharczyk K, Florczak A, et al. (2015) Silk as an innovative biomaterial for cancer therapy. Rep Pract Oncol Radiother 20: 87–98. doi: 10.1016/j.rpor.2014.11.010
    [30] Coleman RE (2012) Bone cancer in 2011: prevention and treatment of bone metastases. Nat Rev Clin Oncol 9: 76–78.
    [31] Gupta GP, Massague J (2006) Cancer metastasis: building a framework. Cell 127: 679–695. doi: 10.1016/j.cell.2006.11.001
    [32] Ehrlich P (1913) Address in pathology, on chemiotherapy: delivered before the seventeenth international congress of medicine. Brit Med J 2: 353–359. doi: 10.1136/bmj.2.2746.353
    [33] Mottaghitalab F, Farokhi M, Shokrgozar MA, et al. (2015) Silk fibroin nanoparticle as a novel drug delivery system. J Control Release 206: 161–176. doi: 10.1016/j.jconrel.2015.03.020
    [34] Zhao Z, Li Y, Xie MB (2015) Silk fibroin-based nanoparticles for drug delivery. Int J Mol Sci 16: 4880–4903. doi: 10.3390/ijms16034880
    [35] Ebrahimi D, Tokareva O, Rim NG, et al. (2015) Silk-its mysteries, how it is made, and how it is used. ACS Biomater Sci Eng 1: 864–876. doi: 10.1021/acsbiomaterials.5b00152
    [36] Eisoldt L, Thamm C, Scheibel T (2012) Review the role of terminal domains during storage and assembly of spider silk proteins. Biopolymers 97: 355–361. doi: 10.1002/bip.22006
    [37] Xu G, Gong L, Yang Z, et al. (2014) What makes spider silk fibers so strong? From molecular-crystallite network to hierarchical network structures. Soft Mat 10: 2116–2123.
    [38] Ha SW, Gracz HS, Tonelli AE, et al. (2005) Structural study of irregular amino acid sequences in the heavy chain of bombyx mori silk fibroin. Biomacromolecules 6: 2563–2569. doi: 10.1021/bm050294m
    [39] Asakura T, Ohgo K, Ishida T, et al. (2005) Possible implications of serine and tyrosine residues and intermolecular interactions on the appearance of silk istructure of bombyx mori silk fibroin-derived synthetic peptides: high-resolution 13c cross-polarization/magic-angle spinning NMR study. Biomacromolecules 6: 468–474. doi: 10.1021/bm049487k
    [40] Asakura T, Okushita K, Williamson MP (2015) Analysis of the structure of bombyx mori silk fibroin by NMR. Macromolecules 48: 2345–2357. doi: 10.1021/acs.macromol.5b00160
    [41] Zhou CZ, Confalonieri F, Jacquet M, et al. (2001) Silk fibroin: structural implications of a remarkable amino acid sequence. Proteins 44: 119–122. doi: 10.1002/prot.1078
    [42] Zhou CZ, Confalonieri F, Medina N, et al. (2000) Fine organization of bombyx mori fibroin heavy chain gene. Nucleic Acids Res 28: 2413–2419. doi: 10.1093/nar/28.12.2413
    [43] Jin HJ, Kaplan DL (2003) Mechanism of silk processing in insects and spiders. Nature 424: 1057–1061. doi: 10.1038/nature01809
    [44] Greving I, Dicko C, Terry A, et al. (2010) Small angle neutron scattering of native and reconstituted silk fibroin. Soft Mat 6: 4389–4395. doi: 10.1039/c0sm00108b
    [45] Lu Q, Zhu H, Zhang C, et al. (2012) Silk self-assembly mechanisms and control from thermodynamics to kinetics. Biomacromolecules 13: 826–832. doi: 10.1021/bm201731e
    [46] Wang X, Yucel T, Lu Q, et al. (2010) Silk nanospheres and microspheres from silk/pva blend films for drug delivery. Biomaterials 31: 1025–1035. doi: 10.1016/j.biomaterials.2009.11.002
    [47] Myung SJ, Kim HS, Kim Y, et al. (2008) Fluorescent silk fibroin nanoparticles prepared using a reverse microemulsion. Macromol Res 16: 604–608. doi: 10.1007/BF03218567
    [48] Gupta V, Aseh A, Rios CN, et al. (2009) Fabrication and characterization of silk fibroin-derived curcumin nanoparticles for cancer therapy. Int J Nanomed 4: 115–122.
    [49] Lammel AS, Hu X, Park SH, et al. (2010) Controlling silk fibroin particle features for drug delivery. Biomaterials 31: 4583–4591. doi: 10.1016/j.biomaterials.2010.02.024
    [50] Kundu J, Chung YI, Kim YH, et al. (2010) Silk fibroin nanoparticles for cellular uptake and control release. Int J Pharm 388: 242–250. doi: 10.1016/j.ijpharm.2009.12.052
    [51] Seib FP, Jones GT, Rnjak KJ, et al. (2013) pH-dependent anticancer drug release from silk nanoparticles. Adv Healthc Mater 2: 1606–1611. doi: 10.1002/adhm.201300034
    [52] Wongpinyochit T, Johnston BF, Seib FP (2016) Manufacture and drug delivery applications of silk nanoparticles. J Vis Exp DOI: 10.3791/54669.
    [53] Zhang YQ, Shen WD, Xiang RL, et al. (2007) Formation of silk fibroin nanoparticles in water-miscible organic solvent and their characterization. J Nanopart Res 9: 885–900. doi: 10.1007/s11051-006-9162-x
    [54] Zhao Z, Xie M, Li Y, et al. (2015) Formation of curcumin nanoparticles via solution-enhanced dispersion by supercritical CO2. Int J Nanomed 10: 3171–3181.
    [55] Lozano PAA, Montalban MG, Aznar CSD, et al. (2015) Production of silk fibroin nanoparticles using ionic liquids and high-power ultrasounds. J Appl Polym Sci 132: 41702–41709.
    [56] Gholami A, Tavanai H, Moradi AR (2010) Production of fibroin nanopowder through electrospraying. J Nanopart Res 13: 2089–2098.
    [57] Wenk E, Wandrey AJ, Merkle HP, et al. (2008) Silk fibroin spheres as a platform for controlled drug delivery. J Control Release 132: 26–34. doi: 10.1016/j.jconrel.2008.08.005
    [58] Lu Q, Huang Y, Li M, et al. (2011) Silk fibroin electrogelation mechanisms. Acta Biomater 7: 2394–2400. doi: 10.1016/j.actbio.2011.02.032
    [59] Rajkhowa R, Wang L, Wang X (2008) Ultra-fine silk powder preparation through rotary and ball milling. Powder Technol 185: 87–95. doi: 10.1016/j.powtec.2008.01.005
    [60] Mathur AB, Gupta V (2010) Silk fibroin-derived nanoparticles for biomedical applications. Nanomedicine 5: 807–820. doi: 10.2217/nnm.10.51
    [61] Xiao L, Lu G, Lu Q, et al. (2016) Direct formation of silk nanoparticles for drug delivery. ACS Biomater Sci Eng 2: 2050–2057. doi: 10.1021/acsbiomaterials.6b00457
    [62] Wongpinyochit T, Uhlmann P, Urquhart AJ, et al. (2015) PEGylated silk nanoparticles for anticancer drug delivery. Biomacromolecules 16: 3712–3722. doi: 10.1021/acs.biomac.5b01003
    [63] Subia B, Chandra S, Talukdar S, et al. (2014) Folate conjugated silk fibroin nanocarriers for targeted drug delivery. Integr Biol 6: 203–214. doi: 10.1039/C3IB40184G
    [64] Rabanel JM, Hildgen P, Banquy X (2014) Assessment of PEG on polymeric particles surface, a key step in drug carrier translation. J Control Release 185: 71–87. doi: 10.1016/j.jconrel.2014.04.017
    [65] Pasut G, Veronese FM (2012) State of the art in PEGylation: the great versatility achieved after forty years of research. J Control Release 161: 461–472. doi: 10.1016/j.jconrel.2011.10.037
    [66] Wang S, Xu T, Yang Y, et al. (2015) Colloidal stability of silk fibroin nanoparticles coated with cationic polymer for effective drug delivery. ACS Appl Mater Interface 7: 21254–21262. doi: 10.1021/acsami.5b05335
    [67] Tian Y, Jiang X, Chen X, et al. (2014) Doxorubicin-loaded magnetic silk fibroin nanoparticles for targeted therapy of multidrug-resistant cancer. Adv Mater 26: 7393–7398. doi: 10.1002/adma.201403562
    [68] Chung H, Kim TY, Lee SY (2012) Recent advances in production of recombinant spider silk proteins. Curr Opin Biotech 23: 957–964. doi: 10.1016/j.copbio.2012.03.013
    [69] Lammel A, Schwab M, Hofer M, et al. (2011) Recombinant spider silk particles as drug delivery vehicles. Biomaterials 32: 2233–2240. doi: 10.1016/j.biomaterials.2010.11.060
    [70] Humenik M, Smith AM, Scheibel T (2011) Recombinant spider silks-biopolymers with potential for future applications. Polymers 3: 640–661. doi: 10.3390/polym3010640
    [71] Schierling MB, Doblhofer E, Scheibel T (2016) Cellular uptake of drug loaded spider silk particles. Biomater Sci 4: 1515–1523. doi: 10.1039/C6BM00435K
    [72] Doblhofer E, Scheibel T (2015) Engineering of recombinant spider silk proteins allows defined uptake and release of substances. J Pharm Sci 104: 988–994. doi: 10.1002/jps.24300
    [73] Elsner MB, Herold HM, Muller HS, et al. (2015) Enhanced cellular uptake of engineered spider silk particles. Biomater Sci 3: 543–551. doi: 10.1039/C4BM00401A
    [74] Florczak A, Mackiewicz A, Dams KH (2014) Functionalized spider silk spheres as drug carriers for targeted cancer therapy. Biomacromolecules 15: 2971–2981. doi: 10.1021/bm500591p
    [75] Neubauer MP, Blüm C, Agostini E, et al. (2013) Micromechanical characterization of spider silk particles. Biomater Sci 1: 1160–1165. doi: 10.1039/c3bm60108k
    [76] Anselmo AC, Zhang M, Kumar S, et al. (2015) Elasticity of nanoparticles influences their blood circulation, phagocytosis, endocytosis, and targeting. ACS Nano 9: 3169–3177. doi: 10.1021/acsnano.5b00147
    [77] Numata K, Kaplan DL (2010) Silk-based delivery systems of bioactive molecules. Adv Drug Deliver Rev 62: 1497–1508. doi: 10.1016/j.addr.2010.03.009
    [78] Numata K, Subramanian B, Currie HA, et al. (2009) Bioengineered silk protein-based gene delivery systems. Biomaterials 30: 5775–5784. doi: 10.1016/j.biomaterials.2009.06.028
    [79] Numata K, Hamasaki J, Subramanian B, et al. (2010) Gene delivery mediated by recombinant silk proteins containing cationic and cell binding motifs. J Control Release 146: 136–143. doi: 10.1016/j.jconrel.2010.05.006
    [80] Numata K, Kaplan DL (2010) Silk-based gene carriers with cell membrane destabilizing peptides. Biomacromolecules 11: 3189–3195. doi: 10.1021/bm101055m
    [81] Numata K, Reagan MR, Goldstein RH, et al. (2011) Spider silk-based gene carriers for tumor cell-specific delivery. Bioconjugate Chem 22: 1605–1610. doi: 10.1021/bc200170u
    [82] Seib FP, Herklotz M, Burke KA, et al. (2014) Multifunctional silk-heparin biomaterials for vascular tissue engineering applications. Biomaterials 35: 83–91. doi: 10.1016/j.biomaterials.2013.09.053
    [83] Seib FP, Maitz MF, Hu X, et al. (2012) Impact of processing parameters on the haemocompatibility of bombyx mori silk films. Biomaterials 33: 1017–1023. doi: 10.1016/j.biomaterials.2011.10.063
    [84] Murphy AR, Kaplan DL (2009) Biomedical applications of chemically-modified silk fibroin. J Mater Chem 19: 6443–6450. doi: 10.1039/b905802h
    [85] Kambe Y, Yamamoto K, Kojima K, et al. (2010) Effects of RGDS sequence genetically interfused in the silk fibroin light chain protein on chondrocyte adhesion and cartilage synthesis. Biomaterials 31: 7503–7511. doi: 10.1016/j.biomaterials.2010.06.045
    [86] Teule F, Miao YG, Sohn BH, et al. (2012) Silkworms transformed with chimeric silkworm/spider silk genes spin composite silk fibers with improved mechanical properties. P Natl Acad Sci USA 109: 923–928. doi: 10.1073/pnas.1109420109
    [87] Xia XX, Qian ZG, Ki CS, et al. (2010) Native-sized recombinant spider silk protein produced in metabolically engineered Escherichia coli results in a strong fiber. P Natl Acad Sci USA 107: 14059–14063. doi: 10.1073/pnas.1003366107
    [88] Wray LS, Hu X, Gallego J, et al. (2011) Effect of processing on silk-based biomaterials: reproducibility and biocompatibility. J Biomed Mater Res 99: 89–101.
    [89] Rockwood DN, Preda RC, Yucel T, et al. (2011) Materials fabrication from bombyx mori silk fibroin. Nat Protoc 6: 1612–1631. doi: 10.1038/nprot.2011.379
    [90] Duncan R, Gaspar R (2011) Nanomedicine(s) under the microscope. Mol Pharm 8: 2101–2141. doi: 10.1021/mp200394t
    [91] Sheridan C (2012) Proof of concept for next-generation nanoparticle drugs in humans. Nature Biotechnol 30: 471–473. doi: 10.1038/nbt0612-471
    [92] Shi J, Kantoff PW, Wooster R, et al. (2017) Cancer nanomedicine: progress, challenges and opportunities. Nat Rev Cancer 17: 20–37.
    [93] Maeda H, Nakamura H, Fang J (2013) The EPR effect for macromolecular drug delivery to solid tumors: improvement of tumor uptake, lowering of systemic toxicity, and distinct tumor imaging in vivo. Adv Drug Deliver Rev 65: 71–79. doi: 10.1016/j.addr.2012.10.002
    [94] Duncan R (2006) Polymer conjugates as anticancer nanomedicines. Nat Rev Cancer 6: 688–701. doi: 10.1038/nrc1958
    [95] Juliano R (2013) Nanomedicine: is the wave cresting? Nat Rev Drug Discov 12: 171–172. doi: 10.1038/nrd3958
    [96] Venditto VJ, Szoka FC (2013) Cancer nanomedicines: so many papers and so few drugs! Adv Drug Deliver Rev 65: 80–88.
    [97] Wilhelm S, Tavares AJ, Dai Q, et al. (2016) Analysis of nanoparticle delivery to tumours. Nature Rev Mater 1: 1–12.
    [98] Cleal K, He L, Watson PD, et al. (2013) Endocytosis, intracellular traffic and fate of cell penetrating peptide based conjugates and nanoparticles. Curr Pharm Design 19: 2878–2894. doi: 10.2174/13816128113199990297
    [99] Duncan R, Richardson SC (2012) Endocytosis and intracellular trafficking as gateways for nanomedicine delivery: opportunities and challenges. Mol Pharm 9: 2380–2402. doi: 10.1021/mp300293n
    [100] Whitehead KA, Langer R, Anderson DG (2009) Knocking down barriers: advances in siRNA delivery. Nat Rev Drug Discov 8: 129–138. doi: 10.1038/nrd2742
    [101] Gratton SE, Ropp PA, Pohlhaus PD, et al. (2008) The effect of particle design on cellular internalization pathways. P Natl Acad Sci USA 105: 11613–11618. doi: 10.1073/pnas.0801763105
    [102] Herd H, Daum N, Jones AT, et al. (2013) Nanoparticle geometry and surface orientation influence mode of cellular uptake. ACS Nano 7: 1961–1973. doi: 10.1021/nn304439f
    [103] Rejman J, Oberle V, Zuhorn IS, et al. (2004) Size-dependent internalization of particles via the pathways of clathrin- and caveolae-mediated endocytosis. Biochem J 377: 159–169. doi: 10.1042/bj20031253
    [104] Oh P, Borgstrom P, Witkiewicz H, et al. (2007) Live dynamic imaging of caveolae pumping targeted antibody rapidly and specifically across endothelium in the lung. Nat Biotechnol 25: 327–337. doi: 10.1038/nbt1292
    [105] Sabharanjak S, Mayor S (2004) Folate receptor endocytosis and trafficking. Adv Drug Deliver Rev 56: 1099–1109. doi: 10.1016/j.addr.2004.01.010
    [106] Mosesson Y, Mills GB, Yarden Y (2008) Derailed endocytosis: an emerging feature of cancer. Nat Rev Cancer 8: 835–850. doi: 10.1038/nrc2521
    [107] Vercauteren D, Vandenbroucke RE, Jones AT, et al. (2010) The use of inhibitors to study endocytic pathways of gene carriers: optimization and pitfalls. Mol Ther 18: 561–569. doi: 10.1038/mt.2009.281
  • This article has been cited by:

    1. Victor Diogho Heuer de Carvalho, Thyago Celso Cavalcante Nepomuceno, Thiago Poleto, Ana Paula Cabral Seixas Costa, The COVID-19 Infodemic on Twitter: A Space and Time Topic Analysis of the Brazilian Immunization Program and Public Trust, 2022, 7, 2414-6366, 425, 10.3390/tropicalmed7120425
    2. Daniel Alejandro Rossit, Fernando Tohmé, Máximo Méndez-Babey, Mariano Frutos, Diego Broz, Diego Gabriel Rossit, Special Issue: Mathematical Problems in Production Research, 2022, 19, 1551-0018, 9291, 10.3934/mbe.2022431
    3. Victor Diogho Heuer de Carvalho, Thyago Celso Cavalcante Nepomuceno, Thiago Poleto, Jean Gomes Turet, Ana Paula Cabral Seixas Costa, Mining Public Opinions on COVID-19 Vaccination: A Temporal Analysis to Support Combating Misinformation, 2022, 7, 2414-6366, 256, 10.3390/tropicalmed7100256
    4. Hamed Karimian, Qin Fan, Qun Li, Youliang Chen, Juan Shi, Spatiotemporal transmission of infectious particles in environment: A case study of Covid-19, 2023, 335, 00456535, 139065, 10.1016/j.chemosphere.2023.139065
    5. Caitriona Murphy, Wey Wen Lim, Cathal Mills, Jessica Y. Wong, Dongxuan Chen, Yanmy Xie, Mingwei Li, Susan Gould, Hualei Xin, Justin K. Cheung, Samir Bhatt, Benjamin J. Cowling, Christl A. Donnelly, Effectiveness of social distancing measures and lockdowns for reducing transmission of COVID-19 in non-healthcare, community-based settings, 2023, 381, 1364-503X, 10.1098/rsta.2023.0132
    6. Germano B. dos Santos, Fabrício A. Silva, Thais R. M. Braga Silva, 2024, Análise do impacto da pandemia de COVID-19 na mobilidade no Brasil sob uma visão semântica, 155, 10.5753/courb.2024.3276
    7. Thyago Celso Cavalcante Nepomuceno, Miguel Gomes da Silva, Maria Eugênia Vergilio Mori, Wilka Maria do N. Silva, Isaac Pergher, Benchmarking non-pharmacological policies from an efficient administration perspective: a panel DEA approach with strategic insights for the post-pandemic, 2023, 0306-8293, 10.1108/IJSE-11-2022-0767
  • Reader Comments
  • © 2017 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(13919) PDF downloads(1885) Cited by(47)

Figures and Tables

Figures(5)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog